When I joined Altitude in 2025, the company had a working rules-engine product and a thesis that LLMs would let it serve more conditions, more workflows, and more enterprise customers. The pivot from a rules-based core to an LLM-and-agent core was the year's defining technical bet.
The pivot took longer than a refactor and shorter than a rebuild. The shape of the work, in order:
- Stabilize the existing rules engine. Don't refactor what you're about to replace. Freeze it where you can. Triage incoming work into "do in old system" and "wait for new system."
- Design the new substrate end to end. Skill files, retrieval, governance, evaluation, logging. On paper, before code.
- Migrate one scenario. The simplest, least-clinical, lowest-stakes one. Build the substrate around it.
- Migrate three more in parallel. Same substrate. Each scenario shakes out a different gap.
- Reach pivot threshold. Half the volume on the new substrate, golden cases passing, governance running. The rules engine is now a fallback, not the system of record.
- Sunset the rules engine scenario by scenario, on a schedule the business can plan around.
The full playbook — including the org changes (who owns what during transition), the budget shape, the milestone cadence, and the way to communicate the pivot to the board — is in the Healthcare AI Automation Playbook.
If your team is staring at a similar pivot, book a call. I'll tell you what's two months of work and what's a year of work, on the call.