HSP Gruppe and partner network Kanzleipakt logged 500,000+ ChatGPT Enterprise conversations across six months and 81 organizational groups. Employee productivity increased 98.6%; weekly active usage hit 84%. The network estimates 40,000 additional annual capacity hours unlocked.
| Metric | Value |
|---|---|
| ChatGPT Enterprise conversations logged | 500,000+ |
| Deployment period | 6 months |
| Organizational groups covered | 81 |
| Employee productivity increase | 98.6% |
| Weekly active usage rate | 84% |
| Additional annual capacity hours unlocked | 40,000+ |
HSP deployed two Agents on ChatGPT Enterprise. AI Client Communication scaffolds first drafts and enforces consistent, client-oriented tone across correspondence. Booking Assistant SKR03 & SKR04 classifies and prepares booking inquiries. Neither agent outputs final work—professional review is mandatory before client-facing delivery. Legal liability stays with the licensed tax advisor or lawyer.
| Agent Name | Function | Scope | Output Handling |
|---|---|---|---|
| AI Client Communication | Scaffolds first drafts; enforces consistent, client-oriented tone | Client correspondence | Mandatory professional review before delivery |
| Booking Assistant SKR03 & SKR04 | Classifies and prepares booking inquiries | Booking workflows | Mandatory professional review before delivery |
Governance is deliberate. ChatGPT Enterprise's built-in security forms the baseline; HSP layers its own data protection policies for client information. Monthly forums surface new use cases for evaluation and standardization, preventing institutional knowledge from staying siloed to individual power users.
Examples: Partner Magdalene Posnak reduced real estate investment analysis from nine hours to two. Managing Director Marco Sell runs AI-powered dashboards for live financial scenario modeling during client meetings. Managing Partner Frank Heibel uses the model as a technical sparring partner for tax questions. In each case, AI surfaces information; the professional does interpretive and liability-bearing work.
This model avoids headcount cuts—a regulatory risk in professional services. Kanzleipakt was operating at capacity with backlogs. Time savings translate to more client engagements, faster turnaround, and higher-margin advisory work.
The binding constraint is organizational absorption speed, not model capability. CEO Carsten Schulz: "Technology is moving faster than organizations can transform. Our challenge wasn't introducing AI—it was helping the organization absorb it." Monthly forums, mandatory review gates, and deliberate rollout per agent address the governance problem the model itself cannot solve.
HSP is piloting ChatGPT Work, OpenAI's agentic platform, with a small group before broader rollout. The goal is mapping how agentic automation interacts with existing governance, quality controls, and per-task cost. Year-end accounting workflows are the redesign target.
Viable deployment in regulated professional services requires: Enterprise-tier model with internal governance layered on top, agents that prepare rather than decide, and mandatory human sign-off at every client-facing output. Agentic expansion follows only after the governance loop is stress-tested at scale.
