As we near the year’s end, it’s the perfect time to step back from the AI media storm. Spend 5 minutes on LinkedIn and it’s patently obvious that AI is a magic wand that solves every problem…or an overhyped bubble waiting to burst. Exhausting.
In the reality of mid-market businesses, the highs and lows aren’t dramatic, but they are important. I was struck this week by a study from a US bank that went deep with CFOs and PE managers.
What struck me from the Citizens 2025 AI Trends in Financial Management Survey was that the reality for mid-sized companies is found in the practical areas where AI is becoming a reliable workhorse.
Here are three takeaways to consider as you reflect on 2025 performance and 2026 planning:
1. The ROI is Becoming Tangible
We’ve moved past the "experimental" phase. The survey shows that 61% of middle market CFOs report that AI has made financial processes easier. And there’s real math behind the finding; companies are seeing an average 35% ROI. While not quite at the 41% average "success threshold" CFOs have set, the gap is closing. AI is proving it can handle the heavy lifting in fraud prevention, cybersecurity and real-time transaction monitoring.
2. Valuations are Now Linked to AI Strategy
If you are considering a sale or seeking investment in the next few years, your AI roadmap is now a line item in due diligence. A staggering 97% of private equity firms say a successful AI strategy is an attractive trait in potential acquisitions. In the eyes of investors, AI is a marker of a modernized, scalable business.
3. The Pivot to Internal Expertise
Perhaps the most interesting trend in this sector is the shift away from external partners. Mid-market companies are increasingly bringing AI development in-house, with external partnerships dropping from 64% to 58% in the last year. This suggests that leaders are less likely to see AI as a "plugin" they buy from someone else, but as a core competency they need to own and nurture.
The Bottom Line for the New Year
As you wind down 2025, ignore the hype cycles and the doomsday headlines. Instead, look at the practicalities: Are your financial processes getting smoother? Is your data more secure? Are you building the internal muscle to manage these tools yourself?
The mid-market "winners" of 2026 will be like the CFOs in the Citizens survey, who are quietly integrated AI to drive measurable efficiency and long-term value. That, and they’ll have attended aiblLive of course.
Playbook of the Week
Why Your Agent Costs More Than You Think (And How to Fix It)
A 500 employee SaaS company built an agent to handle refund requests and in the demo, it looked great checking the policy, drafting a reply and even knowing when to offer a discount.
Then the CFO saw the bill. Each request was costing about £1.50 in compute to recover roughly £5 in margin. Worse, around one in ten replies misinterpreted the policy and that was enough to shut down the project.
Most firms think about agents like software subscriptions, but they behave more like digital workers. Every step they take has a cost and their reasoning shows up on the bill.
Treat them like software and you won't see the cost until it's too late. If you treat them like headcount, you can ask a harder question. Is this role actually paying for itself?
The real aim is simpler. Build agents where machine thinking costs comfortably less than the human time spent on the same problem. That gap is where the value appears.
News worth reading
We’re still collecting real-world wins, misfires and everything in between. If you’ve got a case study that deserves a spotlight, we want to hear it.
Drop a line to [email protected]
Internal friction carries a real economic cost. Workday suggests AI could lift productivity across Europe by around 1.5%, yet legacy ways of working and the EU AI Act together threaten to wipe out close to a third of that gain. Firms are being pushed to move faster while they’re still figuring out how to stay compliant.
For mid-market leaders, waiting it out isn’t realistic. Around 40% of employees need some level of AI training, yet only about 15% are getting it. Productivity gains won’t arrive on their own, especially when the wider economic climate leaves little room for drift.
This is where agentic maturity matters. The firms that pull ahead stop treating AI as a vague productivity boost and start using it to remove specific sources of operational drag. Miss that window and the value seeps away, swallowed by compliance work and the very friction AI was meant to reduce.
New research from Founders Forum Group and Microsoft shows founders moving past experimentation and into tougher operational decisions.
The basics still aren’t in place. One in three don’t trust their data or infrastructure, and security now outweighs cost and speed as the main concern. Around half are building in-house, but a third remain stuck in trial mode, unsure how AI fits into the business.
The clearest movement is on people. Sixty per cent are upskilling existing teams and reshaping roles around day-to-day human–AI collaboration. Governance hasn’t kept up. A third still struggle to formalise Responsible AI practices, even though they recognise the need.
For the mid-market, the problem isn’t the technology. It’s the basics.
When agent costs are the constraint, model choice stops being a technical debate and becomes an economic one. NVIDIA’s Nemotron 3 is built around that trade-off. Beneath the flagship models sits a Nano tier designed to handle the background work, like summarising, routing, and internal search. It’s the steady cognitive load that has to stay cheap and predictable if agents are going to scale.
The open setup matters because it gives teams more control over their data, avoids some of the black-box compromises of closed APIs, and leaves room to adapt as usage grows.
Nemotron 3 doesn’t simplify agentic AI in theory. It simplifies the economics. That’s what makes disciplined agent design viable. Adoption won’t follow benchmark scores. It’ll follow cost. Once an agent’s cheaper to run than a person doing the same job, decisions stop being theoretical.
Mid-market firms are writing big cheques for AI. The Baker Tilly 2026 Mid-Market Report puts average spend at over $600,000, with nearly one in five firms pushing towards $1m.
Seventy-six per cent say the goal is to improve efficiency and automate work. But the investment pattern tells a more cautious story. The survey doesn’t include an option for transforming how work gets done, and the biggest bets aren’t on new systems. They’re on enablers: letting employees use external tools and paying for advisory support.
That points to task-level automation. Until firms move beyond tools and start reshaping the work itself, AI will deliver a faster version of today, not a different future.
Product spotlight of the week
This week we’ve been digging into Ramp, a finance platform built to help mid-market firms move away from messy spreadsheets and disconnected expense tools. It brings corporate cards, bill payments and spend controls into a single system.
Where it stands out is how it uses AI in practice. In July, Ramp launched AI agents that work like an extension of the finance team. They review expenses, enforce policies, and flag issues, learning from past approvals to handle routine work on their own, with a clear audit trail.
It’s a good example of AI being used to solve a practical, everyday business problem. Smaller teams can get started with a free tier that includes their corporate card and core expense management features.
Quote of the week
“The question is no longer “Can AI do this?” but “How well, at what cost, and for whom?”” .