TL;DR: Four AI transformation questions from a leaders' dinner
The Challenge:
Most leaders at GAP's dinner had AI pilots to point to, but few had results they could defend to a board. Boards now ask to see the workflow that changed, the cost that came out or the risk that fell.
Key Takeaways:
- Value Comes From Redesign: Roughly 9 in 10 organizations are raising AI budgets, yet a large share still report no meaningful return. Winning teams redesign the workflow first, then apply AI where it changes the economics.
- Governance Built In: Working controls embedded in the AI system, from data access to agent accountability, let leaders move faster later.
- Readiness Is Specific: Upskilling has to fit each person's workflow, and a serious assessment names which teams are ready, squeezed or resistant.
- Token Spend Discipline: Leaders track spend by workload, route tasks to the right model and judge each use case by business outcome divided by token and human cost.
- Decision Rights Must Change: Problems once too costly to attempt are now in scope, and leaders must set what stays human, what is human-led with AI support and what runs autonomously.
Bottom Line: The leaders who win with AI start with an honest assessment and tie every dollar of spend to a result they can defend.
Good dinners tell you the truth in a way that conference panels rarely do.
At GAP, we recently pulled together technology experts and business leaders for a dinner to talk about AI transformation: the challenges, and just as importantly, the solutions. And these were not people dabbling in artificial intelligence. They lead banks, telecom operations, healthcare and technology organizations, financial services firms and mission-driven teams. They have budgets, boards, regulators and workforces counting on them to get this right.
We were joined by Dwayne Stoltz, GAP's Vice President of AI Transformation, who has spent over a decade researching, consulting on and practicing AI frameworks, process automation and team readiness across industries. Having Dwayne at the table turned a good conversation into a practical one.
Four questions kept returning, in different words, from seats around the table. If you lead an organization today, I suspect at least one of these questions is keeping you up at night too.
The board wants return on investment, not more pilots
Question 1: "What are the pain points keeping you up at night? Is it pressure from the board to produce AI ROI, turning pilots into value?"
Almost every leader at the table had a pilot to point to. But very few had a result they were ready to defend in a boardroom.
That disconnect between running numerous AI pilots and delivering tangible, boardroom-ready results is now the defining problem in enterprise AI. Boards funded experimentation in good faith. What they are asking for next is different. They want to see the workflow that changed, the cost that came out, the revenue that moved or the risk that fell. And they want proof (or at the very least, a great deal of confidence) that it holds up at scale.
Pilots stall for predictable reasons. They are layered on top of processes that were never redesigned. They depend on one enthusiastic sponsor. They measure activity, such as users onboarded, prompts run and demos delivered, instead of outcomes. And when the pilot ends, nobody owns what happens next.
The leaders who are breaking through share a pattern the research confirms: value comes from redesign, not from tools. Roughly 9 in 10 organizations are raising AI budgets, yet a large share still report no meaningful return. The difference is structural. Winning teams redesign the workflow first, then apply AI where it changes the economics of that workflow. They start with a problem worth solving, define the business outcome in advance and hold the initiative to that number.
If your board conversation is getting harder, that is not a failure of the technology. It is a signal the conversation needs to move from experimentation to operating model.
Governance cannot be an afterthought
Question 2: "Ensuring effective governance and compliance to protect data?"
The second conversation was quieter, but more serious. Leaders are moving fast with AI at exactly the moment when the data underneath it is most exposed. Customer data, employee data, proprietary processes and institutional knowledge is flowing into models and tools that were not all procured the same way.
Governance, as the table discussed it, is not a policy binder. It is a set of working controls embedded into the AI system itself: who can access what data, which models can be used for which work, how outputs are checked, how decisions are logged, and who is accountable when an agent acts (and not just when a person clicks).
That last point matters more every month. We have moved from chatbots that answer, to copilots that assist, to agents that act. Intelligence scales easily. Accountability does not. Leaders must keep a human in the lead, with clear ownership, escalation paths and audit trails, especially in regulated industries and workflows involving customers, money, health or safety.
AI safety and cybersecurity belong in the same conversation too. A governance framework that ignores model risk, prompt and data leakage, third-party tool sprawl and agent permissions is incomplete. The good news from the table was consensus: governance done early does not slow transformation down. It is what allows leaders to move faster later, because the board, the regulators and their own teams can trust what has been built.
Readiness is a people problem first
Question 3: "Evaluating team readiness across the firm and implementing effective upskilling?"
If there was one theme that connected every story at the table, it was time: human time and what it is worth.
Several guests made the same point in different ways: the real prize is not doing the same work with fewer people. It is redirecting people toward the higher-value work they never had time for, such as the client conversation, the design judgment, the relationship, and the problem that actually requires experience. Growth from redirecting freed capacity can be worth far more than the labor savings alone.
That only happens if leaders know, honestly and continuously, what their teams can do today. Readiness is not a one-time survey. Roles are changing under people's feet, AI skills are commanding a real premium in the market, and the gap between confident early adopters and everyone else widens quickly when training is generic.
Effective upskilling is specific. It is tied to a person's actual workflow, measured against the work they will do next quarter, and reinforced by managers who use the same tools themselves. It also has to be honest about cohorts: some teams are ready to run, many are capable but squeezed, and a few will resist until they see a peer succeed. A serious readiness assessment names those groups and plans for each of them, rather than pretending a single training day fixes all three.
Token spend is now a boardroom line item
Question 4: "Optimizing token consumption to drive the right balance between spend and value?"
A year or two ago, executives might not have used the word “tokens” unprompted. But it was a big topic at this year’s dinner because token consumption is how enterprise AI is metered, and unmanaged consumption compounds fast.
The pattern is familiar to anyone who has watched a cloud bill grow up. Adoption succeeds, usage spreads, more work is routed to the most powerful and most expensive models whether the task needs them or not, and finance discovers the total only after it has moved. In practice, fewer than one in ten tasks genuinely need a frontier model, yet more than half of requests are commonly routed as if they do.
The leaders managing this well treat tokens like any other production input. They instrument spend at the workload level, so they can see which use case consumes what. They route work intelligently: the right model for the right task, and they use chargeback to business units, which turns an abstract technology bill into a set of owned decisions. Most importantly, they judge spend against return: quantified business outcome divided by token cost plus human cost. If that ratio cannot be defended, the use case gets redesigned or stopped, no matter how impressive the demo was.
Token economics are not a reason to slow down. They are the discipline that keeps a successful program fundable.
The problems we never attempted are now in scope
The most energizing part of the evening came near the end, when the conversation turned from constraints to possibility.
Every organization carries a backlog of problems it has never seriously attempted, not because they do not matter, but because the human labor cost of attempting them was prohibitive. Reviewing every contract the same way is one of them. So is following up on every exception, personalizing every client interaction, and auditing every process every time, instead of sampling and hoping.
AI changes that math. Those previously unsolvable problems are often the best first redesign candidates, alongside the workflows and standard operating procedures leaders already know are broken. They carry clear value, they were never optimized for the old cost structure, and success is easy for a board to recognize because the work simply was not happening before.
Pursuing them does require one uncomfortable step: reorganizing how decisions get made. When an agent can prepare, check and in defined cases act, the old approval chains do not automatically fit. Leaders need to redesign decision rights at the same time they redesign workflows: what stays human, what is human-led with AI support, and what can run autonomously inside clear guardrails. Organizations that skip that step end up with fast tools and slow decisions.
How GAP approaches AI transformation: Assess first, then build in workshops
At GAP, we do not start with a platform pitch. We start where our dinner conversation started, with an honest assessment. Our experience across AI frameworks, process automation and team readiness has taught us that the right second step depends entirely on what the first step finds.
Our proof comes in the order we recommend to clients. First, our own engineers, more than 500 of them, have been upskilled to work autonomously with AI as part of how they build, test and deliver, and they now work as forward-deployed engineers, embedded inside client organizations.
Second, we have applied the same transformation to our own non-engineering functions. Third, and only on that foundation, we deliver transformation for clients. We are not asking leaders to attempt anything we have not already done ourselves, at scale, inside our own operation.
Where GAP Transforms: Products, Processes and People
That work sits inside a simple frame. GAP combines strategic consulting with hands-on engineering to turn AI into measurable outcomes, across the three places transformation actually happens: the products an organization offers, the processes it runs on and the people who do the work. Every question from the dinner maps onto one of those three.
Board return is won or lost in products and processes, governance and token discipline live in processes, and readiness lives in people. If a product initiative cannot name the revenue, retention, or cost line it moves, it does not make the roadmap.
| What GAP Transforms | How GAP Helps |
| Products |
|
| Processes |
|
| People |
|
None of that works as a handoff. GAP's forward-deployed engineers are embedded inside client organizations, alongside their own teams, bringing deep expertise in AI transformation to the initiatives that matter most. They sit in the stand-ups, the design reviews and the board-readiness conversations, and they stay accountable for outcomes, not hours. That is the difference between adding capacity and actually transforming how an organization works.
Ready to get started?
AI Assessments: Find out where you actually stand
An assessment gives leaders a shared, evidence-based picture across the four questions from the dinner:
- Value and workflow: Where value is leaking, and which workflows are ready for redesign rather than automation of a broken process.
- Governance and risk: What data an AI system would touch, what controls exist today, and where governance must be embedded before scale.
- People and readiness: What each team can do today, where the gaps are by role, and what specific upskilling closes them.
- Token economics: What is being spent, on which models and use cases, and what the return is per workload.
The output is not a generic maturity score. It is a sequenced view: which opportunities are worth funding first, what must be fixed before they scale, and what success will be measured against, in language a board can act on.
Workshops: Turn findings into a plan the team owns
Assessments fail when they arrive as a report and leave as a report. Dwayne's workshops are working sessions with the leaders and teams who will own the change, built around their real processes, not case studies from someone else's company.
- Redesign workshops: Leaders leave with a short list of redesign candidates: existing workflows plus the previously unsolvable problems, each tied to an owner, an outcome and a measure.
- Team readiness workshops: Teams map their current capability against the work ahead and build role-specific upskilling plans, with managers equipped to lead them.
- Governance-by-design workshops: Governance, safety and decision rights are designed into each candidate from the start: data access, model choice, human-in-the-lead checkpoints and auditability, rather than reviewed after the build.
- Token and value workshops: Spend is instrumented from day one, with routing and chargeback agreed before usage scales, so the program stays fundable as it grows.
Because GAP's forward-deployed engineers embed inside the client organization and work alongside its own teams, a workshop roadmap does not have to sit waiting for capacity. It can move quickly into build and delivery, scaling fast while holding the governance and measurement discipline the workshop established, delivering speed and lower risk along with better price-performance.
Start with an honest assessment, not another pilot
The takeaway I keep returning to is simple. The leaders who will win with AI are not the ones with the most pilots. They are the ones who assessed honestly, redesigned deliberately, governed from the start, invested in their people's readiness, and kept a clear line from every dollar of AI spend to a result they could defend.
The path ahead is a progression, not a switch. Today, humans lead and machines amplify. People set direction and make the calls, while AI drafts, checks and accelerates the work. As workflows prove reliable and the controls around them earn trust, proactive agents can take on more autonomy, acting inside clear guardrails and escalating to a person when a decision requires it.
Over time, the pattern moves toward autonomous, trusted workflows in which machines lead the way and humans contribute judgment at the points that matter most. The time to start planning for that shift is now, not after the technology forces the question, and trust will depend on the governance and measurement built today.
If the four questions above about board pressure for return, governance you can trust, a team whose readiness you actually know, and AI spend tied to value feel right, the next step is not another experiment. The next ste should be an assessment, followed by a workshop that turns it into work your team owns.
Dwayne and I would be glad to continue the conversation. If you’re ready to find out where your organization stands, reach out and request an AI assessment from GAP.
