Funder reporting

The operational job: assembling a per-funder report across the CRM, the MEAL system, the finance ledger, and four years of prior submissions - every material claim sourced back to the data row it came from, so that no trustee or auditor can ask "where did this number come from?" and not be answered immediately.

The shape Loop takes here is data assembly into structured tables, narrative drafting per funder template, and a claim-source chain - click any sentence's source badge, the page jumps to the originating row. Four-stage internal sign-off (drafter → grants manager → head of programmes → director) baked in as a hard requirement: rejecting at any stage abandons the chain; correcting threads forward.

How we'd shape it

Funder-reporting deployments start with a session with whoever currently writes the reports. We watch them assemble one - usually a painful three-day job - and read their last submitted report alongside the spreadsheets, exports, and inboxes it was built from. The discovery output is the template - what the funder asks for, what your team currently provides, what the audit gap looks like.

What gets built and stays running

We'd build:

• Data assembly that pulls from the CRM + MEAL + finance + risk register into structured tables, preserving lineage
• Narrative drafting per funder template, with every material number tagged to its source row
• The claim-source chain - every sentence in the drafted report carries a chip linking back to the row, decision, or document that justifies it
• The four-stage sign-off pipeline with role-aware queues + decision history
• A case-study screening surface flagging consent boundaries and de-identification concerns before stories leave the building

Funder templates change. New funders come in. The first deployment shapes the engine; subsequent funders are configuration, not new builds. We hold the configuration alongside you and bring in new templates as they arise.

Where this stands today

Built end-to-end against a synthetic £32m homelessness charity reporting to four different funders. Real funder PDFs extracted at 83% headline-field recall. Zero invented citations across 94 claims in two test cycles. Ready for a charity that wants to land it on a real CRM + MEAL + finance integration.