Q&A: Payhawk’s CFO Konstantin Dzhengozov on AI in Finance

With a background in management consulting and some experience in leading FP&A and investments, Konstantin Dzhengozov is one of three Co-Founders of spend-management platform Payhawk. 

The team has since elevated Payhawk to be the first company with Bulgarian roots to achieve a US$1bn valuation, or, unicorn status. 

Konstantin notes that his role as CFO focuses on “driving major growth” for Payhawk. 

“We recently announced that we've surpassed US$100m ARR with 159% YoY growth and a 75% increase in ARR per employee.” he explains. 

“With my experience spearheading capital efficiency, growth strategy, risk and sound unit economics at Payhawk, I can influence how our products are delivered and am also committed to helping companies build sustainable growth models.”

In this exclusive Q&A with Finance Chief, Konstantin explains how the office of the CFO uses AI, why some finance chiefs are hesitant to use it and of course that the dashboard is no longer the only way in. 

What trends have you noticed in the CFO office lately? 

This has to be how finance teams interact with AI.

We’ve noticed a hard behavioural shift in Payhawk’s own CFO customer base. A year ago, customers were asking for better dashboards and reporting. 

However, over the last two months, they've said they don't want the dashboards at all and want to connect to the system via MCP and ask Claude questions directly. Companies seem to be replacing a lot of the reporting layers they used, for example, even Power BI, with a more conversational interface.

We’ve also noticed that some of the problems you would expect to be easily solved with technology like AI are still a constant source of frustration. 

One of these is invoices. That is, accessing the company portal, retrieving the relevant invoice, extracting and coding its information, and matching it to the correct payment. All of this takes way more time than it should. 

This is why we’ve seen our new Financial Controller Agent’s invoice retrieval capability become our most adopted beta feature ever, with over 550 businesses activating it before official launch.

Could this be a remarkable shift in how CFOs use AI? 

Undoubtedly. The combination of AI, modern payment infrastructure and enterprise-grade systems is supercharging payments innovation. Finance teams are already changing how they operate daily compared with just a year ago. Over the next few years, that transformation will fundamentally change how finance professionals manage spend. 

With mundane everyday tasks, for instance, like chasing and collecting receipts, AI and automation will replace manual handling because finance teams want to save time. 

And these are significant amounts of time. The standard industry benchmark is 35 hours a month per person on reconciliation, roughly 420 hours a year, or close to a quarter of that person's working time, spent cleaning up transactions that happened weeks earlier.  

As CFOs stop treating AI as another tool to learn and begin understanding the real impact it has on productivity, AI will undeniably be interwoven into the finance fabric.

Is there potential for a shadow AI CFO to be present given the rise in conversational AI models in business ventures?

Not in the literal sense of an AI making CFO-level decisions. The more realistic risk is a repeat of shadow IT: finance teams using consumer AI tools because the approved alternatives are slower, less capable, or simply do not exist. 

That's already happening in most organisations, and it's a governance problem rather than a technology one.

The answer is the same as it was for shadow IT. If you don't give people capable tools inside your controls, they will use capable tools outside them. 

What matters is whether the AI your team uses operates within your permission structure, leaves an audit trail, and escalates decisions it shouldn't make alone.

So rather than a shadow CFO, the real risk is a shadow finance process: work being completed in systems where nobody can reliably see what data was used, what was recommended, who approved it or why the decision was made.

Will dashboards become a signifier for out-of-date finance professionals?

I'd be careful with such framing. Dashboards aren't going anywhere, and there's nothing outdated about wanting a consistent view of your numbers.

What's changing is their role. 

A dashboard answers a question somebody decided was important when it was built, often months earlier. That's useful for the metrics you track every week. It's useless the moment you need to ask something nobody anticipated, which is most of the interesting questions a CFO faces.

So the skill that ages isn't using dashboards. It's only being able to ask questions someone has pre-built a view for. 

Finance professionals can go straight to the data instead of waiting for someone to build them a view. Dashboards still matter – they're just no longer the only way in. 

Could BI progress more as innovation with AI/ gen AI progresses, the same way automation was adopted into BI a few years ago?

Partly, yes. Every BI vendor is adding a conversational layer, and some of it will be good.

But it doesn't resolve the underlying problem. Most BI tools sit on top of a warehouse fed by six or seven disconnected systems: ERP, expenses, procurement, payroll, travel and more. 

Putting a chat interface on that stack makes the questions easier to ask, but it doesn't make the data underneath any more connected, and it doesn't let the AI act on what it finds.

That's the distinction I'd watch. There's a difference between AI added to a reporting layer and AI that has native access to the transactions themselves, where it can see the invoice, the policy, the approval and the payment as one thing. 

The first gives you faster answers, and the second changes what finance can actually do.

What is your advice to CFOs hesitant to adopt AI? 

Any CFO hesitant to adopt AI probably has their fears grounded in a real concern, especially when it comes to security. Finance is still a high-risk and, as a result, risk-averse industry after all. 

But you can adopt the technology without loosening the rigour of the financial process, provided the AI works inside your permissions, approval thresholds and audit trail rather than around them. 

Our own agent is verified through Cloudflare's Verified Bots programme, so the sites it accesses can authenticate it as authorised traffic rather than an unidentified bot.

The real risk, though, is being left behind. Every day spent on manual work rather than automation is a missed opportunity to drive growth and strategic insight. 

Our Financial Controller Agent now gathers vendor invoices up to 12 times faster than humans. That's time back for work only a person can do.