The Fintech Pilot Is Becoming a Trap.

I've spent much of my career on different sides of the technology equation.

I've helped build a fintech company, worked inside a global asset manager, and led the introduction of new technology into established financial institutions.

One thing I've learned is that developing great technology and getting an organization to adopt it are two entirely different challenges.

And as AI accelerates the pace of innovation, I think that distinction is becoming more important than ever.

My concern is that the financial services industry is getting exceptionally good at testing new technology—and not nearly good enough at turning those experiments into meaningful change.

Everyone wants to innovate. Fewer are ready to change.

Think about the typical enterprise technology evaluation.

A business team identifies a problem. A technology provider demonstrates a potential solution. Stakeholders are enthusiastic. A proof of concept is launched.

The pilot produces encouraging results. Employees see the potential. Leadership recognizes the opportunity.

And then things become complicated.

Who owns the budget? Which department is responsible for implementation? How does the technology integrate with existing systems? What does compliance require? Who maintains the data? Which processes need to change?

Eventually, the initiative can become trapped between enthusiastic business users and an organization that hasn't decided how to operationalize the opportunity.

I've seen enough variations of this challenge to believe the obstacle is rarely just the technology.

It's often the operating model surrounding it.

AI is exposing the problem

Generative AI has made it remarkably easy to demonstrate new capabilities.

In a relatively short period, we've moved from systems that retrieve information to systems that can synthesize research, generate content, analyze complex questions, and increasingly coordinate multiple tasks.

Many of these demonstrations are genuinely impressive.

But the gap between an impressive demonstration and an enterprise-grade business capability remains substantial.

Consider a relatively straightforward use case: an AI assistant helping an asset management sales team prepare for advisor meetings.

The initial demonstration might summarize fund information, identify relevant talking points, and generate suggested follow-up communications.

But making that capability useful every day introduces a more demanding set of requirements.

Can it access approved investment data? Does it understand which products are appropriate for different audiences? Can it distinguish between public, internal, and restricted information? How does it handle stale data or uncertainty? Can employees use its outputs within existing compliance policies?

And perhaps most importantly, does it fit naturally into how the sales organization already works?

A successful AI demonstration proves what technology can do. A successful implementation proves what an organization is willing and able to change.

Those are very different things.

The hidden cost of innovation

I think we're approaching a period when financial institutions will begin questioning the number of disconnected AI initiatives operating across their businesses.

It's not difficult to imagine a large organization with separate AI efforts underway in sales, marketing, product management, operations, investment research, and client service.

Each initiative may make sense independently. But collectively, they can introduce duplicated costs, inconsistent data, governance complexity, and fragmented employee experiences.

The irony is that a technology intended to reduce complexity can actually increase it.

That suggests the next phase of AI adoption will be less about adding new tools and more about rationalizing where intelligence lives, how it is governed, and which workflows it should support.

Some successful experiments will need to be discontinued. Others will need to be combined. And some organizations will discover that their biggest technology problem is not a lack of innovation, but too many disconnected attempts at it.

The opportunity for fintech companies

There's an equally important lesson here for emerging fintech businesses.

For years, technology companies have often sold innovation through features, demonstrations, and technical differentiation.

Those elements still matter. But I believe enterprise buyers will increasingly demand something more practical: a clear path from proof of concept to measurable business outcomes.

That means understanding the buyer's operating environment, organizational incentives, governance requirements, and implementation constraints.

The strongest fintech companies will need to become as good at organizational adoption as they are at building technology.

And they may need to be more willing to challenge prospective customers.

If a company cannot identify who owns a business problem, how success will be measured, or what changes after a pilot succeeds, is it really ready to evaluate a solution?

Sometimes the most productive commercial conversation is not about what the software can do.

It's about whether the organization is prepared to use it.

My prediction

I believe the next wave of enterprise AI adoption will create a divide between organizations that experiment continuously and those that develop the discipline to operationalize innovation.

The winners won't necessarily be the companies with the largest AI budgets, the most pilots, or the newest models.

They'll be the ones that can connect technology to accountable business ownership, trusted data, embedded workflows, and measurable outcomes.

For fintech providers, that creates both a challenge and an opportunity. Differentiation will increasingly come from delivering a business capability, not simply access to better technology.

The next competitive advantage in financial services may not be how quickly an organization adopts AI. It may be how effectively it turns AI into a different way of working.

And that requires far more than a successful pilot.