Future CFO Talks Series: How Finance Leaders Are Applying AI Without Losing Control

Two weeks ago in Vilnius, Payhawk hosted the first Future CFO Talks event of 2026, an invite-only session for CFOs and finance leaders facing the same challenge: more control, faster decisions and less tolerance for inefficiency.

That challenge often shows up in ordinary moments. A department lead needs to order new laptops before a group of new employees starts next week. They’re thinking about delivery dates, supplier availability and getting people set up on time. They’re not thinking about procurement policy. But finance still needs that spend to be approved, visible and under control.

That gap between how people actually work and how finance needs to operate came up repeatedly in Vilnius during the fireside chat “Scaling Finance Operations: Lessons from Orbio World and NordSecurity,” featuring Ieva Terminaite, Procurement Tech Lead at NordSecurity, and Martynas Nenėnas, CFO at Orbio World.

The conversation focused on the practical changes finance teams are making to close that gap, from stronger purchase order discipline to automation that reduces manual work without weakening control.

The discussion also reflected a broader pattern seen in our latest CFO research. In a global study of 1,520 finance professionals, we found that AI adoption in finance is no longer simply early-stage; it’s uneven. Many teams are experimenting, but far fewer have the controls, accountability and workflow design needed to scale AI safely inside finance.

Control starts with behaviour, not systems

If there was one consistent lesson from Vilnius, it was this: better control starts with behaviour.

Ieva Terminaite, Procurement Tech Lead at NordSecurity shared what it took to roll out purchase order discipline across more than 2,000 employees, and why passive training quickly fell short.

We tried videos. Nobody watched them! You need to sit with people and explain why this matters.

Instead, the team focused on direct engagement — meeting teams face-to-face and tailoring the message to how they worked.

Alongside this, they introduced a clear policy: No purchase order, no payment.

“Once people realise the invoice won’t be paid without a PO, the conversation changes,” explained Ieva Terminaite. “Planning starts earlier.”

Within six months, the company reached 60% PO coverage, including adoption across engineering and marketing.

This is where many finance transformation efforts fall short. Tools are implemented, but behaviour does not change.

The practical takeaway for CFOs is simple: Automation is key, but you can’t automate what people don’t follow.

Why change management still determines success

Nikolay Pohlupkov, GM Accounts Payable at Payhawk, who moderated that session, pointed out that simply launching a tool isn’t enough to drive behavioural change, and that finance teams often underestimate the level of support users actually need.

New systems don’t really fail because of functionality; they fail because they’re just not adopted.

Different teams respond differently:

• Administrative teams often adopt first
• Marketing teams need flexibility
• Engineers require deeper explanations

One approach that worked consistently for Ieva Terminaite was to start with the most receptive teams, then expand:
“You build momentum with the teams that get it. Then you bring the rest along.”

This sequencing reduces friction and creates internal proof before scaling further.

Why spend control is moving earlier in the process

For fast-growing organisations, especially those with significant marketing spend, retrospective control is too slow.

The speakers agreed, ‘bad spend’ needs to be stopped before it happens, not cleaned up after.

This shift changes how finance operates completely, meaning:

• Approvals happen before commitments
• Budgets guide decisions in real time
• Visibility exists before money leaves the business

And this is also where true finance orchestration becomes critical. It’s not enough to connect systems. Finance needs to guide decision-making as work moves through the business.

Where AI is already delivering value

That same pattern came through in the AI discussion. The use cases that mattered most were those that eliminated repetitive work and aligned with existing approval, policy, and audit requirements. That matters because, in our second CFO AI readiness report, only 26% of self-identified AI leaders said they had the full operating conditions needed to scale AI confidently across finance.

In practice, the discussion pointed to three areas where AI is already delivering value.

Automated invoice processing

When invoices match approved purchase orders and budgets, they can be processed automatically.

“Users love it because they don’t have to chase approvals. Finance loves it because it stays within policy,” said Ieva Terminaite, Procurement Tech Lead at NordSecurity.

This reduces manual work while maintaining control.

OCR and data capture

Manual data entry remains a hidden cost in many finance teams.

As Martynas Nenenas, CFO at Orbio World described:

It sounds basic, but removing manual data entry saves accounting hours every week.

OCR allows invoices to be processed automatically, reducing both effort and error rates.

Policy enforcement

AI is also improving consistency in how rules are applied:

• Late invoices can be blocked automatically
• Non-compliant spend is flagged early
• Policy enforcement becomes systematic rather than manual

These are incremental improvements, but they compound quickly across high-volume finance workflows.

Why finance is adopting AI more cautiously

Compared to other functions, finance is moving more carefully with AI, and the reason is clearly not rocket science.

AI in finance does not just generate outputs. It influences decisions that must withstand audit, compliance, and accountability scrutiny.

Payhawk’s research highlights this gap, as even among AI leaders, skills are rarely the constraint, whereas rules are. The data revealed:

• 32.1% of leaders rate skills as strongly in place, but don’t have minimum rules in place
• While 21.7% report strong execution measures without minimum rules

In other words, skills and experimentation are widespread, but governable execution is not.

The real constraint: trust and delegation

This distinction was clear in Vilnius. Finance leaders aren’t asking whether AI works. They’re asking where it can be trusted.

As Nikolay Pohlupkov explained:

Most AI conversations still focus on what the technology can do. In finance, the harder question is what you’re prepared to delegate, and under which rules.

That question defines how far AI can go inside finance.

Today, most teams are comfortable using AI to:

• Analyse trends
• Support reporting
• Assist decision-making

But when it comes to autonomous purchasing or approval decisions, caution remains.

Why AI scale depends on operating design, not tools

A further insight from Payhawk’s research helps explain why so many finance teams see early success with AI pilots, but struggle to scale them across the function: AI in finance doesn’t scale because the model improves. It scales when the controls, ownership, and data foundations are strong enough to carry it.

That is the real threshold. In finance, scale depends less on the tool itself and more on whether the surrounding operating model can support it. AI needs clear rules on what it can do, defined ownership of outcomes, data that reconciles back to systems of record, and processes that can handle exceptions without creating risk.

Without those foundations, AI may still save time in isolated tasks, but it remains a productivity layer rather than becoming part of the finance operating model.

One area highlighted for future impact was bank reconciliation. As Martynas Nenenas noted, “If reconciliation is fully automated, you remove hours of work every month.”

That matters because the upside is not just efficiency. Automation at this level gives finance teams more time for planning, analysis and business partnering. Over time, that shifts finance away from operational processing and towards a more strategic role in the business.

What this means for finance leaders now

Across both the event discussions and Payhawk’s research, one idea stands out:

AI stalls in finance when it can’t operate within financial rules, approvals and audit requirements.

And that’s why many organisations feel stuck. They’re operating in the middle; active but not yet able to scale safely.

For CFOs, this reframes the problem around a few practical priorities. They must;

• Focus on behaviour before automation
• Design processes that people will actually follow
• Introduce AI where rules and data already support it
• Define clear guardrails before expanding automation
• Scale only where decisions remain auditable and accountable

This is how finance teams move from experimentation to execution.

Your next steps

If the discussions in Vilnius and our research point to one practical conclusion, it’s this: the next step is not to add AI everywhere at once. It’s to identify where it’s already influencing finance work, and whether those decisions can hold up under your existing rules, approvals and audit requirements.

If you want to assess where your own organisation stands, start with a simple check:

1. Where is AI already influencing finance decisions?
2. Which of those decisions could you defend under audit?
3. Where do manual workarounds still exist?
4. What would break if you scaled those workflows today?

Those answers will show you where to focus first, whether that means tightening policy, improving data quality or redesigning a workflow before scaling it further. That gives finance leaders a more useful starting point: where AI can operate safely, consistently and with clear accountability.
For a deeper look at what is helping finance teams scale AI safely, explore Payhawk’s CFO AI Readiness Report.