I’m a big fan of meeting people where they are.
Throughout my career, that has meant paying attention to where an industry, sector, geography, buyer, or individual user sits on the adoption curve, and how much risk they are willing to tolerate.
AI does not reduce that tension.
If anything, that tension may never have been greater.
01
We have been here before
When I started my career in Product Management, we were living through a different technology transition: from on-premises software, often heavily customized for individual customers, to SaaS.
In the earliest days of SaaS, many of the benefits were immediately obvious to vendors before they were obvious to customers.
We saw lower deployment costs, easier maintenance, faster upgrades, and the efficiencies of supporting a common codebase.
Customers often experienced something else first.
Less control.
Fewer customizations.
New privacy and security concerns.
And sometimes, the disappearance of work that had once made certain people valuable inside their organizations.
The person responsible for installing patches and coordinating monthly, quarterly, or annual updates suddenly had less to install. The administrator maintaining servers saw infrastructure move somewhere else. Customization increasingly gave way to configuration.
Those concerns weren’t irrational. The technology really was changing how work got done, and who did it.
But little by little, adoption moved up the curve.
Customers began to experience the upside themselves: instant updates, broader access, less infrastructure to maintain, and faster improvement.
Eventually, SaaS became the norm.
For a long time.
02
This time, the pressure comes from users
This new transition feels different for all the reasons AI researchers, technologists, politicians, and philosophers discuss.
But one difference especially interests me:
This time, much of the pressure is coming from the user base.
Users are experimenting before many companies have decided what they are comfortable offering them.
They are bringing ChatGPT, Claude, Copilot, and other tools into work on their own.
They are asking why something that still takes ten clicks in enterprise software can be completed conversationally somewhere else.
They are beginning to expect software not just to present information, but to interpret it, synthesize it, recommend an action, and increasingly take the action for them.
That creates an interesting product dilemma.
I still believe deeply in meeting people where they are.
But meeting people where they are cannot mean leaving them there.
03
The faster horse
There is a famous idea often associated with Henry Ford: if you ask customers what they want, they may ask for a faster horse.
Whether or not he actually said it is less interesting to me than the product lesson underneath it.
Customers are very good at telling us where something hurts.
They understand their constraints. Their jobs. Their risk. Their organizations. The compromises they already make every day.
But they are far less likely to describe a solution outside the mental model they already have.
That has always been part of Product’s job: listen carefully for the problem without assuming the request is the answer.
AI raises the stakes because the distance between the “faster horse” and a fundamentally different mode of transportation may now be surprisingly small.
A customer may ask us to make a workflow faster.
The better answer may be to remove the workflow.
They may ask for better search.
The better answer may be a system that understands the question, gathers the relevant context, and returns the answer.
They may ask for another dashboard.
The better answer may be an agent that notices the condition, interprets it, and acts before anyone opens the dashboard at all.
04
The other half of the dilemma
But that does not mean we should drag customers into an autonomous future they do not trust.
That is the other half of the dilemma.
The adoption curve still exists.
Industries still differ.
Risk tolerance still differs.
And perhaps most importantly, trust differs.
A small business owner might happily let an agent draft and send a routine customer follow-up.
A compliance officer in a highly regulated enterprise may reasonably require explainability, approvals, and a complete audit trail before allowing anything close to autonomous action.
Both can be perfectly rational.
So the product challenge is not simply choosing between the old way and the new way.
It is designing the bridge between them.
And increasingly, I think that bridge is really about trust.
05
A ladder of trust
AI adoption may not happen as one dramatic leap from “the human does the work” to “the agent does the work.”
It may look more like this:
Each step represents a different degree of delegated agency.
Different customers will stop at different places on that curve.
So will different tasks.
A low-risk administrative process may move toward autonomy quickly. A consequential employment, financial, legal, or compliance decision may retain a human checkpoint for a very long time.
That means one of the most important product questions in the AI era may not be:
“What can AI do?”
It may be:
“What is this customer ready to trust it to do?”
And that is a much more interesting problem.
Because the answer is unlikely to be static.
Trust can be earned.
A system that consistently explains its recommendation may eventually earn permission to draft the action.
A system that drafts accurately may eventually earn permission to execute with approval.
A system that performs reliably within clearly defined boundaries may eventually earn greater autonomy.
The destination can be ambitious without pretending every customer is ready to arrive there at the same time.
06
Willing to move
That is increasingly how I think about AI-native product strategy.
Not:
“What AI feature can we add?”
And not:
“What did the customer ask us to build?”
But:
“What outcome are they actually trying to achieve, what has newly become possible, and how far are they ready to trust us to take them?”
That last question matters.
Because trust may become one of the most important dimensions of product-market fit in the AI era.
The technically superior solution is not necessarily the one people will adopt.
And the familiar solution is not necessarily the one we should keep building.
Product leaders have to hold both truths at once.
Meet people where they are.
Understand why they are there.
Understand what has newly become possible.
Then build a credible path toward something they may not yet know to ask for.
