When Expertise Spreads, Timing Wins

AI is changing the economics of expertise.

That may sound like an abstract statement, but we can already see it happening. Work that once required considerable time, specialized knowledge, or access to the right expert can increasingly be supported by AI. People can pull together information faster, explore alternatives, challenge assumptions, and get a reasonable first analysis without starting from a blank page.

For engineering organizations, there is a lot to like about that. Expertise has always been difficult to scale. The person who understands why a requirement changed three years ago, or remembers the reasoning behind an unusual configuration decision, cannot be in every meeting. AI offers another way to bring knowledge into the work.

But easier access to expertise creates a problem of its own.

When almost anyone can generate a credible analysis, the scarce resource is no longer the analysis itself. It is knowing which information matters, what deserves attention, and when there is enough evidence to act.

I think that shift will have a bigger impact on companies than simply giving more people access to AI.

AI is changing what expertise means

For years, organizations concentrated expertise in people. Experienced engineers knew where to look when something went wrong. Product leaders recognized patterns because they had lived through earlier programs. A few people could connect decisions made in one part of the lifecycle to consequences elsewhere.

Their value lay not simply in knowing more facts. They understood context.

AI changes part of that equation. It can help people find relevant information, compare alternatives, summarize technical material, and identify relationships that would otherwise take considerable effort to uncover.

That raises the baseline.

An engineer does not necessarily need twenty years of experience to begin asking better questions about a problem. A product manager can explore implications outside a familiar domain. Someone investigating an issue can examine a much broader body of information before calling in an expert.

None of this makes deep expertise less valuable. In some respects, it makes expertise more important because experienced people are better equipped to assess AI’s output.

What changes is the advantage created by access alone.

If five competitors can use similar AI models to analyze the same market shift, produce similar scenarios, and reach roughly similar conclusions, the analysis itself does not provide much differentiation. What happens next does.

The new bottleneck is knowing what matters

Engineering organizations already live with more information than any individual can reasonably absorb.

A change to a product can touch requirements, parts, software, suppliers, manufacturing processes, quality records, certifications, and service information. Those connections accumulate over years, sometimes decades.

AI can help make sense of that complexity. But it can also create another layer of output.

Ask an AI system for possible causes of a quality problem, and it may give you ten. Ask it how a supplier change could affect a program, and it can generate several scenarios. Ask for potential responses to a new requirement, and there may be half a dozen plausible paths forward.

All useful. None of them is a decision.

This is where I think organizations need to be careful not to equate more analysis with better decision-making.

AI makes it cheaper to keep thinking. There is always another scenario to run or another interpretation to consider. In an organization that already struggles to make decisions across functional boundaries, that can easily become one more reason to wait.

The real question is whether AI helps a team recognize the signal sooner.

That signal may not look particularly important at first. It could be a recurring workaround on the manufacturing floor. A supplier issue that keeps appearing in different programs. An unusual pattern in engineering changes. Service technicians are replacing the same serialized component earlier than expected. A requirement that repeatedly creates downstream confusion.

Individually, these can look like routine operational issues. Connected over time, they may tell a very different story.

That is where access to product context matters.

Engineering organizations need better signals, not more ideas

We often talk about leaders who can “see around corners.” I have never been entirely comfortable with the phrase because it makes foresight sound like a mysterious personal trait.

In an AI-rich environment, it should become much more systematic.

An organization ought to be able to detect a pattern forming across engineering, manufacturing, quality, suppliers, and service without waiting for one exceptional person to piece everything together manually.

AI can contribute to that, but only when it has useful context to work with.

Consider an engineering change. Looking at the change request alone tells an AI system relatively little. The meaning comes from the relationships around it. Which requirements are affected? What configuration does it apply to? Has a similar change been made before? Which parts, suppliers, manufacturing processes, or service procedures could be affected? What decisions were made previously, and why?

This is where the digital thread becomes important to AI.

When product information and its relationships are connected and governed, AI has something more useful than a collection of documents to search. It has context for interpreting what is happening across the product lifecycle.

That does not mean AI decides what matters on its own. Engineers and other domain experts still bring judgment that the technology lacks.

What AI can do is improve the organization’s field of view.

It can help surface relationships that are easy to miss, bring prior decisions into the current work, and expose patterns across information that no single person sees in its entirety. That gives people a better chance of noticing something while it is still an early signal rather than an obvious problem.

And early signals are valuable precisely because they are not obvious yet.

The advantage is shortening the distance between the signal and the decision

Detecting something sooner only matters if the organization can act on it.

This may be the harder part.

Many companies are built to reduce uncertainty before making a decision. That is understandable, particularly in engineering environments where mistakes can be expensive. Reviews, approvals, validation, and traceability exist for good reasons.

But there is a difference between rigor and waiting for certainty.

AI could make that distinction harder because it gives us so many ways to continue analyzing a problem. Or it could help us get comfortable making a well-supported decision sooner.

I would rather see engineering organizations pursue the second path.

That might mean using AI to identify the information needed for a decision rather than simply collecting everything available. It could mean finding comparable changes and showing what happened afterward. It might surface affected product configurations, summarize previous decisions, or identify where information is incomplete.

The goal is not for AI to make an engineering decision and move on.

The goal is to give people enough relevant context to make a better decision while there is still time for that decision to matter.

That is a meaningful distinction.

A company that recognizes a supplier pattern six months earlier has options. A team that sees configuration risk before a change reaches production has options. An organization that identifies an emerging service problem while only a small number of assets are affected has options.

By the time everyone agrees that something is a problem, many of those options may be gone.

Use AI to accelerate decisions, not just ideas

There is a temptation with every new technology to measure what it makes easier.

With generative AI, early measures have naturally focused on productivity. How quickly can we create something? How much information can we summarize? How many tasks can we automate?

Those are reasonable questions, but I do not think they take us very far toward the real opportunity in engineering.

I am more interested in questions like these:

• Did we recognize the problem sooner?
• Did we understand its impact more completely?
• Did the right people have the context they needed?
• Could we trace the reasoning behind the decision?
• Did we act while we still had meaningful choices?

Those measures move AI closer to the work that actually determines product and business outcomes.

They also put the emphasis back where it belongs. AI can broaden access to knowledge and make sophisticated analysis available to far more people. It can help an organization see patterns that were previously buried across systems, documents, and teams.

But somebody still has to decide what those patterns mean.

That is why I do not think the spread of AI makes expertise irrelevant. It changes the job expertise required.

The differentiator becomes less about possessing information that other people cannot access. It becomes the ability to bring together experience, product context, and emerging signals and make a sound decision before the answer becomes obvious.

There will be no shortage of AI-generated ideas.

The harder capability to build is an organization that knows when one of them is telling you something important and is ready to act on it.