A customer asks an AI shopping assistant for a waterproof travel backpack under $180. It needs to fit a laptop, arrive by Friday and come with a returns policy that is easy to understand.

The customer may still make the final decision. But before they visit a website, read a review or see the product photography, the assistant may already have decided which options are worth showing them.

That is where the commercial shift starts.

Not autonomous checkout. Not a robot quietly buying the weekly groceries before breakfast.

The shortlist.

Earlier this year, I revisited an idea I first explored many years ago: the global shopper.

The technology had changed, but the pattern had not. Commerce keeps widening what customers can reach while forcing organisations to rethink how they are discovered, compared and trusted.

That article was about how AI was changing where customers looked.

This one is about whether AI can confidently understand what an organisation offers when it does.

The first shift was AI helping the shopper.

The next is AI doing the first comparison for them.

Before the customer arrives

A customer-side assistant can now take a broad need and begin narrowing the market.

It can compare products, apply price and delivery constraints, summarise reviews and identify trade-offs. It may decide which brands look credible, which offers are too uncertain and which options appear to fit the request best.

For the customer, that can be genuinely useful. It reduces searching and makes a complicated market easier to navigate.

For the organisation, it changes where the commercial journey begins.

Brands have spent years optimising for human attention through search, imagery, merchandising, reviews, offers and checkout design. All of that still matters.

But if an agent filters before the human arrives, the organisation also needs to be legible to software.

That brings us back to the backpack.

To place an offer on the shortlist, the assistant needs to establish some fairly basic things:

• Is it waterproof?

• Will it fit the customer’s laptop?

• Is the current price under $180?

• Is it in stock?

• Can it arrive by Friday?

• Is the returns policy clear?

None of those questions is especially difficult.

Answering them consistently often is.

What the agent needs to trust

People are good at working around ambiguity.

They interpret vague wording, read between the lines and sometimes forgive clumsy product data because they already know the brand or like the design.

An agent has less room for that kind of generosity.

If product attributes are incomplete, the advertised price does not match checkout, stock information is stale or the returns policy is buried in dense prose, the assistant has to decide how much confidence to place in the offer.

It may guess, add a qualification or move on to a competitor whose offer is easier to verify.

That is why the shortlist matters now, long before autonomous checkout becomes normal.

The organisation may not lose after a careful comparison.
It may simply never be included.

A brand can still persuade a person through photography, copy, reviews, service or familiarity. But in an agent-filtered journey, it may first need to pass a more functional test:

Can the offer be verified?

For the backpack, that means product attributes, current price, inventory, delivery timing, returns conditions and supporting customer evidence all need to tell roughly the same story.

If those signals conflict, the assistant has a confidence problem.

Soon enough, the customer does too.

This is not just a marketing issue

It would be easy to treat this as another content-optimisation problem.

It is not.

Machine legibility spans product, commerce, data, inventory, policy, technology and customer service.

The people who own product attributes, stock signals, eligibility rules, delivery promises and policy content may not think they are shaping brand discovery.

Increasingly, they are.

The commercial team may own conversion, but another team owns the inventory feed. Legal may own the returns policy. Operations may own fulfilment. Product may own the specification. Customer service may be left dealing with the gap when those promises do not line up.

That is the operating issue underneath agent-mediated commerce.

The organisation needs reliable answers to basic commercial questions: current enough to trust, structured enough to travel and clear enough for software to use without filling in the blanks.

Retail makes this visible first, but the same issue appears elsewhere.

An insurer’s eligibility rules, a university’s entry requirements or a utility’s service commitments may all be screened before a person reaches the organisation.

In each case, software is trying to work out whether the offer fits the need and whether the evidence supports the recommendation.

Run the shortlist test

Before building a large agentic-commerce roadmap, start with one real customer mission.

Use the backpack example or choose something more relevant to the organisation.

Then trace the evidence an external agent would need.

Where does the product or service attribute come from? Which price is current? How reliable is the availability or eligibility signal? Can delivery or fulfilment be verified? Is the returns, cancellation or dispute policy clear?

Then ask who owns each answer, how often it changes and what happens when channels disagree.

Only after that should you run the comparison.

Ask a customer-side assistant to compare the offer with several competitors. Then look beneath the result.

What did the tool get right? Where did it guess? What did it leave out? Did a competitor make the shortlist because its offer was stronger, or simply because its information was easier to understand and verify?

That distinction matters.

The audit may reveal an AI interpretation problem. It may also reveal that the organisation’s own commercial information is more fragmented than everyone assumed.

The response may involve clearer product data, stronger structured content, more reliable stock and delivery signals, simpler policies or clearer ownership between teams.

Those improvements will help human customers as well.

The investment question

A more immediate question for leaders may not be:

Do we need our own shopping agent?

It may be:

Can someone else’s agent verify our offer without guessing?

That is the more useful test.

Brand still matters. Customer experience still matters. The owned channel still matters.

But each becomes less valuable if the offer is filtered out before the customer gets there.

The organisations best placed to respond may not be those with the most ambitious agent strategy.

They are more likely to be those whose products, services and policies are clear enough to be understood by both people and machines.

Brand still matters once it gets seen.

Machine legibility may increasingly decide whether it does.

Executive note for leaders

Choose one real customer mission and ask:

Can a customer’s agent verify enough about our offer to place it on the shortlist?

Trace the attributes, price, availability, delivery promise and policy conditions back to their sources and owners.

Then run the comparison and inspect where the agent verifies, guesses or leaves the offer out.

The aim is not to optimise for every AI tool.

It is to make sure the organisation provides reliable enough evidence for software to understand what it is offering.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

In 2001, when I launched the small digital agency called Lava that later became Wolf Interactive, I had what felt like a bold idea. We would make our own website multilingual.

The internet was young enough that this felt exciting. A small Australian business could present itself to a wider market, and digital was becoming more than an online brochure.

In practice, the idea was also slightly mad.

Money was tight. Every change meant involving a translation specialist, updating duplicated content and spending more time than the budget justified.

The ambition was good. The operating model was not.

The multilingual website did not last.

That experience is one reason I find the current multilingual AI moment so interesting. A capability that once required translation providers, duplicated content and considerable patience can now appear inside a chatbot or live-chat interaction almost instantly.

That is genuine progress.

But the operating-model question has not gone away. It has simply moved further into the service.

When the conversation gets ahead of the service

AI can make the conversation multilingual before the organisation itself is ready to provide a multilingual service.

A customer can ask a question in Vietnamese, Arabic, Mandarin, Cantonese or Punjabi and receive a fluent response. A live-chat platform can translate both sides of an exchange. A contact-centre agent can receive an English summary.

For straightforward interactions, this can be very useful. Opening hours, appointment times, order status and simple service information are sensible places to begin. The consequence of a minor translation issue is usually limited and recoverable.

The difficulty starts when a fluent conversation is treated as evidence that the whole service is now multilingual.

Most services do not begin and end inside one chatbot interaction. They move through portals, forms, documents, payments, records, contact centres and human escalation.

A chatbot might explain a hardship process clearly, only to send the customer to an English-only application. It may describe an insurance claim accurately while the policy wording and dispute process remain difficult to understand. Or it may hand the case to an employee who receives a short English summary with much of the context removed.

I think of this as the channel cliff: the point where translated guidance ends and the customer is expected to continue through a service that has not changed with it.

The conversation has improved. The customer’s ability to finish the task may not have.

The failure often happens after the chatbot

The risk is not necessarily in the translated conversation. It is what happens next.

This is where performance reporting can become a little too reassuring.

Chatbot use is increasing. More languages are being selected. Containment is improving. The numbers look tidy in a dashboard.

All of that may be true.

But containment is not completion.

A closed conversation does not tell us whether the customer understood the next step, completed the application, challenged a decision or received the support they needed.

When the journey breaks after the chatbot, the organisation may simply move the cost into repeat contact, manual assistance, complaints or rework.

Low demand for interpreters can also be misleading. It may mean formal support is not needed. It may equally mean customers have stopped asking.

People often find another way through. A child helps a parent navigate a portal. A grandchild explains an aged-care form. A community worker walks another household through the same process for the third time that week.

That support can be generous and practical.

It should not quietly become the service model.

If an essential service depends on a teenager interpreting a financial, legal or health obligation, the organisation has transferred the risk to the household.

The real test is service completion

The first question most organisations ask is:

What languages do our customers speak?

It is a reasonable starting point, but it does not tell us whether the service is working.

A more useful question is:

What were customers who speak those languages actually able to complete?

That shifts the conversation from language availability to service outcomes.

Take one important journey, such as hardship support, eligibility, a complaint or an essential-service interruption, and follow it from the first multilingual interaction to the final result.

Where does English reappear? Where does context disappear during a hand-off? What does the customer need to understand, agree to or challenge? Can the organisation reconstruct what was explained and acted on if the outcome is disputed later?

Not every interaction carries the same consequence.

Checking an appointment time is different from agreeing to a payment arrangement. Asking where to upload a document is different from consenting to care or accepting a legal obligation.

AI translation is particularly useful when the consequence is low and any error can be corrected easily.

More care is needed when the journey involves money, access, safety, consent, dispute or essential support. In those situations, translation quality, human escalation, record keeping and accountability need to be part of the service design.

A practical place to begin

This does not mean every document, channel and service needs to be translated immediately. Most organisations will need to sequence the work.

Start with one consequential journey and identify where multilingual support begins and ends. Look at the customer outcome, not only the interaction.

Then decide which parts can sensibly rely on AI translation, where human language support is required and what information needs to travel with the customer between channels.

The goal is not a perfect multilingual service overnight.

It is to avoid creating a polished front door that leads customers into the same barriers that were already there.

AI has made multilingual interaction faster and far more affordable. That creates an important opportunity for organisations serving increasingly diverse communities.

The risk is mistaking a translated conversation for an inclusive service.

The practical test is whether the customer can understand, act, challenge and finish.

That is the point where multilingual capability becomes service capability.

Executive note for leaders

Before approving multilingual AI in a customer channel, ask what the customer can actually complete.

Follow the journey beyond the chatbot into the form, portal, hand-off, record, payment, escalation and final outcome.

Find where English reappears, where context is lost and where someone outside the organisation is quietly expected to finish the task.

A fluent conversation proves the front door can speak.

A completed service shows the organisation can support the customer through to the outcome.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

The first wave of enterprise AI is producing plenty of useful results. A team turns on a copilot. A service channel adds an assistant. A business function builds an agent to summarise, triage, route, draft or recommend.

The work gets faster. A queue moves. A pilot looks promising.

Those results matter, but they do not necessarily mean the organisation is building a platform.

The harder question is what remains after the first use case is finished.

Did it improve the data, clarify ownership, strengthen a workflow or put in place a control another team can use?

Or will the next use case require the same scramble for access, integration, approvals and change support?

If everything begins again from zero, the organisation may have bought a useful product.

It may not have built much capability.

The next use case is the real test

Copilots, assistants and agents are appearing inside enterprise software, service channels, knowledge bases, sales platforms, HR systems and operational workflows. Some are bought directly. Others arrive as features inside products the organisation already uses.

Some are built quickly because, quite reasonably, the work still needs to get done.

In many cases, this is exactly the right way to begin.

Local use cases allow teams to test value and understand the practical constraints before making a larger commitment.

But each use case should also answer a quieter question:

Does this make the next use case easier?

A product solves a local problem.
A platform leaves behind something reusable.

That might be a stronger knowledge base, a clearer data source, an agreed workflow, a tested privacy control or a better integration pattern.

The platform is not necessarily the product suite, dashboard or AI model.

It is the organisation’s growing ability to solve the next problem without repeating the same work.

Why local wins can become expensive

Product wins are easy to see.

A team can demonstrate time saved. A business unit can point to a working assistant. A vendor can show an agent moving confidently through a process, helped by the fact that demo environments rarely contain the full organisational mess.

The frontline team sees whether the tool connects to the system they use every day, whether the answer reflects the latest process change, and what happens when the assistant reaches a situation it was not designed to handle.

The CFO sees something else.

More licences. More vendors. More integrations. More productivity claims.

But not always evidence that the organisation now owns a capability it did not own before.

That is where AI investment starts to drift.

The value remains local while cost and complexity accumulate across the enterprise.

Architecture inherits another connector. Risk inherits another exception. Operations inherits another workflow. Procurement inherits another contract.

The next team inherits very little.

Over time, individually sensible decisions can add up to another layer of legacy complexity.

The lesson is familiar

We have seen versions of this before.

Customer 360, CRM and personalisation programs all promised better customer understanding, more relevant service and stronger commercial performance.

The promise was reasonable. The difficulty was all the work underneath it: data quality, identity resolution, consent, integration, ownership, workflow and adoption.

The same pattern is returning with new language: copilots, agents, orchestration, embedded assistants and systems of intelligence.

A useful tool can still be mistaken for reusable capability.

This is particularly important at the data and knowledge layer.

An assistant drawing from inconsistent sources will produce inconsistent answers with impressive fluency. An agent moving across an unclear process may simply make the fragmentation move faster.

The organisations that gain most from AI will not necessarily have the greatest number of tools. They will improve the foundations each time they introduce one.

What should each use case leave behind?

This does not mean every AI pilot should become an enterprise architecture program.

Please do not invite everyone to that meeting.

Small experiments matter. Teams need room to test ideas, learn quickly and stop when the value is not there.

But speed and reuse are not opposites.

A well-designed use case can deliver a near-term result while improving the organisation’s ability to deliver the next one.

Before approving the next assistant, agent or copilot rollout, I would want to know four things.

Has the data or knowledge source improved?

Is the information more reliable, better governed and easier for another workflow to use?

Is ownership clearer?

Is it clear who maintains the content, owns the decision and supports the workflow once the pilot team moves on?

Is there now something another team can reuse?

That might be an integration pattern, escalation pathway, governance control or repeatable approach to change and adoption.

Will the next use case be easier because of the work already done?

Will the second team move faster, with less cost and uncertainty, because the first team left something useful and reusable behind?

If the answer is yes, the organisation may be making platform progress.

If not, the use case may still be worthwhile. A local tool that saves time, improves service or removes frustration can create meaningful value.

There is nothing wrong with a product win.

The problem begins when every successful pilot is described as enterprise progress.

The leadership decision

A product win proves that something worked in one place.

Platform progress means the organisation is becoming better equipped for what comes next.

Leaders need a clear view of which initiatives are delivering a useful local result and which are genuinely strengthening enterprise capability. Both can be funded, but they should not be described or valued in the same way.

The more durable return on AI is unlikely to come from the number of pilots launched or licences activated.

It will come from how much less effort, cost and organisational friction is required to deliver the next valuable use case.

That is the product-to-platform shift.

Not buying a larger suite.

Building an organisation that does not have to start again every time.

Executive note for leaders

Before approving the next AI use case, ask:

What will this leave behind?

Look for practical residue: better data, clearer ownership, a reusable workflow, a tested governance control, a stronger knowledge base or a faster second use case.

A product win proves something worked once.

Platform progress proves the organisation is becoming better equipped for what comes next.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Reputation used to feel like something an organisation accumulated over time: a pattern of delivery, a memory of good service and a general sense that it did what it said it would do.

That still matters. What has changed is the speed at which people now form a view.

A customer thread starts moving. A support forum fills in some of the gaps. A review is quoted elsewhere. An employee post adds context. Then an AI search tool or answer engine pulls those signals together and presents a confident summary of the organisation.

The organisation may still be working out what happened while the market is already reading the answer.

That creates a different kind of leadership problem. It is not simply about responding faster or producing better messaging.

The harder task is making sure the organisation can stand behind the version of itself that customers, employees and the market are already encountering.

Reputation now forms from the evidence trail

AI search and answer engines do not wait for an organisation to complete its internal alignment.

They assemble what is available.

A review site, an employee forum, a support thread, an old product page and a current help article can all appear in the same answer. The contradictions may not be identified as contradictions. They can be smoothed into a single explanation that sounds more settled than the evidence really is.

Sometimes the result is fair. Sometimes it is incomplete, stale or more confident than it should be.

Either way, it is usually built from a much broader evidence trail than the organisation’s website or approved statement.

The homepage says one thing. The support forum suggests another. The product experience creates a third impression. Employee commentary adds something else again.

From inside the organisation, these may look like separate ownership areas. From outside, they look like the same organisation.

AI does not see a complicated governance model. It sees source material.

That is why reputation is becoming less about controlling a message and more about reducing contradictions across decisions, channels and experiences.

Where organisations get caught

Most organisations still align sequentially.

Product is checking the roadmap. Legal is reviewing the wording. Customer teams are working out what the change means for service delivery. Employees may know something is coming, but not enough to answer questions confidently.

Meanwhile, the market is not waiting for that process to finish.

A product change can go live before the support model is ready. A pricing decision can land before customers understand the trade-off. An AI-enabled service can provide an answer that is technically explainable but difficult to defend publicly. A platform migration may make sense internally while customers experience it as lost functionality.

The 2024 Sonos app redesign is a useful example. What may have been intended as a platform and product reset was experienced by many customers through missing basics, disruption and reduced confidence.

The same pattern appears in banking migrations, utility channel changes, government service redesigns and AI-enabled support.

The original decision may have been rational. What can be underestimated is how quickly it will be interpreted through the customer experience, frontline response, employee commentary and public evidence around it.

A sensible internal decision can still produce an incoherent external experience.

Once those signals begin circulating, the organisation is no longer explaining the decision on its own timetable.

What coherence requires

This is where reputation becomes more than a communications issue.

A communications team can make an organisation sound clear for a while. It cannot make an incoherent operating model feel trustworthy indefinitely.

A product team can be technically right and still miss the lived customer experience. An executive team can have sound strategic intent and still underestimate what people will infer from timing, tone or silence.

Reputation coherence is therefore narrower and more practical than brand control. It is about reducing the distance between the decision that was made, the action the organisation took, the explanation given to customers and employees, and the experience people actually had.

Speed matters, but speed and coherence are not the same thing.

A fast response can add another contradiction. A slower response can arrive after the market has already settled on its own explanation.

Coherence also has a cost. It may mean delaying a launch, absorbing extra support demand, briefing employees earlier or accepting some short-term commercial friction so frontline teams are not left defending a decision they cannot properly support.

That is not excessive reputational caution. It is part of making a significant change operationally credible.

In an AI-mediated environment, inconsistency can travel further than intent. If customers, employees, service channels and digital evidence point in different directions, the formal statement becomes only one input among many.

Who owns the next 12 hours?

Not every reputation issue is a crisis.

Treating each one as a crisis usually creates hierarchy, theatre and a considerable number of calendar invitations.

Many situations are simply moments where public interpretation is moving faster than internal alignment.

In those moments, the useful question is:

Who owns the next 12 hours?

Not the whole reputation system. Not a new permanent committee. The immediate window in which uncertainty can either be reduced or allowed to harden into the story.

I would want a small group to establish a few things quickly.

What are customers actually experiencing now, rather than what the release plan says they should be experiencing?

Are the website, chatbot, contact centre and support teams giving customers the same explanation?

Do employees know enough to answer basic questions, or are they filling in the gaps themselves?

What are reviews, search results, forums, social posts and AI-generated answers already surfacing?

And what can the organisation say now that it will still be able to stand behind tomorrow?

The goal is a clear, supportable position that does not need to be quietly reversed once the internal picture catches up.

Reputation coherence does not require every decision to be centralised. It requires clarity about which issues need enterprise judgement, which can be handled locally and which need a simple line that customer and employee teams can use immediately.

Some uncertainty is inevitable.

Allowing every part of the organisation to improvise its own version is not.

The organisations that manage this well will not necessarily be the loudest or most polished. They are more likely to be the ones whose customers, employees and markets encounter roughly the same organisation wherever they touch it, including when an AI system gets there first.

Executive note for leaders

Before treating a reputation issue as a communications problem, run a fast coherence check.

Look across the customer experience, service channels, employee understanding, legal or policy position, search visibility and public messaging.

The aim is not perfect certainty.

It is to reduce contradiction before the market, or an AI-generated summary, turns the organisation’s loose ends into the story.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Springsteen is a Border Collie, which means he combines high intelligence, unreasonable enthusiasm, and the quiet operational belief that everyone else in the household is underperforming.

This becomes especially clear around evening walk time.

If the walk has not commenced within what he considers the acceptable delivery window, he begins stakeholder engagement: eye contact, pacing, strategic sighing, toy placement, and eventually the sort of stare that suggests the roadmap has failed.

The interesting thing is that Springsteen is not especially moved by intent. He does not care that the walk was planned, that the household had competing priorities, or that the delivery team needed five minutes to get organised before mobilisation.

He responds to observable signals: shoes, keys, movement, tone and timing.

Which, unhelpfully, makes him a fairly good market analyst.

There is an exception, of course. If you are a technology company, some customers will analyse your roadmap with the seriousness of an intelligence briefing. Springsteen does something similar when someone picks up a cup of tea during evening-walk hour. To him, that is not a beverage. That is a signal.

But for most organisations, customers are not studying the roadmap. They are responding to what they can see, feel, use, resolve, buy, access or trust today.

They do not usually know what was discussed in the quarterly planning session, which initiative is in phase two, or that a dependency has been moved into the next release window.

And, in most cases, they do not care.

It is just reality doing what reality does.

A roadmap is comforting because it turns uncertainty into sequence. It gives leaders something structured to point at. It shows intent, ambition and order. Done well, it is useful. Necessary, even.

But the roadmap is not the customer experience, the employee experience or the market signal — it is a theory of movement. Reality is the feedback loop.

Springsteen understands this deeply, although I suspect he would describe it differently, possibly by placing a tennis ball on someone’s foot and maintaining unsettling eye contact.

His view is simple: if the signals suggest walk, then walk should happen. If the signals suggest movement, but no movement follows, confidence begins to fall. If the pattern repeats, escalation becomes reasonable.

In organisational life, the same thing happens more quietly.

A company announces a refreshed customer strategy, but the contact centre still cannot see the right information. A new digital channel launches, but customers still have to repeat themselves when they move to a person. A transformation program says the organisation is simplifying, while employees experience more tools, more meetings and more internal friction.

No one is necessarily being dishonest. The roadmap may be real, the intent may be good, and the sequencing may even make sense internally.

The roadmap explains what we meant to do. The market responds to what it can see.

That does not mean leaders should abandon roadmaps and chase every signal that moves. That would be exhausting. Springsteen would probably support it, but he is not the Chief Operating Officer.

The more useful reminder is smaller: leaders need to hold the roadmap lightly enough to notice when reality has started giving feedback.

Sometimes the signal is customer behaviour. Sometimes it is frontline stress, reputation moving faster than expected, adoption lagging behind a product everyone believed was ready, or a Border Collie standing in the hallway, making it abundantly clear that the current operating model lacks credibility.

The leadership work is not to throw away the plan every time the room changes; it is to notice when the room has changed. Because the market does not care that the roadmap was carefully considered. It cares what is landing.

And if what is landing does not match what was intended, the useful response is not to point harder at the roadmap. It is to read the signal, adjust the work, and keep moving.

Preferably before the Border Collie escalates.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

There is a question that often gets skipped after a successful AI pilot: Who approved the knowledge the assistant is using, and who keeps it current when the pilot team moves on?

During a pilot, most things remain close enough to manage. The knowledge base is limited, the scenarios are familiar and the project team is still around to notice an awkward answer before it becomes awkward in public.

Production is less forgiving.

Policies change. Service definitions shift. Customers arrive with disputes, hardship issues, accessibility needs, language barriers and circumstances that have shown very little interest in the pilot plan.

Pilot success shows that the capability can work. Production rollout shows whether the organisation is ready to stand behind it.

Those are different tests.

What changes in production

The specific capability here is not a general chatbot drawing answers from the open internet.

It is a grounded AI assistant: a system that generates responses from approved organisational sources such as policy libraries, service guides, hardship procedures, complaint pathways, internal knowledge articles and customer-service scripts.

In practice, it might support a contact-centre employee, help a chatbot answer more consistently, guide a customer through an after-hours issue or summarise the next best step from approved material.

For high-volume service environments, the appeal is obvious. Responses can be faster and more consistent, with less dependence on every frontline employee remembering every policy variation on a difficult Tuesday.

But grounding an assistant in approved content does not settle the leadership question. It brings the question into sharper focus.

Once the assistant begins turning organisational knowledge into customer-facing answers, the organisation is no longer simply managing documents. It is operationalising its policies and promises.

Where the operating model gets exposed

A grounded assistant is only as dependable as the knowledge it is allowed to use.

That sounds straightforward until a hardship policy changes and the source article is not updated. Or the website reflects a new complaint pathway while the internal knowledge base still carries the old one. Or a customer asks a question that is technically answerable, but sensitive enough that a human should be involved.

In each case, the assistant may be doing exactly what it was designed to do.

The model can be grounded while the organisation around it is not.

Customers do not see the retrieval process, confidence score or internal debate about whether the pilot was ready for production. They experience the organisation answering them.

If the response is wrong, outdated or poorly escalated, the issue is no longer confined to model performance. It becomes a service, governance and accountability problem.

It can also create very ordinary operating costs: repeat contact, complaints, manual correction, employee workarounds and senior attention being pulled into issues that should have been resolved in the service design.

This is one reason AI initiatives often slow down between pilot and rollout. The technology may have worked, but production forces the organisation to decide what it is prepared to own.

Knowledge becomes part of the service

A grounded assistant can retrieve, draft, summarise and respond. It can make language more consistent and reduce the time employees spend searching for information.

It can also expose gaps in policy, content ownership and channel governance that have been sitting quietly in the background because no system had tried to apply them at scale.

That makes it useful in another way: it is diagnostic.

Once organisational knowledge becomes response logic, content governance is no longer administrative housekeeping. It becomes part of the operating model.

An answer about hardship eligibility may shape what a customer believes they can access. Guidance in a portal may influence whether someone escalates a matter, accepts an outcome, complains, waits or gives up.

The production question therefore sits one level above whether the answer looks correct.

It is whether the organisation can continue to own that answer when the policy, customer circumstances or service context changes.

The ownership test

Before moving a grounded AI assistant from pilot to production, I would want three things made explicit.

Who owns the knowledge?

Not the vendor, the project team or “the business” in the abstract.

The actual person or function accountable for the policy, service rule or customer commitment being expressed.

That owner also needs a workable process for reviewing and updating the source. A named owner without time, authority or a maintenance rhythm is really just a name in a governance document.

Where must the system stop?

Some answers can safely go directly to a customer. Others should support an employee rather than replace their judgement.

Hardship, vulnerability, complaints, disputed charges, accessibility needs and unclear policy are clear examples where escalation needs to be designed before launch.

The assistant should not discover its boundaries one customer at a time.

Can the answer be traced later?

If a customer challenges a response six months from now, the organisation should be able to establish what source the assistant used, what it produced, whether the source was current and what happened next.

Not every interaction requires the same audit trail. But where an answer affects money, access, rights, safety or a formal decision, traceability should not depend on someone finding an old screenshot.

Before approving rollout

This does not require a 47-page governance framework that everyone politely agrees is important and no one can quite apply.

A practical production-readiness check is more useful.

Take the highest-consequence scenarios the assistant is expected to handle and work through them with the people who own the policy, service, risk and frontline response.

Confirm which source is authoritative, who keeps it current, when the assistant can answer directly, when it must hand over to a person and what needs to be recorded when the issue carries material consequence.

If those answers remain vague, the organisation may not yet have a production system.

It may have a pilot wearing a production badge.

There is nothing wrong with extending the pilot while those questions are resolved. The greater risk is moving into production because the demonstration was successful and discovering later that the organisation cannot support the service it has created.

Confidence is not about assuming nothing will go wrong. It comes from knowing what the organisation is prepared to own, where judgement sits and who acts when the context changes.

That is the difference between good enough to ship and safe enough to sleep.

Executive note for leaders

Before a grounded AI assistant moves from pilot to production, ask:

Who owns the knowledge, where must the system stop, and can the answer be traced later?

If those questions do not have clear owners and workable processes behind them, the organisation is still carrying pilot risk into production.

The practical test is not simply whether the assistant can answer.

It is whether the organisation is ready to stand behind the answer when the situation changes.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

The productivity case for AI copilots is easy to see. A briefing note appears faster. A contract can be summarised in minutes. Research is pulled together before the meeting finishes. A senior person can ask for a comparison, a first draft and a tidy follow-up email without waiting for someone junior to assemble it all by hand.

That is useful, and in many cases overdue.

But there is a quieter question underneath the efficiency gain:

What if some of the work organisations are most eager to remove is also where people used to learn?

Not the glamorous work. The rough first draft. The overlong briefing note. The document review where someone slowly learned what mattered and which detail a senior colleague would ask about immediately.

AI is beginning to absorb more of that work.

The organisation may get the output faster while weakening the pathway that produced its future experts.

Which would be an impressive little productivity surprise.

When first-pass work disappears

Embedded workflow AI can now draft, summarise, compare, analyse and review material inside the tools people already use.

In some workflows, it produces the first pass, the messy pass and sometimes most of the finished result.

That matters because first-pass work has never been only about producing an artefact. It has also been part of the apprenticeship.

People learn to write a useful brief by first writing an unhelpful one. They learn to separate evidence from noise by working through too much of both. They learn what a weak recommendation looks like when someone asks them to defend it.

The feedback was not always delivered with the care now recommended in leadership programs. Many of us have the tracked changes to prove that.

But the repetition built pattern recognition, judgement and professional instinct.

Why the risk is easy to miss

Most organisations are trying to move faster, lower cost, improve consistency and make better use of experienced people. All reasonable objectives.

The productivity gain appears immediately. Work is completed faster, junior employees produce more polished material earlier, and senior employees spend less time correcting basic drafts.

The capability loss, if it occurs, appears later.

A junior employee reviewing an AI-generated analysis may become faster at improving the output. They may not develop the same understanding of how the evidence was gathered, which alternatives were discarded or why one trade-off mattered more than another.

We may end up asking people to check work they have never properly done themselves.

Then, several years later, the organisation wonders why the mid-career bench is thinner than expected, why senior specialists are still being pulled into routine decisions, and why succession plans appear stronger in the spreadsheet than they do in the room.

Where judgement is really built

This is not mainly an argument about preserving entry-level jobs. It is about how organisations continue to produce judgement.

Early-career roles have always done two things: helped produce today’s work and helped produce the people who will handle harder work tomorrow.

The first is easy to measure. The second becomes visible when it has gone missing.

That is why the productivity discussion can become too narrow. It asks how much faster a task can be completed, but not always what developmental value was hidden inside it.

The slow draft, the repeated review and the analysis rebuilt after a senior colleague asked one annoying but entirely correct question were not simply delays. They were how people learned to see the problem properly.

No one needs to preserve busywork out of nostalgia. There is plenty of low-value work that AI should remove.

But leaders need to distinguish between removing waste and removing practice.

Copying information between systems may teach very little. Reviewing customer cases, comparing patterns and explaining what changed may teach a great deal.

If both disappear without understanding the difference, the organisation may be borrowing capability from its own future.

Redesign the pathway, not just the task

Before compressing early-career work with AI, I would start with a practical question:

Where does judgement currently form?

Not where the learning framework says it forms. Where it happens in the real work.

Choose one role, function or career pathway and look at the tasks AI is beginning to absorb: drafting, analysis, research, document review, meeting preparation, case summaries or contract triage.

Then separate the work into two broad categories.

The first is genuine waste: repetitive activity with little learning value that can be removed without weakening the pathway.

The second may also be repetitive, but it builds pattern recognition, context, trade-off judgement or confidence in handling ambiguity.

That work does not need to remain unchanged. It needs to be redesigned.

A junior employee might prepare an initial view before using AI to improve it. They might explain what the AI missed or why they rejected part of its recommendation. Teams can use AI output as the start of a review conversation rather than the finished answer.

People may also need exposure to difficult, incomplete cases before they are asked to supervise polished machine output.

The goal is not to slow the organisation down for the romance of apprenticeship. It is to make sure efficiency does not remove the experiences people still need to become capable.

That is why AI adoption, workforce planning and learning design should not sit in entirely separate conversations. If one team is redesigning work while another is responsible for capability development, someone needs to connect the two.

The point is not to preserve every junior task. It is to understand which work is producing more than output.

Where a task helps someone interpret evidence, manage uncertainty, recognise risk or recover when the first answer is wrong, design another way for that learning to happen.

AI should help people reach capability faster. It should not allow organisations to skip the process through which capability is formed.

The quiet risk is not that employees become more productive.

It is that the organisation stops noticing where expertise used to be made.

Executive note for leaders

Before removing a task with AI, ask:

Is this task only producing output, or is it also producing judgement?

If it is mainly waste, automate it carefully.

If it also builds pattern recognition, context or professional instinct, redesign the development pathway before the task disappears.

Identify where first drafts, analysis and review are being compressed, then decide how future experts will still get the repetition, exposure and feedback they need.

Organisations do not simply inherit expertise.

They create the conditions in which it develops.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Spend enough time around a border collie and you realise something fairly quickly. They’re always watching you. Not in a demanding way. More in a quietly attentive way, as if they’re constantly checking in on what you might be about to do next.

My border collie Springsteen does this every morning. Before work, he’ll sit near the door, looking at me with that very particular collie expression that seems to say:

So… are we doing something interesting today?

Border collies are remarkable animals. But what makes them remarkable isn’t just intelligence. It’s the relationship they build with their person.

They read tone.

They notice small changes in movement.

They pick up signals you didn’t realise you were sending.

And over time, something interesting happens.

You start reading them too.

Without thinking about it much, you develop a kind of quiet coordination. A glance, a shift in posture, a change in pace, and both of you know what’s happening next.

It’s less like giving instructions and more like being a small two-person team that understands each other.

Recently, while studying Organizational Leadership at Harvard Business School Online, a theme kept surfacing: as leadership scales, it becomes less about authority and more about calibration.

People sometimes assume leadership at the senior level becomes more about command - clearer direction, stronger decisions, sharper strategy.

And those things absolutely matter.

But what colleagues consistently describe when they talk about the leaders they trust most are slightly different qualities: empathy, presence, curiosity and clarity of direction.

And perhaps most importantly - the ability to genuinely listen - to catch the signal before the noise.

In other words, leadership isn’t just about directing energy.

It’s about tuning into it.

Which brings me back to Springsteen.

Border collies don’t work well with constant instruction. In fact, if you try to over-control them, things tend to fall apart pretty quickly.

What works much better is awareness.

You pay attention.

They pay attention.

And somewhere in that space, a kind of rhythm forms where both of you move in the same direction.

The longer I work in organisations, the more I suspect the same thing is true there as well.

The strongest teams rarely form because someone is simply issuing orders from the front.

They form because people feel understood. Because trust builds. Because the leader creates enough space for the unspoken signals to be noticed.

Which means leadership, at its best, is a little less like command…

…and a little more like a really good border collie partnership.

I sometimes joke that I’m Springsteen’s GM.

But in reality, he’s a constant reminder that leadership often starts with awareness.

And on most mornings, watching how closely he reads the world around him, I’m reminded just how powerful paying attention can be. 🐾

Leadership, it turns out, often starts with paying attention.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

The interesting problem with AI measurement is that most organisations are not short of data. They are short of evidence.

They can count licences, logins, prompts, active users, pilot numbers and reported time savings. They can show adoption dashboards and point to teams using copilots, assistants and AI-enabled workflow tools across the business.

There is usually no difficulty producing a slide that proves activity is increasing.

The harder question is whether any of it changed something the organisation genuinely cares about.

Did a customer issue get resolved faster? Did a manager make a better call? Did a team reduce rework, catch a risk earlier or move through an approval with more confidence?

That is the measurement gap sitting underneath a lot of AI strategy in 2026.

The practical question is not simply how much AI is being used. It is what AI is improving, who owns the outcome and what evidence would make the result credible.

Activity is easy to see

AI is now embedded across everyday work.

Enterprise copilots, generative AI assistants and workflow tools are helping people draft, summarise, analyse, compare, search, recommend and decide faster.

That creates plenty of visible activity.

A team generates more summaries. A pilot reduces drafting time. A workflow produces recommendations. Usage rises, the line on the dashboard moves up and to the right, and everyone feels a little better about the investment.

Sometimes that confidence is justified.

But activity and value are not the same thing.

A tool can be popular while having only a loose connection to the decision it was meant to improve. A team can save time on drafting while the approval still waits in an executive inbox for two weeks. A model can produce a useful recommendation that nobody trusts enough to act on.

The pilot may have improved one part of the process without changing the outcome that justified the investment.

That is why another dashboard of prompts and licences will not solve the measurement problem. Leaders need a clearer view of what happens between AI-assisted work and the decision, service outcome or risk position that follows.

Where value gets lost

Adoption data is useful, particularly early on.

Licence activation, repeat use, pilot participation and feedback from early adopters all help show whether people are engaging with the capability. None of that should be dismissed.

But those measures do not answer the question a board, CFO or business owner will eventually ask:

What changed?

That question tends to make the room slightly quieter.

Often, the value is there. A customer case was handled with better context. A manager had a clearer view before making a call. Analysis that once took days took hours.

But if the organisation cannot connect that improvement to a decision, outcome or accountable owner, the value story becomes difficult to defend.

The dashboard may be accurate and still measure the wrong thing.

It can tell you who used the tool, how often they used it and perhaps how much time they believe they saved. It may not tell you whether the work improved, whether the decision changed or whether the benefit remained after the novelty wore off.

At that point, the organisation may be measuring attendance rather than value.

Shadow AI makes the picture less complete. Useful work may be happening in tools the organisation cannot see, while official reporting gives a very precise account of only the environment it can.

Name the decision, not the tool

The deeper issue is usually ownership.

One team buys the licence. Another runs the pilot. Someone reports adoption. A business area is expected to realise the benefit.

When scrutiny increases, everyone can explain their part, but no one can quite explain the outcome.

That is not only a measurement problem. It is an operating-model problem.

A useful shift is to name the decision or outcome AI is meant to improve.

If the ambition is better customer retention, faster triage, sharper pricing, stronger forecasting or more consistent service recovery, the relevant business leader needs to own that result.

The tool owner can report usage. The business owner needs to explain whether anything important changed.

AI creates value when it helps someone act sooner, make a better judgement, avoid rework, reduce risk or improve a service outcome.

Without a named decision and an accountable owner, it can become another layer of activity across an already complicated organisation: more tools, more dashboards and more governance meetings, without much more confidence.

A practical way to measure it

The answer is not to build a perfect enterprise-wide AI measurement framework before the next steering committee.

That way lies a long workshop and a very tired room.

Start with one decision stream.

Choose something repeatable and close enough to the operating model to show what is really happening: a customer escalation, a pricing approval, a claims decision, a field response, a campaign allocation or a workforce-planning call.

Then work through four questions.

How was the decision made before AI?

Establish the original process, including the time, hand-offs, rework, escalation and judgement involved.

Where is AI now contributing?

Be precise about whether it is drafting, summarising, recommending, prioritising or changing the information available to the person making the call.

What changed?

Look at speed, quality, rework, escalation, risk, confidence and the customer or business outcome. Not every measure will matter in every case.

Who owns the result?

Name the leader or function accountable for deciding whether the change is valuable enough to continue, scale or stop.

Perfect attribution is unlikely. In most organisations, AI will be one influence among several.

What matters is a credible line between the activity and the consequence.

Some use cases will prove valuable. Others will show that the real bottleneck was elsewhere in the process. Both are useful findings.

Progress is more likely when organisations stop treating AI consumption as the outcome and start connecting it to decisions the business already cares about.

Executive note for leaders

Before scaling an AI pilot, dashboard or licence rollout, ask:

What decision is this meant to improve?

If the answer is unclear, the organisation may be measuring activity rather than value.

Name the decision, assign the owner and agree on the evidence that would show whether AI changed the result.

Measure the decision, not the tool.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Years ago, I wrote a feature for Desktop magazine about the global shopper: a customer who crossed borders, compared at speed and expected digital commerce to keep up.

The channels have changed, but the shopper’s orientation has not. Customers have continued to become better informed, less patient with friction and more willing to compare across brands, markets and platforms.

Now the journey is shifting again.

It may still end on your website, in your app, in your store or with your service team. But part of it may begin inside a generative AI tool, answer engine or AI browser that has already formed a view before the customer arrives.

The customer is no longer only searching. They are asking a tool to interpret the market for them.

That tool may read product pages, reviews, delivery promises, returns policies, support content and third-party commentary before deciding which options appear credible enough to mention.

It is always nice to meet the customer after someone else has explained you first.

The journey now starts earlier

AI-mediated discovery allows customers to begin with a need rather than a brand or category.

They might ask:

I need something reliable for this use case. I care about price, delivery, returns and support. What should I consider?

The tool may compare specifications, summarise reviews, interpret policies and identify trade-offs. It may also decide which brands are worth including and which appear too difficult, expensive or uncertain.

Sometimes the result will be useful. Sometimes it will be incomplete, stale or more confident than the evidence supports.

Either way, it may shape the customer’s first understanding before the organisation’s own digital experience has had a chance to do much at all.

The website is not dead. Most customers will still need an owned channel to explore the detail, check availability, make a purchase or complete the service.

But the website may no longer be where the first meaningful interpretation is formed.

That changes the role of the owned experience. It may need to confirm, correct or expand an understanding formed somewhere else.

Where the mismatch appears

Most customer journeys are still designed around the person who arrives.

Organisations optimise imagery, content, navigation, offers, comparison tools, service prompts and checkout flows. That work remains important.

The difference is that the customer may now arrive carrying an interpretation of the organisation already.

They may believe the product is best suited to a particular use case. They may have a view about whether the price premium is justified, how easy the returns process will be or whether support appears reliable.

The organisation may not have written that interpretation, even if it supplied some of the source material.

A generative AI tool may describe a slightly different promise from the one the organisation believes it made. A condition in the returns policy may disappear from the summary. A premium product may be reduced to “similar, but more expensive.” A service may be recommended for a need it was not designed to support.

The customer then arrives expecting the AI-generated version, and the website, store or service team inherits the gap.

That can produce abandoned journeys, avoidable service contact, complaints, returns and pressure on frontline teams to explain expectations they did not create.

Being visible is useful. Being confidently misrepresented is not.

The organisation needs to be legible

For a long time, organisations could assume they would have an opportunity to explain themselves. The customer would arrive, browse, compare and decide, and the brand could shape at least part of that understanding through the experience it controlled.

Some of those moments are now moving upstream.

The question for leaders is whether the organisation is clear and consistent enough to be represented reasonably when it is not in the room.

Not perfectly. No organisation can control every external summary, review or answer.

But the underlying evidence should be strong enough that customers are not sent into the journey with the wrong expectation.

This makes some previously unglamorous work much more important.

Product data is not just back-office hygiene. Policy content is not fine print that can be revisited eventually. Delivery commitments, support promises and eligibility conditions cannot quietly vary between the product page, help centre and contact-centre script.

In an AI-mediated journey, those inconsistencies become source material.

If the source material is fragmented, the interpretation is likely to be fragmented too.

The organisation may think it is managing a customer journey when the customer is actually arriving with an interpretation formed somewhere else.

That is now part of customer experience.

Start by observing, not optimising

The practical starting point does not need to be a large AI-discovery strategy.

Choose one product category, service journey or customer need and ask several major generative AI tools the questions a real customer would ask.

Not brand-approved questions. Real ones.

Which option is best for my situation? What should I watch out for? Which provider has the clearest returns policy? Why is this brand more expensive? Which service appears most reliable?

Then compare the answers with the organisation’s actual product data, policies, service promises and frontline reality.

I would want to know what the tools consistently get right, what they flatten or misunderstand, which sources they appear to rely on and whether the expectation created can be supported once the customer arrives.

The first step is observation.

I would not rush straight into optimisation, new content production or another technology purchase. Start by understanding the gap between what the organisation believes it is communicating and what customers are being told elsewhere.

That audit is relatively inexpensive.

The harder part may be reading the results without immediately explaining why the tool is wrong.

Some issues will sit in the AI-generated interpretation. Others will reveal that the organisation’s own data, policies or service messages were never as consistent as everyone assumed.

Both findings are useful.

The response may involve clearer product information, stronger policy content, better structured data, improved third-party listings or a more deliberate handover into owned and human channels.

The point is not to control every answer. That is neither realistic nor a particularly good use of an executive team’s time.

The aim is to make the organisation easier to understand and harder to misrepresent.

The global shopper has changed again.

This time, the customer may not simply arrive informed. They may arrive with expectations shaped by a version of the organisation that the organisation did not create.

Executive note for leaders

Ask a practical question:

What is GenAI telling customers before they reach us?

Choose one product category, service journey or customer need. Ask the questions a real customer would ask, then compare the answers with your product data, policies, service promises, support content and frontline reality.

Look for what is being flattened, omitted or misunderstood, and whether the organisation can support the expectations being created.

The aim is not to control every AI-generated answer.

It is to make sure the organisation is clear and coherent enough to be represented fairly before the customer arrives.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Many organisations may already have the beginnings of an AI strategy. It may simply be taking shape somewhere leadership has not looked yet.

It is sitting in browser tabs, AI note-takers, unofficial prompt libraries, pasted documents, personal subscriptions and small workarounds people are using to get through the day.

In one team, someone uses ChatGPT to summarise a policy because the internal search is hopeless. In another, an AI note-taker appears because meeting actions keep disappearing. A useful prompt library starts circulating in Teams before the official training arrives. A personal tool sits open beside the approved one because, awkwardly, it gives the better answer.

That does not mean the organisation is full of reckless rule-breakers. Usually, it means people are trying to solve the work in front of them.

Quiet AI use is not always the problem. Sometimes it is evidence of where the problem sits.

The strategy may already be forming

The capability here is not one platform.

It is the easy availability of copilots, public generative AI tools, meeting assistants, research tools, prompt templates and lightweight automations that people can use before formal strategy catches up.

That changes the starting point for leaders.

AI strategy no longer begins when the board approves a roadmap, IT selects a platform or the steering committee signs off an acceptable-use policy. By then, the organisation may already have developed a shadow operating model around the work.

These workarounds are rarely random. They tend to appear where the organisation is slow, unclear or poorly supported.

Weak knowledge access creates unofficial summarising. Too many meetings create silent note-takers. Slow review pathways encourage AI-assisted drafts. Confusing policy creates personal judgement calls. Limited training produces prompt libraries built by whoever happened to work it out first.

Some of this activity will be valuable. Some will be risky. None of it should remain invisible to leadership.

Start with the risk, but do not stop there

Leadership often encounters quiet AI use through the risk lens.

That is reasonable.

Customer data entered into a public tool is not a charming innovation story. A meeting bot recording a sensitive conversation without consent is not evidence of cultural progress. An unofficial prompt library containing unreviewed legal, HR or customer language can scale poor judgement very efficiently.

But if leaders stop at enforcement, they miss a more useful question:

What problem was the workaround trying to solve?

If people are pasting documents into public AI tools, the internal knowledge environment may not be doing its job. If employees are using personal subscriptions beside approved enterprise tools, the approved tool may not support the work particularly well. If teams are building their own prompt libraries, capability may be forming faster than the organisation can recognise and support it.

The issue is not simply that people are using tools without permission. It may also be that the organisation is learning about AI faster than its formal processes are.

Much of that learning is happening in places leaders do not routinely see.

Quiet adoption reveals the operating model

Informal AI use can be a useful mirror, though not always a flattering one.

It shows where systems are difficult to use, where work moves too slowly, where policy is unclear and where the official process is technically correct but practically avoided.

It also shows where the approved tool exists, but the real work has moved somewhere else.

The right first response is neither panic nor applause. It is curiosity with clear boundaries.

The useful question is not only:

Who is breaking the rules?

It is:

What are people trying to get done that the current operating model is not helping them do safely?

That does not weaken governance. It gives governance something real to work with.

A useful AI policy should help people understand which tools are approved, what information can go where, when human judgement must remain involved and who to ask when the work falls into a grey area.

In practical terms, governance needs to spend less time defaulting to no and more time setting clear conditions for yes.

That still requires control, but it gives people a clearer way to use AI responsibly.

Map the behaviour before writing another policy

Before tightening AI governance or adding another framework, leaders should understand what people are already doing.

Not as a witch-hunt. Good luck getting honest answers that way.

Treat it as an operating scan.

Choose a small number of teams and ask straightforward questions:

• Which AI tools are you using?

• What work are you using them for?

• What information are you entering?

• What problem is the workaround solving?

• What remains unclear?

• What would you be reluctant to disclose because you are unsure how leadership would respond?

That final question may tell you more than the policy document.

Then sort what you find into four practical categories.

Stop now

The activity creates unacceptable data, privacy, legal or control exposure.

Govern better

The behaviour is useful, but it needs clearer rules, approved tools, stronger data boundaries or better oversight.

Legitimise

The workaround is solving a real organisational problem and should no longer be treated as unofficial.

Scale carefully

The experiment is useful, understandable and narrow enough to govern with confidence.

The point is not to endorse every quiet experiment. Some should stop immediately.

It is to recognise that the formal AI strategy may not be the only strategy operating inside the organisation.

A small, regular review can help. Bring together technology, risk and a business owner close to the work. Review actual examples rather than abstract scenarios.

Thirty minutes is probably enough.

The leadership decision

The real AI strategy may already be forming through the workarounds.

Leaders can wait until those behaviours appear as an incident, an audit finding or an awkward discovery during procurement. Or they can treat them earlier as signals of weak tools, unclear rules, accumulating risk or useful capability beginning to emerge.

The aim is not to reward unauthorised behaviour.

It is to govern from reality rather than from the assumption that policy has already shaped the work.

Quiet AI experiments can become liabilities. They can also reveal where better tools, clearer rules, stronger knowledge access and more practical training are needed.

It is worth understanding the difference before useful learning disappears underground or walks out the door.

Executive note for leaders

Before tightening AI governance, ask:

What are people already using AI for, and what is that behaviour trying to tell us?

Quiet experiments can reveal risk. They can also reveal poor knowledge access, slow systems, unclear policy, weak training and useful capability forming before the organisation has named it.

Map the actual use first.

Then decide what to stop, govern, legitimise or scale.

The aim is not permission without control.

It is governance that starts from reality.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.

Welcome, and thanks for being here. Every meaningful transformation I’ve been part of, whether building something from scratch, working with clients or leading change inside a large organisation, has started in much the same way.

Not with a dramatic leap, but with a quiet, intentional shift.

A clearer question.

A reframed assumption.

A small spark that ripples outward into something bigger.

That is the idea behind The Ripple Effect, and it is why I’m launching this publication for executives and senior leaders carrying responsibility for digital, customer experience and AI-enabled change.

Why this space exists

We are entering a moment where AI is no longer only a technology conversation. It is a leadership conversation.

It is reshaping how organisations design, build, learn, collaborate, serve customers and make decisions.

For leaders with P&L, portfolio or transformation accountability, that creates a different kind of pressure: where to place the next dollar, how far to lean into AI, what needs to change around the technology, and how to bring people with you while the ground is moving.

A few things have become clear to me along the way.

AI can accelerate possibility, but it does not replace perspective.

Tools evolve quickly. Understanding, culture and organisational capability take longer.

My view is simple.

We do not need more hype.

We do not need to pretend everything is changing overnight.

And we do not need to abandon what makes us human to embrace what makes us capable.

What we do need is space to think clearly.

To explore how well-considered, human-centred shifts can help organisations make better strategic choices, move from experimentation into delivery, and modernise without losing trust, momentum or control.

That is what this publication is for.

What you can expect

I’ll share reflections and practical ideas for executives and senior leaders across:

• AI and human-centred leadership

• Digital and customer experience modernisation in complex organisations

• Strategy, systems and design thinking

• The move from experimentation into accountable delivery

• How small decisions create meaningful momentum

• The judgement, sequencing and operating discipline required as AI becomes part of everyday work

• Stories and perspectives from people doing thoughtful, impactful work around the world

I try to write the way I lead: calm, curious and practical, with a touch of creative spark and a preference for clarity and momentum over noise.

Some posts will be short. Others will go deeper. Some will be conversational. Occasionally, I’ll explore questions I’m still wrestling with myself.

Over time, I also hope to bring in interviews, perspectives and ideas from beyond my own experience, because the most interesting insights often come from places we do not expect.

For now, this is the beginning.

A new publication, a few ideas worth exploring, and a small push away from the shore.

Let’s see where the ripple leads.

Subscribe

You can subscribe for free. Every edition is delivered directly to your inbox and published here on Substack.

If a piece raises a question, surfaces a pattern, or helps you think more clearly about a decision, I’d value the conversation.

Thanks for reading,

Stuart Gonsal MAICD
With occasional help from Springsteen, my Border Collie, who reminds me that clarity comes from movement 🐾.

Connect

LinkedIn – Follow for practical leadership on AI-era opportunity. ↗

Disclaimer

Everything shared in The Ripple Effect reflects my personal views and does not reflect those of my current or past employers, clients or partners. Any examples are illustrative, drawn from publicly known patterns or anonymised experience.