AI in a business banking product is usually a layer sitting on top of it, a chat window answering questions about data it has been handed. Finom, the European financial management platform, has built its assistant inside the business account itself, where it draws on the customer’s own balances, transactions, invoices, cards and accounting records.

Alex Gromadzki, head of product at Finom, on building AI into an SME business account, and what has to be true before it is allowed to move money.

Finom AI launched first in Germany, France, Italy and the Netherlands. The Fintech Times put written questions to Alex, on what the assistant does for a small business today, how much of it has to be rebuilt for each European market, and what has to be true before software is allowed to move a customer’s money on its own.

1. Finom has built AI into the business account rather than bolting it on top. What does that difference mean in practice for an SME user?

It means Finom AI can understand the user’s actual business context and give responses that are specific to that business, not generic finance answers. Because it sits inside the Finom account, it can work from the user’s real balances, transactions, invoices, cards, accounting signals and financial patterns, so the answer is based on what is actually happening in the business.

In practice, an entrepreneur can ask a plain-language question and get a relevant answer or suggestion without exporting data, uploading files or explaining their company from scratch. This will become even stronger as we roll out memory and background “thinking” capabilities, so Finom AI can remember business context, follow up on open items and surface useful suggestions even when the user is not actively engaging with it.

2. What specific tasks does Finom AI take on for entrepreneurs and small businesses today?

Today, Finom AI helps entrepreneurs and small businesses handle the everyday financial admin that normally means clicking through dashboards, invoices, cards and accounting screens. Users can ask in plain language, for example: “Which invoices are still unpaid?”, “How much did I spend on this card last month?”, “What was this transaction for?”, “How much came in and went out this week?”, “Which cards are active on my account?”, “What bookkeeping records or reports are ready for my accountant?”, or “What are my cashback and card benefits?”

We are also starting to move from answers to user-approved actions: some are already available, such as freezing a card, while others are in testing or development, including creating and sending invoices, sending invoice reminders, preparing payments from details or an uploaded invoice, paying a known contact, creating wallets and ordering cards. Because these actions touch sensitive financial workflows, they are built around user control: simpler actions ask for confirmation, while higher-risk actions require stronger authorisation such as 2FA.

So the short version is: today it helps entrepreneurs understand, find and interpret their business-finance data faster; the next step is moving from answers to approved actions.

3. You have launched first in Germany, France, Italy and the Netherlands. Why those markets, and what varies between them?

We started with Germany, France, Italy and the Netherlands because they are core Finom markets, where we have both a strong customer base and a deep understanding of how entrepreneurs already use the product.

But the bigger reason is that we do not believe business-finance AI can be generic: entrepreneurs do not run generic businesses; they operate within specific tax systems, accounting expectations, regulators and local financial habits. The underlying goal is the same everywhere: less admin, more clarity and more confidence about money. But delivery differs by market.

In Germany that means dealing with DATEV, Elster, VAT and detailed booking categories; in Italy it means the weight of the tax system and strong reliance on accountants; and in France and the Netherlands it means adapting to their own compliance rules, language and local financial-product expectations, so the product is not just translated but genuinely localised.

4. What are the risks of putting AI inside a business account, from errors to trust, and how do you manage them?

The main risk is that finance is a high-trust, low-error-tolerance environment: users may tolerate a weak chatbot answer, but they will not tolerate a wrong payment, incorrect tax categorisation, misleading cash-flow figure or unexplained recommendation. Customer research shows trust is not binary; users are comfortable fully delegating low-risk tasks like Q&A, feature discovery or document search, but want review or explicit approval for forecasts, categorisation, payments, tax matters and anything irreversible.

Finom’s approach is therefore gradual: start with information and recommendations, expose uncertainty instead of guessing, show the user what the AI is using, and require confirmation or approval before high-impact actions.

The clearest design principle from the research is “ask, don’t guess” and “start small, earn more”: trust compounds when the AI proves itself on low-risk workflows before autonomy expands.

5. What is next on the roadmap for Finom AI?

What is next comes down to two things: moving from answers to doing the work, and becoming a more proactive financial partner.

First, we are expanding the actions customers can take directly from chat, with confirmations and 2FA for sensitive workflows; including creating and sending invoices, chasing unpaid invoices with reminders, preparing and making payments, paying a known contact or an uploaded invoice, managing cards, and opening or organising wallets.

Second, Finom AI will increasingly work in the background, reviewing what is happening in the account and surfacing what matters without waiting to be asked: an overdue invoice worth chasing, a cash gap before a scheduled payment, money sitting idle that could be earning, or VAT and tax reserves to plan for.

The goal is simple: help customers spend less time on financial admin and more time on their business and with their families.

The post Finom on the Jump From AI that Answers to AI that Pays appeared first on The Fintech Times.