Most Canadian AI teams leave SR&ED money on the table because the R&D documentation was written months after the fact by someone who did not do the work. CFRI ships the AI system and the SR&ED-aligned R&D record together — technological uncertainty, hypotheses, experiments, and results, captured as the work happens.
Identifiable technological uncertainty, a hypothesis-driven experimental approach, systematic investigation, and a record that the work advanced the team's technological knowledge. We frame every AI engagement around those four pillars.
Each sprint produces a structured R&D log: uncertainty statement, hypothesis, experiment design, result, knowledge advancement. By the time the claim is filed the technical narrative is already written and dated.
When CRA asks questions, the engineers who did the work can answer them — because the documentation reflects what they actually did, not a reconstruction.
The R&D portions are when there is genuine technological uncertainty — novel retrieval methods, eval harnesses, agent coordination, or fine-tuning research. Routine integration work is not eligible.
We produce the technical R&D documentation. We partner with experienced SR&ED accountants who file the claim itself.
We can backfill the technical narrative if the underlying engineering evidence (commits, evals, decision records) is intact, and capture the remaining work in real time.
IRAP-funded R&D is typically not double-claimable as SR&ED for the funded portion, but unfunded sister work usually is. We map the boundary explicitly.
Yes — we structure documentation to support both federal SR&ED and Quebec-specific credits where the entity qualifies.
Email info@cfri.io with the project scope and the fiscal year-end.
Thirty minutes with operators who have already shipped what you're trying to figure out. CAD-billed. SR&ED-aware.