The real reason your AI investment isn't paying off
Find out how Pleo syncs to Europe’s most used finance tools in real-time, letting you scale without complexity.
Finance teams have been testing if AI tools and platforms can provide a complete picture of company spending. But ask leaders whether that investment is paying off, and the answer gets murky. According to new research from Pleo*, 61% of finance leaders agree they’ve invested in AI in pockets where the actual ROI is unknown.
That gap between investment and earned value isn’t a sign of AI’s failure. Look closer, and the real problem is usually the spend data AI is asked to work with, not the AI itself.
Signs of scattered spend data
For most finance teams, spend information lives in several places at once. Expenses sit in one platform, invoices in another and approvals often happen over email or Slack. Each tool also structures the same information a little differently. For example, one uses categories where another uses tags, currency conversion happens at different points or vendor names can vary depending on how they’re entered.
This lack of cohesion forces finance teams to stitch together multiple reports, folders and spreadsheets just to get an accurate understanding of spend. It’s the type of manual work AI was supposed to own so teams and leaders could focus on acting on the numbers, instead of just piecing them together and double-checking accuracy.
“We have different entities onboarded. If you prepared transactions for export for one entity and accidentally dragged them into the wrong folder, things could go completely wrong. Then you'd have to revert everything.” - Shafier Saboerali, Financial Controller at SILICON
Why fragmentation affects AI
AI needs consistent, structured and trustworthy data to give a reliable answer and take action. When spend data models don’t line up across tools, AI can’t reconcile the full picture. It either flags a gap or fills it with a best guess that looks confident but isn’t accurate. Ask it for total spend on a vendor, and it might miss the invoices sitting in a separate system entirely, with no obvious sign that anything is missing.
The mismatch also shows up in ordinary questions, not just edge cases. If you ask AI whether a purchase falls within policy, it needs the policy, the category and the approval history to agree with each other. When the category recorded in the expense tool doesn't match the wording in the policy document, or the currency conversion happens at a different rate in each system, AI can’t tell a genuine breach from a labelling mismatch. It either overcorrects and flags spend that’s actually accurate, or misses a real problem because the numbers looked close enough to pass.
Losing trust is the real cost
Once the finance team catches AI getting something wrong, they (rightly) stop trusting it for everything else, including the spend questions it would have answered correctly. Rebuilding that trust takes far longer than the original task would have and often means falling back to the manual process AI was supposed to replace.
According to Pleo research*, 68% of finance leaders agree that AI skills, training and confidence are severely lagging in their teams. Inconsistent, unreliable outputs are a big part of why. It’s hard to build confidence in spend intelligence when it gives a different answer depending on which system it checks.
Consolidation earns ROI for spend intelligence
AI-powered platforms and tools only earn their place once they can query one connected view of spend, where expenses, budgets and approvals are in one place and an audit trail is behind every figure.
When configured strategically, AI assistants like Claude, ChatGPT and Microsoft Copilot can plug into that single, connected source of spend data. This turns a plain-language question into a real answer instead of one that still needs manual checking against a spreadsheet. Take it further, and the same consolidated system can turn a repetitive task—like chasing a receipt or fixing a category—into a single workflow handled across the tools you already use.
That’s the difference between spend intelligence worth the investment and disjointed AI tools that fall short of what they promised.
*This research is taken from a forthcoming report from Pleo, launching in September 2026. The survey was conducted by Sapio Research in April 2026, among finance decision-makers and finance professionals in organisations with 51 to 1,000 employees in Germany, Spain and the UK.
