Why the best AI use cases are now emerging from business teams

When companies talk about artificial intelligence, the conversation often centers on models, platforms, or technology investments

Yet the use cases that truly transform employees’ daily work are sometimes far more modest.

At Deel, the first automations to see mass adoption weren’t big IT projects or sweeping transformation programs. They involved preparing for client meetings, writing up call summaries, sorting internal alerts, or documenting incidents.

“Taken individually, these are just details. Put together, they add up to 60 to 70% of the team’s time. You only discover them by watching the actual work on the ground — never from a steering committee,” explains Anne-Lise Bouaziz-Klotz, Director of Customer Operations at Deel.

That observation points to one of the most interesting phenomena in AI adoption today. The most transformative use cases no longer necessarily originate in IT departments. They emerge closest to operations, where employees accumulate hundreds of tiny, invisible frictions every day — invisible, that is, to the rest of the organization.

Until AI came along, technological innovation followed a fairly stable path. Business teams would articulate their needs, IT departments would select the technology, run the project, and organize the rollout. Innovation was rare, expensive, and largely centralized.

Generative AI is upending that balance.

The end of technological scarcity

For a long time, development capacity was the main limiting factor. Companies could spot plenty of automation opportunities without having the resources to actually build them. Every project required trade-offs, budgets, and specialized teams.

Generative AI fundamentally changes that equation. For the first time, non-technical employees can build their own tools to address their day-to-day needs. The hard part is no longer building the solution — it’s identifying the problem worth solving.

That knowledge rarely sits with technical teams; it belongs to the employees who deal with operational irritants every day.

At Nabla, this dynamic emerged very early. The first use cases didn’t come from a technology roadmap drawn up by an IT department. They emerged from go-to-market teams facing a very concrete constraint: taking notes during client calls.

“The first concrete use case came from the go-to-market teams: note-taking during client calls. The problem was simple — taking notes pulled attention away, hurt the quality of listening with our clients, and required a lot of time spent writing things up afterward,” explains Delphine Groll, Co-founder and Chief Operating Officer of Nabla.

The use case sounds almost trivial — and that’s exactly what makes it interesting. The value doesn’t come from a technological feat, but from solving an everyday problem identified by the people who actually face it.

Once that first proof of concept landed, momentum picked up. “That use case served as an internal demonstration. Once teams saw what AI could concretely solve, the range of possibilities opened up naturally,” Groll continues.

Business know-how becomes a technological edge

One of the paradoxes of artificial intelligence is that it raises the value of business know-how at the very moment it’s democratizing access to technology.

For decades, the scarce skill was mastery of technical tools. Today, that scarcity is gradually shifting toward a fine-grained understanding of processes. Generative models already know how to write, summarize, search, or analyze. What they can’t do on their own is identify an organization’s pain points or understand which tasks are wasting time unnecessarily.

That knowledge remains deeply embedded within business teams. This is exactly what Anne-Lise Bouaziz-Klotz observes at Deel. In her view, the most useful initiatives aren’t the big projects traditionally run by technology functions.

“An IT department prioritizes big platform projects. We automated the small, invisible day-to-day frictions instead: the write-up after every client call, sorting internal alerts, documenting bugs, or figuring out which team a problem should be routed to.”

Artificial intelligence, then, looks especially well suited to optimizing the real work — the work that happens between formal processes and often stays invisible both in org charts and in IT roadmaps.

Support functions become AI’s first labs

This shift explains why some of the most advanced experiments today are showing up in departments rarely associated with technological innovation.

HR, customer support, operations, and marketing are among the first to benefit from this new generation of tools. These functions rely heavily on analysis, synthesis, information retrieval, and coordination — tasks where generative models deliver immediate value.

At Deel, the first widely adopted use cases involved automatically flagging at-risk payroll runs or onboardings, preparing for client meetings, drafting emails, and generating daily briefings for managers.

“Adoption was immediate because AI was replacing hours of manual work,” sums up Anne-Lise Bouaziz-Klotz.

At Nabla, the momentum spread across the whole organization too. Teams gradually adopted automated note-taking tools, conversational assistants like ChatGPT or Claude, and various conversational-intelligence solutions for sales teams. AI features built into Slack or Notion were connected to the existing work environment to smooth access to information and streamline certain daily tasks.

Innovation is no longer flowing only from technical teams down to business teams — it’s now flowing back up from the ground to the organization.

A new generation of builders

This shift is gradually transforming the very nature of certain roles. In Deel’s Customer Success teams, for instance, the role has started to evolve as automations have multiplied.

“Before, the Customer Success Manager spent their time executing — passing along requests, following up, writing reports. Today, AI prepares the work — a draft email, a report, an enriched alert — and the human validates and decides.”

The shift is significant: value is concentrating less on execution and more on judgment, client relationships, and decision-making.

Even more interesting, teams are gradually becoming responsible for their own automations.

“I build them, then I train the team to keep them running,” explains Anne-Lise Bouaziz-Klotz.

At Nabla, this shift has even led to the creation of new roles. “We’ve created new roles like prompt engineer and AI Lead,” says Delphine Groll. The former design and evaluate the prompts used in the company’s products. The latter work directly with business teams to build agents that meet their operational needs.

Artificial intelligence, then, is no longer just augmenting existing roles — it’s already helping new specialties emerge within organizations.

IT’s role is changing

None of this means IT departments are losing ground. If anything, their importance could grow. It’s simply that their role is shifting. Until now, they were responsible for building systems; going forward, they’ll be more responsible for orchestrating them.

Security, data governance, regulatory compliance, architecture, and access policies become even more critical issues as business teams gain autonomy.

The more employees build their own agents and workflows, the more obvious the need for a shared framework becomes. IT departments are gradually shifting from being the sole producers of innovation to being the guarantors of its consistency.

This shift is especially visible in heavily regulated sectors. “We operate in healthcare, a highly sensitive and regulated sector. That means validation cycles run by security teams, which can sometimes lengthen deployment timelines. It’s not a blocker in itself — it’s a structural constraint you have to factor in,” notes Delphine Groll.

A new geography of innovation

The accounts gathered from Nabla and Deel converge on the same observation: the most relevant use cases aren’t necessarily the ones dreamed up during big strategic-planning exercises.

They often emerge from everyday frustrations, operational irritants, or repetitive tasks identified by the teams themselves.

“The best ideas don’t come from a single team or from leadership. They come from the people who are testing things, learning, and asking themselves the right questions,” observes Delphine Groll.

Anne-Lise Bouaziz-Klotz makes a similar point when she explains that the best automations arise directly from difficulties encountered on the ground.

“Management provides the vision and unlocks the resources. The business team identifies the problem and builds the solution.”

This convergence is probably the most important signal here. For decades, companies treated technology as a centralized capability. Artificial intelligence seems to be introducing a different model — one in which innovation first emerges from business teams before being structured at the organizational level.

So the question is no longer just which technologies to adopt. It becomes: how do you enable the employees who best understand the problems to build part of the solutions themselves?

For decades, digital innovation flowed from the center of the company out to its periphery. Artificial intelligence seems to be taking the opposite path.

It’s now being born on the ground, where a company’s everyday problems live, before being gradually industrialized at the scale of the organization.