Will 2025 be remembered as the year the conversation around AI in finance truly took off or simply the year the noise became impossible to ignore? AI now dominates headlines, board agendas and investor conversations, creating a sense that finance teams should already be deep into implementation.
Yet when we surveyed more than 500 finance professionals about how AI is actually being used in practice, the results told a very different story. While the overwhelming majority acknowledged AI as an important, and often strategic, priority, only 37% reported actively using it in any capacity. That raised an obvious question: if belief in AI is so widespread, why is adoption still so limited?
Highlights from Scaleup CFO Webinar: Roadmap for 2026
In December 2025, we partnered with EmergeOne to host the Scaleup CFO Webinar, bringing together founder of EmergeOne and industry thought-leader, Aarish Shah and Joe Newbold, a highly experienced CFO with years of experience supporting investor-backed businesses. One of the central themes of the discussion was AI in finance, where they challenged prevailing assumptions and shared a practical, experience-led view on what is genuinely working today, what remains unproven, and how finance teams can adopt AI in a way that is grounded, controlled and commercially effective.
1. AI is already useful in finance, just not where most people expect it to be
“You can analyse term sheets, analyse contracts, and create narratives around business performance, but it’s not yet creating usable, investment-ready forecasts.”
There is a clear mismatch between expectations and reality. AI is not yet capable of producing investment-ready forecasts, management accounts, or robust financial models for fast-changing scale-ups. Tools that promise end-to-end automation of FP&A still fall short once complexity, judgement and iteration are required.
Where AI is genuinely delivering value today is in language-heavy and process-heavy areas:
- Contract and term sheet review
- Summarising legacy agreements across long customer histories
- Producing performance narratives alongside numbers
- Transaction processing, reconciliations and invoice workflows
- Capturing, transcribing and structuring meeting outputs and actions
In short, AI is strongest where finance work overlaps with text, repetition and volume, not where it relies on context-driven judgement or dynamic modelling
2. “Human in the loop” is not a temporary limitation – it’s fundamental to finance
Finance cannot tolerate “mostly right” outputs. The discussion reinforces that human oversight is not a transitional phase on the way to full automation, but a permanent requirement.
“Finance is not a function where you can afford to be 70% right. Unless you have absolute trust in the output, there will always be a human in the loop.”
Strategic finance decisions are driven by:
- Market sentiment
- Internal context and nuance
- Informal signals from customers, investors and teams
- Judgement built from experience rather than clean datasets
Even when AI is connected directly to systems like Xero or ERPs, it lacks access to this contextual layer. As a result, AI can support finance teams, but not replace decision-making at CFO or FD level.
3. FP&A and business intelligence remain the biggest unmet opportunity
“None of these tools are able to pull together even simple causal relationships. We still have to ask the right questions, exactly the same as we have for the last five years.”
Both speakers highlight a major gap: AI still struggles to identify causal relationships in complex, multi-variable datasets. Even relatively intuitive patterns – seasonality, regional behaviour, product mix effects, require humans to frame the questions before AI can assist.
This means:
- AI may accelerate dashboard creation
- It may reduce time to insight once questions are defined
- But it does not yet replace analytical thinking
For now, finance teams should assume that insight generation still starts with them, not the tool.
4. The real opportunity is workflow automation, not “killer apps”
“I don’t think it’s about looking for an app. It’s about building the workflows yourself, because every single business is unique.”
The discussion pushes back strongly on the idea that finance teams should wait for a single transformative AI product. Instead, the most immediate value comes from building bespoke workflows using tools like Zapier, Make (NAN), Replit and similar platforms.
Key points:
- Every business has unique workflows, even if the underlying tasks look similar
- Off-the-shelf tools often fail at the margins that matter most
- Micro-automations can remove hours of repetitive work with relatively low effort
Rather than asking “which AI tool should we buy?”, finance leaders should ask:
- What repetitive processes slow us down today?
- Where is manual reconciliation absorbing junior time?
- Which handoffs break under scale?
5. Start now because waiting for cleaner data or better tools is the real risk
“Build it in Lovable, build it in NAN, build something. If it makes your workflow easier, you can always rebuild it properly later, but get started.”
The survey data cited in the discussion is telling: most finance leaders see AI as critical, yet fewer than half are actively using it. Fear around data integrity, compliance and security is a major blocker.
The guidance here is pragmatic:
- Start small and low-risk
- Verify outputs manually at first
- Accept that no system is ever trusted 100% anyway
- Improve data quality as part of using the tools, not before
Delaying adoption does not reduce risk; it simply widens the skills gap as tools continue to evolve weekly.
6. Finance skillsets are shifting faster than job titles
One of the most important underlying points is that future finance leaders will be differentiated less by pure technical modelling ability and more by their ability to integrate AI into workflows.
“Five or ten years from now, it won’t be the people who know how to build a foundational model who are most valuable. It will be the people who know how to use AI tools in the context of their jobs.”
The emerging advantage sits with people who:
- Understand how systems connect
- Can design and iterate workflows
- Are comfortable experimenting, breaking and fixing
- Know how to prompt, validate and refine outputs
This is not about becoming an engineer. It is about knowing enough to make tools work in context.
7. Stop fixating on “AI” – focus on outcomes
“Does it save me time? Does it save me money? Preferably both. If it does, use it. It doesn’t matter whether it’s AI or not.”
A recurring theme is that “AI” as a label is becoming meaningless. Much of what is marketed as AI is incremental automation or algorithmic matching.
The buying criteria remain unchanged:
- Does it save time?
- Does it save money?
- Ideally, does it do both?
If the answer is yes, use it. If not, ignore the hype. Finance teams should remain outcome-driven rather than technology-led.
About our panel
Aarish Shah is an experienced CFO, thought leader and founder of EmergeOne, working closely with founders, CFOs and investors across high-growth and investor-backed businesses. With deep experience navigating the operational and strategic realities of scaling finance functions, Aarish brings a pragmatic perspective on where AI genuinely adds value — and where it currently falls short.
Joe Newbold is an experienced CFO and finance operator who has spent years inside fast-growth businesses, building and refining finance, FP&A and operational workflows. Known for his hands-on approach, Joe has actively tested many of the tools currently being promoted across the AI landscape, giving him a grounded view on what works in practice, what requires caution, and what remains aspirational.
