I keep getting the same ask. The C-suite wants to be AI evangelists. They’ll go through a four-hour workshop on Claude Code or GitHub Copilot, and at the end of it they expect to be able to identify workflow improvements inside their own organization, build prototypes of those improvements, and hand them off to engineers for deployment.

Four hours of training. Workflow engineering capability on the other side.

I’ve heard this version of the request from multiple enterprises in the last two quarters. The leaders making it are smart people. Most of them have technical backgrounds somewhere in their history. Some of them used to write code. They aren’t asking out of ignorance. They’re asking because someone sold them a story about AI as a force multiplier, and they internalized it as “AI multiplies capability across the board.” That isn’t quite how the technology works, and it isn’t what a force multiplier does.

What AI generates, and what you need to evaluate it

A force multiplier amplifies what you bring to it. That’s the whole concept. The tool sharpens the expertise you brought into the room.

AI can also generate things you didn’t bring in. That’s the part the marketing usually leads with, and it’s a real capability. The tool will produce code, proposals, analysis, plans. None of that comes back labeled as right or wrong. Without expertise in the domain it’s operating in, you have no way to tell which of the generated output reflects reality and which is hallucinated. The generation is fast. The verification still takes someone who knows.

This is why my favorite use case for AI is artifact creation. When the tool produces something testable, the loop closes. Code can be run. Articles can be fact-checked. Plans can be reviewed against ground truth. The artifact is right there for inspection, and the inspection is where the actual value gets validated. The expertise required to do the inspection is the part that doesn’t come out of a workshop.

What the tool cannot do is gauge judgment and reactions, evaluate whether a plan accounts for the political and organizational dynamics that make it land or die, or substitute for the years of context that produce engineering taste. Those are things a human still has to bring. AI is good for the work where the artifact can be tested. It is much less useful for the work where the answer depends on knowing the room.

What an executive can actually evaluate

There’s a specific list of things you need to know to evaluate whether a workflow improvement proposal is workable in a real codebase.

Actual dependencies. Level of effort. Brittleness, meaning what cracks first when you change something. Lateral concerns, meaning what gets affected downstream when you modify a workflow. Holistic thinking about how this fits into everything else engineering is shipping this quarter. Implementation reality, not just architectural sketches.

Those are not items in the executive wheelhouse. They aren’t supposed to be. A CTO might have some of them. An SVP of Strategy will not. The Chief Revenue Officer absolutely will not. Asking AI to fill that gap is asking the wrong tool, because AI will produce confident-sounding answers to all of those questions, and it cannot tell anyone whether the answers are right. Correctness here requires context that lives in the codebase and in the heads of the engineers operating it.

So when a leader who has been out of engineering workflows for a decade asks Claude Code to propose workflow improvements, the artifact that comes back is plausible. It looks like an engineering proposal. The leader does not have the expertise to know whether the proposal accounts for the architectural decisions made three years ago that constrain what’s feasible now, or whether the team it’s being handed to has any capacity to absorb it on top of what they’re already shipping. The proposal looks workable to the executive. The engineering team is the group that discovers it isn’t.

And then somebody, usually a director or a senior staff engineer, has to explain in diplomatic language why the executive’s proposal cannot ship as written. That conversation is the cost. It happens every time, and it shouldn’t have to.

What AI does land well for executives

I want to be clear about something. AI applied to the actual work an executive does is enormously valuable, and the workshop should be designed around that work.

An executive can use AI to identify candidate workflows that might benefit from improvement. That’s a strategic spotting exercise, and it’s exactly what an executive is positioned for. They know where the business is feeling friction. They know which teams are blocked. They can use the tool to think through where AI-assisted improvement might land. That work is real, and it’s the right thing for an executive to be doing with AI.

What stays in the executive wheelhouse alongside it is the rest of the strategic and communication work the role actually involves. Framing sharper questions before a meeting with the VP of Engineering. Evaluating proposals coming up from the engineering organization with more analytical rigor than the executive could bring without the tool. Drafting board communications that take dense technical context and compress it into language a non-technical audience can act on. All of that fits the executive role. AI extends it considerably. The executive comes out measurably better at the actual job.

What stays outside that wheelhouse is everything that requires running the generated artifact through expertise the executive doesn’t have. Implementation. Dependencies. Effort sizing. Brittleness. Lateral effects. None of that gets unlocked by four hours of training, and the workshop should not imply otherwise.

The workshop that gets bought versus the workshop that would land

The four-hour Copilot training is easy to procure. The curriculum is already written. Vendors are competing for the booking. The calendar slot is clean. Everyone gets to check a box that says “executive team trained on AI.” Nobody gets fired for booking that workshop.

The workshop that would actually work for an executive audience requires custom design, a facilitator who can hold both the leadership room and the engineering organization in their head at the same time, and content that covers candidate-workflow discovery, evaluation frameworks, and how to have substantive technical conversations without being either credulous or dismissive. Most training catalogs don’t carry that product. The firms that can deliver it charge accordingly.

So the easier one gets bought, and the cost shows up downstream. Executives walk out of the four-hour session believing they understand engineering workflows because they used Claude Code to refactor a Python script during the lab. They propose changes informed by that confidence. Engineering teams absorb the cost of defusing those proposals diplomatically. The organization accumulates skepticism about whether the AI investment is producing returns proportionate to the spend. The original goal, building leadership AI literacy, gets harder to pursue next year because the first attempt produced friction with engineering rather than alignment.

What this actually needs to be

The intent behind the request is sound. Executives need a working grasp of AI. They need to discuss it credibly with their boards, make investment decisions with their eyes open, and evangelize it internally without overpromising. A workshop can absolutely produce that outcome. It needs to be designed around what executives actually do, not a stripped-down version of the engineering curriculum.

If you’re a leader buying training for your C-suite, ask the vendor what the executive workshop is. If the answer is the same four-hour Copilot deck they sell to engineers with the labs pulled out, you’re buying the wrong product. If they don’t have a different answer, find a different vendor.

There is no substitute for expertise. AI does not generate it, and a four-hour workshop does not transfer it. What the workshop can do, when it is designed for the right audience, is teach executives to use AI brilliantly inside their own job, so they can spot the workflows worth investigating and bring those to the engineers who have the expertise to evaluate them. That is the workshop you should be buying.

The four-hour Copilot workshop will not make your executives into engineers. It will make them confident about engineering decisions they aren’t qualified to evaluate. That’s worse than where they started.

Michael Rishi Forrester is a Generative AI Strategist and platform engineering veteran with over 25 years in operations, DevOps, and technical education. He has trained more than one million engineers through every major platform shift, and he currently leads the AI for Leadership and Organizations business stream at Accenture LearnVantage.

Connect: LinkedIn | X | Bluesky | Mastodon | Micro.blog