Cardarelli Group case study spotlights AI agent for SR&ED claims

Cardarelli Group published a case study on a platform it built for GrowWise Partners, showing how an AI agent interviews engineers in more than 20 languages and helps prepare Canadian SR&ED tax credit claims. The system aims to cut client time, speed up claims work and keep every allocation reconciled to the cent.

Why it matters: - SR&ED claims are time-intensive and document-heavy, so automation can reduce the hours consultants and engineers spend on interviews, invoice matching and narrative drafting. - GrowWise Partners says the platform has helped the firm serve more than 100 Canadian innovators, file more than $20 million in claims and keep client time under five hours per claim. - The workflow is designed to preserve consultant judgment while automating the repetitive extraction work that often slows claims preparation.

What happened: - Cardarelli Group published a case study on Sept. 1, 2026 about the SR&ED claim platform it built for GrowWise Partners. - The platform's AI agent, Sam, telephones clients, interviews engineers in more than 20 languages and returns an English transcript for the consultant. - GrowWise Partners prepares Canadian Scientific Research and Experimental Development tax credit claims. - Version 3.4.6 reached production on Aug. 15, 2026.

The details: - The platform is built around an AI-and-human split: Sam handles information extraction, while a qualified consultant applies SR&ED criteria and makes the final call. - Sam calls at the scheduled time, retries if no one answers, and can pause and resume an interview across days without losing the thread. - The system supports French, Mandarin, Punjabi, Persian and Ukrainian, among other languages. - The client's transcript and the consultant's English transcript are displayed side by side. - Sam reads invoices and T4 slips and drafts entries for review. - A person confirms every match before it counts, and fuzzy matches remain suggestions rather than assignments. - Social Insurance Numbers are deleted before Sam reads a slip. - Project documents and completed interviews become searchable evidence within the claim. - The financial engine carries current federal and provincial rules and maps each rule to its legislative source. - The labour allocation export, filing file and accrual report use the same calculation and reconcile to the cent. - The platform blocks a claim from being filed if it is incomplete. - The case study says GrowWise serves more than 100 Canadian innovators, has filed over $20 million in claims and reports 98% audit accuracy. - GrowWise says the traditional SR&ED process typically takes 20 to 40 hours of client time per claim, compared with under five hours on the platform. - GrowWise also says it charges a contingency fee under 20 percent.

Between the lines: - The case study frames SR&ED prep as an information-extraction problem more than a tax-calculation problem. - That matters because the system can automate intake without replacing the specialist judgment needed to decide what qualifies. - The release cadence also signals active product development: 17 versions shipped in 141 days, across 1,797 commits, or about one release every nine days. - Each release added a new automation layer, including voice, multilingual recovery, pause-resume, invoice extraction, approval gating and word-limit enforcement.

What's next: - Cardarelli Group says the platform will continue to combine automated extraction with consultant review as it evolves. - The company made the full case study available online at the full case study. - GrowWise Partners and Cardarelli Group are positioning the workflow as a repeatable model for claims work that depends on interviews, reconciliation and evidence gathering.

The bottom line: - The case study shows how AI can speed up tax-credit prep without handing over the final judgment that determines whether a claim survives review.