With finance leaders under pressure to cut costs and demonstrate returns from AI investment, Sebastien Marchon, CEO of Rydoo, explores the emerging trend of bringing existing business systems into the AI tools employees already use, and how CFOs can reduce complexity and improve productivity.
Finance teams have spent the past few years testing out AI, and many are now finding that using it and getting value from it are different things. Employees have picked up tools like ChatGPT and Claude on their own, while the often expensive systems that finance teams rely on for approvals, spend and reporting stay behind a separate login. That disconnect can help explain why businesses are struggling to see a clear return on their AI investments.
Adoption is high yet returns are hard to find
Wide adoption of AI has not yet turned into a clear payoff. Protiviti's 2026 Global Finance Trends Survey found that 77% of finance organisations are using AI, yet only 14% of those have a defined AI strategy. Similarly, PEX's 2026 State of Finance Report found that just 5% can measure a return on their AI use.
A large part of the reason is how the tools are introduced. They tend to be added one at a time, each with its own interface, permissions and training needs, and so teams end up with a patchwork of capabilities that adds complexity to their day rather than removing it. Not only that, but it creates operational silos and makes data harder to find and track.
With finance teams under increasing cost pressure, maintaining a growing number of AI tools without clear returns becomes difficult to justify.
Recent reporting found that efficiency and cost control are the main priorities for finance leaders this year, and that 47% of general corporates plan to prioritise AI adoption within the next 18 months.
Finance teams are being asked to do more yet with the same headcount and budget. AI is well suited to this challenge, but this operating environment means that every additional platform has to earn its place. Each one brings procurement, security review, onboarding and ongoing support, so a strategic, company-wide approach is needed to ensure there is no duplication or unnecessary spend.
Meeting people where they are
As a result of this challenging environment, a trend is emerging: rather than adding yet another tool, businesses are bringing their existing systems into the AI assistants employees already use, whether that’s ChatGPT, Claude or any major tool the company has chosen. This is made possible by the Model Context Protocol, or MCP, an open standard that allows AI applications to connect to business systems and data, so that employees can get things done through the interfaces they already use.
Expense management is a good example of where this approach can make a difference. Approvals are frequent, small and time sensitive, yet they often wait because the approver has to open a separate system, while answering relatively simple questions about spend can still mean exporting data and searching through spreadsheets.
For finance, this only works if control comes with the convenience. Approving spend or querying data from within an AI assistant must respect the same permissions, approval rules and audit trail as the underlying system, so that speed never comes at the expense of compliance or visibility.
At Rydoo, we’ve developed an MCP to integrate our expense management software into the likes of ChatGPT and Claude. While simplifying employees’ work day, the bigger opportunity extends well beyond expenses.
As more business applications become accessible through AI interfaces, C-suite leaders have an opportunity to rethink how their existing technology is used. Rather than continually investing in additional platforms, the focus should be on making the systems they already have easier for employees to access and use. This is what can help businesses get more value from their existing technology investments while reducing unnecessary complexity. Finance teams don’t need yet another login. They need the systems they already trust to show up where they already work.