Why AI’s next advantage isn’t the model, it’s the people who make it work

Written byJake Grensinger

Executive Consultant at Finatal and a specialist in placing AI talent.

Two announcements from the frontier of AI have caught my attention recently. Not because they involve another leap in model capability, but because of what they say about where the real challenge in enterprise AI is moving.

OpenAI has launched the OpenAI Deployment Company, embedding Forward Deployed Engineers into organizations to identify high-value opportunities and build AI into critical workflows. Its partners include major investment firms that collectively sponsor more than 2,000 businesses.

Anthropic has made a similar move, forming an enterprise AI services company with Blackstone, Hellman & Friedman and Goldman Sachs. Its Applied AI engineers will work directly with businesses to identify problems, build solutions and support their adoption.

AI is becoming an execution problem

For the last few years, much of the enterprise AI conversation has been about technology: model capability, platforms, vendors and use cases. Those questions have not disappeared but, increasingly, they are not the constraint.

The harder question is how you take rapidly improving technology and make it useful inside an actual business, with its existing data, systems, processes, controls and people.

That is partly a technology challenge. But from where I sit in the AI talent market, it is increasingly a people and execution challenge.

We are seeing growing demand for individuals who can operate between the technology and the business: technically credible enough to understand what can be built, commercially minded enough to know what is worth building and pragmatic enough to get something adopted inside an organization.

Call them Forward Deployed Engineers, AI Operating Partners, Applied AI Engineers or something else. While the titles are still evolving, the capability underneath them is becoming much clearer.

Private capital is an obvious proving ground

There is a reason private equity features so prominently in these new deployment models.

A standalone business can identify an AI use case, prove it and scale it. A private equity firm potentially has the opportunity to do that repeatedly across a portfolio, and that creates a different prize.

The question for a fund is not simply, “How should we use AI?” It is increasingly, “Where can AI materially improve performance across our portfolio, which opportunities are repeatable, and what capability do we need to make that happen?”

This, in turn, is beginning to change the talent brief. We are seeing the market develop at two distinct but connected levels.

At Finatal, we have recently recruited for both.

The portfolio company: from finance problem to working AI solution

For a PE-backed e-signature business in Milan, we supported the appointment of an embedded AI engineer to improve the analysis of annual recurring revenue, particularly churn, upgrades and downgrades.

What is important about the mandate is that the starting point was not AI itself, but a specific business problem. It was a real operating problem inside a real finance function, not an instruction to “transform the business with AI”.

Could reporting become more consistent and less manual? Could the CFO understand changes in customer behavior sooner? Could the team spend less time assembling information and more time interpreting it? Could the board have greater confidence in the numbers it was using to make decisions?

This is where I think the conversation around AI can sometimes miss the point. The most valuable opportunities do not necessarily announce themselves as AI projects. They appear as reporting problems, inefficient processes, fragmented data, slow decisions or expensive manual work.

The skill is being able to recognize where AI genuinely changes the economics or effectiveness of that workflow, and then having someone capable of delivering it.

The private equity fund: from individual use cases to portfolio-wide capability

We have seen the other side of the equation through our work appointing an AI Operating Partner for London-based private equity investor TPA Capital.

Here, the challenge was not one workflow inside one company. It was a portfolio of businesses at very different stages of AI maturity.

Some had immediate opportunities for automation and efficiency. Others needed more fundamental work around data, processes, reporting or governance. There was no reason to assume the right intervention for one business would be right for another, and that creates a very different talent requirement.

The role of an AI Operating Partner is not to arrive with a list of fashionable use cases and deploy them everywhere. It is to create the capability to make better decisions about AI across the portfolio: where to focus, where the foundations are not ready, where an intervention can create measurable value and where it can potentially be repeated elsewhere.

That distinction is important. The best AI leaders in private capital will not be measured by how much AI they introduce. They will be measured by the value they extract from it.

What we are seeing in the market

For me, the OpenAI and Anthropic announcements validate a shift that is already visible in the mandates we are working on: AI talent is moving closer to the operating problem.

The market still needs exceptional machine learning engineers, data scientists and technical leaders. It still needs people who can set strategy. But between strategy and the underlying technology, a particularly valuable category of operator is emerging: someone accountable for getting AI out of the presentation and into the way a business actually works.

For private capital, I think that capability will increasingly exist at multiple levels.

Some funds will build central AI capability or appoint an AI Operating Partner. Others will use specialist or interim expertise when required. Portfolio companies will bring in embedded engineers and operators against specific value-creation priorities. The strongest funds will probably use a combination, rather than impose one model across every asset.

The important thing is that the talent decision follows the value-creation problem, rather than the other way around.

From experimentation to accountability

Enterprise AI is moving beyond experimentation. The question is increasingly one of accountability: who is responsible for turning the technology into an operating result?

In private capital, we are already seeing that responsibility translate into new roles and new talent requirements, and I expect that to accelerate.

The next competitive advantage in AI may not come from access to a better model. It may come from having better people at turning those models into results.