Loading content…
Francois Roux
CEO & Co-founder
Most accounting professionals approach AI with a reasonable level of skepticism. The tool sounds impressive in a demo. Real-world reliability is a different question. Robin Idevall came to Taxxa with exactly that skepticism and left with a clearer picture of when and why the tool earns trust.
Freedom Stockholm AB is part of the Freedom Group, one of Sweden's growing independent accounting networks. Robin works as a client-responsible consultant and also carries the IT responsibility for the firm which means he evaluates tools through both a practitioner lens and a systems lens. His use case is grounded and specific.
A client had entered into a car lease agreement. The situation was more complex than it appeared. Multiple shareholders were involved, each with strong opinions about cost fairness, what Robin describes as a demand for 'millimetre justice' in how company costs were allocated. The accounting question involved calculating the car benefit, determining the correct tax treatment, handling the bookkeeping entries, and then comparing gross versus net salary deduction methods to identify which approach was most cost-effective for all parties.
This is exactly the kind of multi-variable problem that takes time to work through cleanly. Each assumption changes the downstream figures. The client communication has to be accurate. There's no room for a rough approximation.
Robin uploaded the full lease agreement with all appendices directly into Taxxa. He described the shareholder structure and the requirements around cost allocation, then asked Taxxa to evaluate both deduction methods and produce calculation documentation supporting the comparison.
The output was substantively correct. But what changed Robin's view of the tool wasn't the answer itself, it was the source references attached to it.
This is a meaningful shift in how AI assistance works in professional practice. Without citations, a correct answer and an incorrect answer look identical. The professional has to re-do the verification work from scratch. With source references, verification becomes targeted: you're checking the primary source, not reconstructing the research path.
Robin also describes a collaborative dynamic with the tool: even when Taxxa's answers were correct, he was able to refine the conversation by contributing his own knowledge, steering the
model toward the right framing. He knew roughly what the answer should be before he started. Taxxa helped him get there faster and with documentation he could hand to the client.
Robin estimates the task would have taken 10-20% longer without Taxxa. That figure might sound modest, but it reflects something important: this wasn't a task he was struggling with. It was a task he was already competent to perform. Taxxa made a competent professional faster and produced a client-ready output along the way.
Without the tool, he would have done more manual cross-checking, consulted colleagues for a second opinion on the method comparison, and taken longer to formulate the client communication. The time savings compounded across those steps.
Robin's closing observation is worth noting for what it reveals about expectations. He was surprised to find that Taxxa could perform calculations and suggest bookkeeping entries that were close to real-world practice not just retrieve information, but work through numerical scenarios in a way that held up to scrutiny.
That gap between expectation and reality is one of the more interesting data points in how accounting professionals discover AI utility. The tool does more than most expect in some areas. It does less in others. Robin's experience is a good illustration of both.
Author: Robin Idevall, Client-Responsible Accounting Consultant & IT Manager · Freedom Stockholm AB - Freedom Stockholm AB | LinkedIn
