Loading content…
Francois Roux
CEO & Co-founder
There is a version of the AI story that accounting professionals have heard too many times: a tool arrives, promises to replace expertise, and ends up producing confident nonsense.
Heli Kemppainen, CEO of Tilitoimisto Helpp Oy, has used Taxxa long enough to have a more nuanced view and she is candid about both sides of it.
Tilitoimisto Helpp is a Finnish accounting firm serving small and mid-sized businesses. Like most firms of its kind, the team operates across a dense mix of daily tasks: drafting client communications, interpreting legislation, working through edge-case scenarios, and verifying figures across financial reports. Each of these is manageable on its own. Combined, they consume a disproportionate share of a senior professional's working day.
Heli came to Taxxa with a clear set of challenges. She wasn't looking for automation in the factory sense; she was looking for speed and confidence across tasks she was already doing herself.
The first area was text formatting and document drafting. Reviewing and polishing communications takes time that rarely shows up on a client invoice. She now feeds anonymized background context into Taxxa and asks for the document she needs then applies her own finishing layer. The result lands in a fraction of the time.
The second was legislation and case interpretation. Finnish accounting law is detailed, frequently updated, and full of edge cases. When a client situation doesn't map cleanly to a standard treatment, Heli uses Taxxa as a thinking partner: she provides the background, challenges the output with follow-up questions, and uses it to surface alternative solutions and source references she can then verify herself.
The third and fourth areas were financial analysis and data verification. Taxxa can ingest attachments, reports, ledgers, comparative figures and return structured interpretations. For large volumes of data where a discrepancy needs to be found, this replaces what would otherwise be manual line-by-line review.
The fifth was collective labour agreement (TES) interpretation, which is where Heli's experience gets more critical and more instructive.
"Taxxa doesn't do anything I couldn't do myself, but it significantly speeds things up."
That framing AI as an accelerant for existing expertise, not a replacement for it is at the heart of how Heli describes the tool's value. She is deliberate about the distinction. Taxxa is useful because she can already evaluate what it produces. She can spot when a source reference doesn't hold, when an interpretation has drifted, when the model has overreached.
And it does overreach, sometimes. On TES interpretations specifically, she found that Taxxa would pull from the wrong collective agreement and couldn't self-correct even when prompted. Her recommendation is direct: this class of task needs a different approach until the sourcing is more reliable.
This kind of feedback is valuable precisely because it comes from someone who has integrated the tool into daily practice rather than evaluated it in isolation. She isn't dismissing AI. She is describing the shape of its usefulness which is more than most assessments manage.
The deeper shift Heli describes isn't about any single feature. It's about cognitive load. When a professional spends less time on text cleanup, lookup, and first-draft analysis, they have more capacity for the judgment calls that matter the interpretation decisions, the client conversations, the scenarios that don't fit the standard mold.
Taxxa's value, in her account, is largely invisible on any given task. It doesn't replace the professional. It reduces the drag on their time so the professional can show up more fully for the work that can't be templated.
For firms considering where AI fits into their workflow: Heli's experience suggests the question isn't whether the tool is perfect. It's whether the people using it have the expertise to direct it well and the judgment to know when it's fallen short.
