Data: The Fuel for AI Agents—An Interview with Michael Cohen, CPO of Intescia

Data: The Cornerstone of AI Strategy

Why has data once again become such a central issue, even though it has always existed?

"Data is nothing new. APIs, open data, and free datasets existed long before the advent of AI agents and protocols like MCP. What has changed is how we can now use them. AI models are built on and fed by data, and this is precisely where Intescia—formed through several acquisitions (including Explore, DoubleTrade, and Codata, among others)—has an advantage: the group owns its own data."

"MCP's new concept should be understood as an enhanced API, capable of telling AI agents what data is available and how to use it. But simply having data isn't enough. You still need to know how to use it."

The real issue: quality and enrichment, not fundraising

Many AI agents simply retrieve data from the Internet. Why is that insufficient?

"This is the blind spot of many generic AI agents: they process raw, unenriched data. Yet open data, as abundant as it may be, poses concrete reliability issues: undetected duplicates, misinterpreted data, and information counted multiple times.

Intescia's work lies precisely in this area: data enrichment, deduplication, and SIRET identification. Our industry experts process this raw data to make it usable, after which our staff activate it for clients and partners to optimize their prospecting or project-search processes."

Toward the Disappearance of Interfaces in Favor of Agents

You're referring to a transformation that goes beyond data alone: the transformation of the interfaces themselves.

"This is a belief I share with a large part of the tech ecosystem: traditional user interfaces will gradually give way to autonomous agents capable of handling entire tasks. I experienced this firsthand this summer while working with several specialized agents (product owner, product manager, developer, product marketing specialist), each of whom processed the same information independently to produce landing pages, user stories, or automation workflows."

"Tasks that used to take me two days can now be completed in a few hours—sometimes even a few minutes."

The Rise of "Product Builders"

This transformation requires new skills. How does that play out internally?

"I'm seeing the emergence of a hybrid role at the intersection of development, product, and design: the Product Builder. These professionals, capable of rapidly prototyping agents using MCPs and pre-enriched data, are automating tasks that are currently performed manually or not at all. Intescia is already hiring for these types of positions.

I anticipate that this trend will accelerate in 2027, with the reinternalization of certain tools that were historically purchased as SaaS. However, I would temper this view: in my opinion, the “SaaSpocalypse” will primarily affect peripheral tools, posing no major operational risk. Rewriting an ERP or an ITSM system remains, in my view, a costly and risky gamble that few companies will take lightly.”

The New Budget Headache: Token Credits

Generative AI also raises a very practical question: that of the budget.

"This is a point I want to emphasize in particular. The 2026 budgets were drafted in late 2025, at a time when no one had anticipated the scale of token consumption associated with AI applications. The question is no longer simply how much AI will cost, but how to find the necessary budgetary room within budgets that have already been finalized."

I’ll take this line of thinking a step further: in the future, an employee’s compensation could include a fixed portion and an AI credit allocation, just like a salary. This raises a fundamental question for management: How many staff members should be assigned to support each team (marketing, product, development), and how does this decision translate into budgetary trade-offs?"

From Creators to Validators

Is the role of employees changing fundamentally?

"Yes, and in my opinion, this is one of the most significant changes: employees are gradually shifting from being executors to validators. A developer will no longer be just someone who writes code, but someone who validates what a team member has produced before it goes live. This shift affects all roles, including those in communications and content."

Intescia Customer Service Representatives

Specifically, what Intescia agents are you developing for your clients?

"We are already well on our way with our roadmap. As for the detection process, which has historically been very labor-intensive—customers used to have to manually sift through lists of RFPs or company data across the group’s various brands (Corporama, Doubletrade, Codata)—agents now make it possible to automate the sorting process based on defined criteria.

A sales engagement module—which is already in production and has just been launched—goes a step further: it helps sales teams automatically identify companies that match the right profile and then initiate automated outreach, including emails and follow-ups.

Another area of focus: scoring models capable of drawing on more than ten—or even twenty—years of historical data to estimate a client’s chances of success in a given bid and identify the criterion (price, timeline, or other) most likely to make the difference. This is an analytical capability that generic AI agents, lacking comparable depth of data, simply cannot replicate."

Governance and Security: The Next Battle

Doesn't this acceleration carry certain risks?

"In my view, this is the issue that will catch up with companies in the coming months. Right now, everyone is developing, testing, and experimenting on their own. It’s a new generation of ‘shadow IT.’ But what happens if the person who developed a critical agent leaves the company? Who documents it, who maintains it, and who is responsible in the event of a security breach or regulatory audit (PCI, DORA, NIS2, GDPR)?"

"My belief is that we must regulate these developments with the same standards as those imposed on developers (testing, traceability, documentation) without, however, stifling the agility of the business teams, which, in my view, remain best positioned to identify the right use cases."

A revolution whose full impact has not yet been realized

Any closing remarks?

"I'm very excited, while remaining realistic about the scale of the change underway. Projects that would have taken me months are now completed in a matter of hours. A problem I would have thought unsolvable was solved in thirty seconds. For me, the real limitation going forward will no longer be technological, but will be related to the resources (energy, water) needed to run these systems on a large scale."

Interview by Alexandre Savigny