“It’s Not 10% More Efficient. Now We’re Talking 10x.”

Finance teams have long relied on spreadsheets and static enterprise software to manage reporting and analysis. At high-growth companies, that balance can strain quickly.

Before co-founding Concourse, Matthieu Hafemeister worked at Jeeves, a global expense management and corporate card company operating across 30 countries. As the company scaled from roughly 15 employees to more than 200, financial complexity expanded across currencies, bank accounts, and lending data.

“Making sense of data at Jeeves was really, really hard,” Hafemeister said to AIM Media House. “We were trying to understand inflows and outflows of cash across 80 bank accounts that were not connected.”

As the business grew, he said, the finance function scaled primarily through headcount and spreadsheets.

“We would just hire more people and scale up the team,” he said. “Software is great, but it is very static. Once you implement it and the business changes, you’ve got to re-implement it. Spreadsheets are flexible, but they’re very manual and very time consuming. Both typically exist within the enterprise.”

That operational tension is unfolding at a time when organizations are increasing investment in generative AI. According to a 2025 global survey by McKinsey & Company, 44% of CFO respondents said their organizations were already using generative AI for more than five use cases, and 65% planned to increase AI investments as usage expanded. The same report noted that many organizations have yet to scale AI beyond pilots across enterprise processes.

Separate benchmarking shows finance teams still devote most of their time to routine tasks. The 2024 survey by FP&A Trends found that only 35% of FP&A time is spent on gaining insights and driving action, while roughly 45% is consumed by data collection and validation. Accenture has similarly reported that finance teams can spend up to 85% of their time on transactional or production work rather than forward-looking analysis.

Hafemeister said those patterns align with what he sees among enterprise customers.

“When you talk to most finance teams, a lot of them are behind, don’t have enough resources, have too much work to do,” he said. “I haven’t heard of a finance team that told me, ‘We have way extra capacity.’”

Recurring Reports Become the First AI Test Case

Hafemeister said many organizations begin with experimentation using general-purpose large language models before attempting to embed AI in structured workflows.

“Our best customers are customers that have definitely been working with AI,” he said. “Most of our customers have tried ChatGPT… but as it relates to finance work, data analysis work, reporting workflows, that’s where the hard parts are.”

Recurring reporting cycles, he said, are a common first target.

“A lot of those teams have weekly reports they’re sending out to different business leaders,” Hafemeister said. “Those reports are extremely time consuming and need to be done manually every week.”

In those cases, AI systems connected directly to ERP, CRM, billing, and data warehouse platforms are used to reproduce reports on a recurring cadence.

“We can reproduce the exact same output over and over, but with updated information and updated analysis,” he said.

He described measurable changes in output once those workflows are implemented.

“One of the metrics we look for is how much more analysis a team is able to do once they implement Concourse,” Hafemeister said. “The benchmark is five to six times more analysis within the first month.”

He also described compression in task duration.

“Work that would typically take an hour takes five minutes,” he said.

For larger enterprise deployments, he said adoption often begins with a proof of concept.

“For enterprise customers, we’ll do proof of concepts,” Hafemeister said. “That allows us to validate how the system works and present the business case around time savings and ROI.”

He said some enterprise customers report “thousands of hours a year” in time savings.

Hafemeister described early adopters as typically mid-market companies in the $50 million to $300 million revenue range, often with 40- to 50-person finance teams, as well as a smaller number of Fortune 500 organizations.

“The larger the company, the more data and work there is,” he said. “Analyses that break Excel are where agents can do a really good job very fast.”

Trust Determines Deployment

While productivity gains may initiate interest, Hafemeister said finance leaders focus heavily on traceability.

“The next question in finance is, ‘How did you calculate that?’” he said.

He said enterprise users are given visibility into underlying SQL queries, code execution, and raw outputs alongside natural language explanations.

“You can see the raw code, the raw result of those queries, and a natural language explanation,” he said.

Rollouts often begin with technical stakeholders.

“We typically start with the data team… then junior analysts… and then move up the chain within finance,” Hafemeister said. “That allows teams to get confident that the system is understanding the way they’re calculating certain metrics.”

Security controls also factor into procurement decisions. Hafemeister said the company maintains SOC 2 Type II compliance, supports SAML-based single sign-on, and does not train models on customer data.

“We don’t train on customer data,” he said. “In many instances, we don’t even host the data on our servers.”

Industry analysts have indicated that governance and integration will shape enterprise AI purchasing decisions. Gartner has projected that generative AI capabilities will become embedded across enterprise software, shifting buyer focus toward reliability, integration, and compliance controls.

For finance leaders, Hafemeister said, the objective extends beyond accelerating dashboards.

“I don’t want to just understand that revenue is up by X percent,” he said. “I want to understand why.”

He described systems capable of traversing multiple datasets to explain drivers such as customer acquisition, churn, upsell activity, and channel mix.

“The real question is, why did it grow 20%? How can I keep growing 20%?” he said.

In his view, enterprise finance teams are evaluating whether AI can reduce the time spent assembling data and increase the time spent interrogating it, while preserving the controls and transparency required in production environments.