A workflow accelerator is not a decision system.
BlueFlame AI accelerates the workflow around a deal — research, document Q&A, memo drafting, knowledge management. Capital Refinery answers the question that comes after: what happens to the IC decision the moment the deal is on the books, and is the case still defensible eighteen months later.
What BlueFlame does well, and where the line is.
What BlueFlame is built for
AI-driven workflow acceleration across the deal lifecycle.
Document Q&A across the diligence corpus, AI-assisted memo drafting, internal knowledge search across prior deals, and integration into the team's day-to-day workflow. For deal teams under time pressure during diligence and IC prep, it shortens the time from raw documents to a working draft.
What it is not built for
Continuous decision governance after the deal closes.
BlueFlame operates around the workflow. It does not maintain a structured investment record bound to the IC decision. It does not connect operator data flowing in post-close to the assumptions that justified the entry price. It is an accelerator, not a substrate. The IC memo it helped you draft is still a slide deck on a shared drive eighteen months later.
Where Capital Refinery sits next to it
On the other side of close.
Capital Refinery starts where BlueFlame's workflow utility ends. The structured investment record locks at IC. Operator updates are tested against the entry assumptions. Decision validity is a live score. The two systems are complementary — BlueFlame accelerates the work of producing the decision; Capital Refinery governs the position once the decision is made.
Where the lines fall.
| Capability | BlueFlame AI | Capital Refinery |
|---|---|---|
| AI memo drafting and document Q&A | Strong — core capability | Not the goal — uses deterministic extraction |
| Cross-deal knowledge search | Yes | Through structured investment record |
| Structured investment record at IC | No — outputs are unstructured artifacts | Yes — bound at approval with thesis and conditions |
| Operator data → entry assumption mapping | No | Yes — the data model |
| Decision validity scoring | No | Yes — live across the position |
| Time-to-consequence ranking | No | Yes — across the book |
| Provenance trail to source page | Citation-style, LLM-generated | Every figure has a candidate ID |
If you already run BlueFlame, that is the right call for diligence.
The natural next question is what governs the position after close. Bring us a deal you closed last quarter and we will show you on real data.