R&D claim review

Open one claim, write its technical narrative against the right framework, and push it into approval. This is where claims actually get filed-quality.

The claim detail URL — /plan/[planId]/research-development/claim/[claimId] — redirects straight to the rationale writer at /plan/[planId]/research-development/claim/[claimId]/rationale. The old standalone claim detail page was a stub with no real data; the rationale writer is the actual workspace.

The rationale writer

Renders the unified RationaleWriter with rationaleType="rnd", scoped to a specific claim:

Param	Source	Default
claimId	Path param	—
projectId	?projectId= query string	Falls back to claimId
region	?region= query string	UK
planId	Path param	—

The writer is the same component the CapEx per-project review uses — the prompts and frameworks differ, the UX is identical.

Three columns

• Left sidebar — interview prompts. Adapts to the framework. UK SME relief asks different questions than US Section 174.
• Editor — the narrative draft. Margin comments from the AI panel surface alongside specific paragraphs. A draft-readiness gate prevents submission until the narrative meets the framework's minimum evidentiary bar.
• AI panel — generation progress, cost evidence summary (links the narrative back to the underlying salaries and contractor invoices), completion dialog.

Cost evidence

The cost evidence summary is the secret weapon. For every paragraph that asserts a qualifying activity, the AI panel surfaces the underlying employees and contractors with their effort allocation. When your tax adviser asks "show me how you got to that qualifying spend figure", the answer is one click away.

What stages a claim moves through

RECOMMENDATION  (AI candidate, before accept)
   -> DRAFT          (rationale being written)
   -> IN_REVIEW      (rationale submitted, waiting on approver)
   -> APPROVED       (signed off, ready to submit)
   -> SUBMITTED      (filed with the authority)
   -> CLOSED         (assessed and finalised)

The claim review page is most relevant during DRAFT and IN_REVIEW. After APPROVED, the rationale is locked.

What the writer does not do

• It does not file the claim. Filing is still a manual step with your tax adviser. What Flowstate produces is the narrative and evidence package they need to file.
• It does not adjust qualifying spend figures. Those come from the underlying allocations and effort data — change them in Workforce and the claim updates.
• It does not pick the framework. The framework is set when the claim is created.

Permissions

• View: any user with access to the org.
• Cost figures and qualifying spend: FINANCIALS_VIEW_SUMMARY.
• Edit and submit the narrative: claim ownership or one of the configured approvers — see approval workflow.

Related

• R&D recommendations — where claims get accepted from the AI candidate queue.
• Approval workflow — the gates a claim has to pass before submission.
• Archive — submitted and closed claims.