German tax advisory firm HSP Gruppe has developed internal AI capabilities to support client-facing tax advisory work, according to a case study published by OpenAI. The piece describes how the firm integrated AI models into its advisory workflow, though specific details on scope, cost and timeline are not fully broken out in the source material and should be treated as unconfirmed until HSP or OpenAI publishes further specifics.
Tax advisory is a useful test case for AI adoption because it combines two things that make automation hard: high compliance stakes and heavy reliance on specialized, frequently updated domain knowledge. Getting it wrong isn't a minor inconvenience — it can mean regulatory exposure for a client. That HSP Gruppe, a professional services firm rather than a technology company, chose to build capability in-house rather than rely solely on off-the-shelf software is the detail worth noting.
For the 10-200 person B2B companies that make up INITE AI's audience, this is a familiar bind. Most operators in this range don't have engineering teams dedicated to AI tooling. They have a handful of people — often the same ones handling sales, support and operations — who understand the workflows that could benefit from automation but lack the resources to build custom systems from the ground up. HSP Gruppe's case suggests that gap is narrowing: firms without large technical departments are increasingly able to stand up AI-assisted processes for research, document drafting and client communication using existing foundation models rather than commissioning bespoke software.
That said, professional services firms and typical B2B operators aren't identical. Tax advisory work is document-heavy and research-intensive in a way that maps reasonably well onto current AI capabilities — reading regulations, drafting summaries, flagging inconsistencies. A B2B company automating its support queue or sales pipeline is solving a different problem: routing tickets, qualifying leads, updating CRM records, generating quotes. The underlying technology may overlap, but the workflow design does not transfer directly.
What does transfer is the sequencing. Firms that succeed at this tend to start with a narrow, well-defined process — one with clear inputs and outputs and a way to check the output for correctness — before expanding scope. HSP Gruppe reportedly focused on specific advisory tasks rather than attempting to automate the entire client relationship at once. That's consistent with what tends to work in smaller B2B operations too: pick one workflow, like first-response drafting in support or lead qualification in sales, get it reliable, then expand.
The unconfirmed part of this story — cost, headcount involved, and how long the build took — matters for any operator trying to size up whether a similar effort is realistic for a company their size. Until those numbers surface, the case should be read as a directional signal that in-house AI capability is increasingly accessible to non-technical firms, not as a template to copy line for line.