Priceless lessons for CFOs

1. 1

   AI without ready data gets scrapped.

   Through 2026, Gartner expects organisations to abandon 60 percent of AI projects that lack AI-ready data.

   Gartner, February 2025Context ›

2. 2

   The ambition is everywhere. The foundation is not.

   87 percent of North American CFOs expect AI to be very or extremely important to their finance team in 2026. Yet when Actuals read the open finance roles of 10 large Dutch e-commerce companies, only 2 were hiring people to build data and systems into finance.

   Deloitte CFO Signals, Q4 2025; Actuals Research, June and July 2026Context ›

3. 3

   Look at who a company hires, not at what it says.

   Hiring shows the plan long before the figures do. 7 of the 10 companies hire to keep the current machine running. 1 is barely hiring in finance at all.

   Actuals Research, June and July 2026Context ›

"I wish you all good luck."

1. 1

   Leadership has to lift AI from islands to the whole team.

   Across Smile Sail's portfolio everyone knew about AI, but people were figuring it out "on their own little island". That is hard to scale without leadership stepping in.

   Keynote, 07:05Context ›

2. 2

   Expect people to drop AI when it gets busy.

   When Ronald's own team got busy in May, they stopped using AI. In June and July their use went up again. Those were the months they trained together and built their own agents.

   Keynote, 10:01

3. 3

   Name champions and make building fun.

   Make it a CEO topic, name 1 AI champion at C-level, let one grow in each team, and run hackathons. It's not about what people build, but how they build it.

   Keynote, 22:53Context ›

4. 4

   Count the 10 times, not the first time.

   Don't judge an agent by the effort of building it once. Look at the saving over the 10 times you do the task, such as accrual estimates, collections or credit-risk signals.

   Keynote, 33:44

5. 5

   Your team owns what it sends you.

   AI can produce endless output, so reviewing it is where people struggle. In a test by a vendor, the same request went to 3 different AI models. The specialised data tool scored 85 percent; the others scored 25 and 50 percent. Even then, Ronald said, your team still needs to own the outcome.

   Keynote, 08:13 and 37:22Context ›

Quotes to share

1. 1

   Connecting an accounting system to AI is harder than it looks.

   Ask Exact's API for the open items of 31 December and you get today's. You need to know details like that. André expects many people to try connecting their books to AI themselves, and he says it is no simple job.

   Session André and Jennifer

2. 2

   Put your expertise into the agents.

   André built his bookkeeping knowledge into Tibata's agents. That is why the system knows what a trial balance is.

   Session André and Jennifer

3. 3

   15 sets of books, checked in a single Saturday.

   Last Saturday André checked 15 sets of books, for his son and people in his neighbourhood, and found the errors. They get a list of what to change, fix it, and he runs it again. He estimates it saves them thousands of euros in accountant fees.

   Session André and Jennifer

4. 4

   AI literacy is now part of risk management.

   In the session they pointed out that since August this year, the AI rules come with supervision. If you can't show your team is AI literate, that needs attention. AI literacy belongs with compliance, with clear frameworks and responsibilities.

   Session André and JenniferContext ›

Quotes to share

1. 1

   A payment and the money in your bank are 2 separate steps.

   When a customer pays, they no longer owe you the money. Your payment provider does, until settlement moves it on to your bank.

   Breakout, 00:08 to 00:12Context ›

2. 2

   Record every payment event, then work out the status.

   A stored status like "paid" can change, and you can miss events. Log every payment and refund, and you can show that nothing is missing.

   Breakout, 00:24

3. 3

   Your payment provider is a black box.

   Between what customers pay and what reaches the bank sit refunds, chargebacks, fees, currency conversion, reserves and holds, each with its own timing. As your volume grows, a provider may ask for bigger reserves.

   Breakout, 00:25 to 00:29Context ›

4. 4

   Use your payment provider's balance report for accounting.

   It works like a bank statement per currency and shows every flow. Settlement reports come late, and payment reports miss later fees. Not every provider offers one, so ask.

   Breakout, 00:29 to 00:33

5. 5

   Agree which truth you are talking about.

   Your shop, your payment provider and your bank each see different open items. The shop view runs on webhooks, which usually arrive, "but sometimes they don't". Miss one and the statement is wrong.

   Breakout, 00:38 to 00:39Context ›

Quotes to share

1. 1

   Design from the business event.

   Marinda starts with each event in the business and asks what it means financially. Like drilling a tunnel in Switzerland, she works from both ends: the event on one side, the report you need on the other. Then she connects the middle.

   Panel, 02:28 and 13:40Context ›

2. 2

   Report the exceptions, not the totals.

   Patrick wants every amount on an invoice, even a 10 euro line, linked to the exact transaction behind it. Once that holds, the monthly report only shows what you didn't expect.

   Panel, 11:10Context ›

3. 3

   Put finance at the start of the process.

   Today finance collects data at month-end and repairs mistakes made earlier. Patrick turns it around: what does staying in control mean for the steps before the report?

   Panel, 12:37Context ›

4. 4

   Agree what your words mean.

   Uwe's team built a shared company context, so that revenue, cancellation and refund mean the same to everyone. "When someone says revenue, everyone knows what revenue means."

   Panel, 22:53

5. 5

   You don't always need one big ERP.

   At Goboony and Yescapa the settlement logic lives in the platform, which keeps changing. Their choice: a small ERP per entity, tooling to consolidate, and their own BI on top.

   Panel, 29:47Context ›

6. 6

   Take the big bite.

   Book big blocks of time for change, in a regular rhythm. Otherwise the backlog keeps growing. You are building the track while the train is running.

   Panel, 15:51 and 17:02

Quotes to share

1. 1

   Hiring more people doesn't fix it.

   As payment methods, entities and currencies grow, the reflex is to hire. Companies still got stuck: the close took longer, open items piled up and balances were written off.

   Launch, 43:58Context ›

2. 2

   A gap tells you nothing until you know what is in it.

   A gap of 40,000 euros tells you nothing on its own. It could be a late payout, partial refunds or a bank transfer someone entered by hand. You need to see which.

   Launch, 45:41

3. 3

   The value is hiding in the data.

   Helloprint handles more than a million order lines a year from over 200 suppliers, each with its own formats. Matching every invoice line against the purchase order surfaced more than 500,000 euros in net overcharges in 2025.

   Helloprint, slides 73 and 75Context ›

4. 4

   Let AI agents work with your finance data.

   Actuals is working on MCP, so AI agents can talk to the platform, for example to sort out unclear open items. Peter Lem calls it a roadmap item. Actuals already started on it together with Helloprint.

   Launch, 52:12

Quotes to share