— 01
R&D spend ingestion
Pulls from accounting systems (Xero, QuickBooks, Polish standard systems) — payroll, contractor fees, direct expenditure. Produces the underlying spend dataset the claim will draw from.
TaxAI
R&D tax credit drafting, HMRC and KAS rubric aware. Same engine as GrantAI, different rubric — ingests your R&D spend, classifies qualifying activities, drafts the technical narrative, computes the relief, files the claim.
Pillar 05Planned Q3–Q4 2026UK RDEC/SME + Polish B+RShares engine with GrantAI
What it does
Capability surface
— 02
Qualifying-activity classification
Maps individual projects and cost items to HMRC / KAS qualifying criteria. Flags borderline activities for human review with explicit reasoning. No vague claims — the reasoning is auditable.
— 03
Technical narrative drafting
Drafts the technical description per HMRC 'BEIS Guidelines for Qualifying R&D' or KAS equivalents. Argues the scientific or technological uncertainty, the advance sought, and the systematic investigation undertaken.
— 04
Relief computation
Computes the relief value at current HMRC rates (UK SME, RDEC, Merged Scheme as applicable) or Polish B+R multipliers. Presents sensitivity analysis on claim value.
— 05
HMRC scrutiny-aware drafting
UK R&D claims are under HMRC scrutiny since 2024. TaxAI drafts with the inspectable, verified reasoning HMRC is now demanding. Not aggressive claim inflation — defensible claim drafting.
— 06
Filing and correspondence
Produces the final claim package in HMRC / KAS expected format. Tracks enquiry correspondence with the firm's own records of the claim reasoning.
Who it’s for
Buyer profile
R&D-active SMEs currently using tax consultancies
UK and Polish SMEs paying 15–30% success fees to R&D tax consultancies.
The displacement target. Firms spending real money on genuine R&D, currently outsourcing the claim preparation to consultancies that take a large share of the relief. TaxAI produces the same claim at a fraction of the cost, with your reasoning defensible and your claim history owned by you.
Accounting firms servicing R&D clients
Accountants with R&D-active clients who currently refer out to tax consultancies.
Bring the R&D claim work back in-house. Add a product line without building R&D tax expertise from scratch — licence the engine, apply your accountancy judgment to the final review.
How it’s built
Agent pipeline
Not a single prompt or a chatbot. A composed pipeline of specialised agents, each with a defined role, a documented toolset, and a declared failure mode.
agent/ingest-rd-spendPulls payroll, contractor, and direct-expenditure data from the firm's accounting system. Maps to claim categories.
agent/classify-qualifyingClassifies each project and cost item against HMRC/KAS qualifying rules. Flags borderline items for human reasoning.
agent/draft-technical-narrativeDrafts the technical narrative per the funder's guidelines (BEIS for HMRC, KAS for Polish B+R).
agent/compute-reliefApplies current rate structures. Produces the claim value with sensitivity analysis.
agent/qa-defensibilityHMRC-scrutiny gate. Audits the claim for defensibility against current UK R&D scrutiny standards. Flags vague claims or unsupported activity descriptions.
Status
Planned Q3–Q4 2026
Status
Planned Q3–Q4 2026. Architecture is inherited from GrantAI — the POC work carries across.
Starting markets
UK (RDEC, SME scheme, Merged Scheme). Poland (B+R ulga) shortly after.
Compliance
HMRC guidelines aware. KAS-aware for Polish market. Full reasoning trail for enquiry defence.
Code reuse
Significant architectural reuse from GrantAI. Same ingestion, rubric-parsing, drafting, and QA layers — different rubric and different funder.
Commercial model
Per-firm licence. Tiered by claim value brackets.
Register interest
enterprise@blackflake.com — be on the Q3 pilot list.
How we engage
Built for your firm. Developed with you.
Every deployment is a bespoke engagement. We don't sell seats or generic SaaS. We build agents against your workflows, your matter taxonomy, your regulatory context — then we operate the system with you under a monthly retainer. Your data, configurations, and outputs are yours and portable. The Blackflake engine is developed by us and runs under your retainer.
— 01
Discovery
Scoped call to map your workflow, constraints, and success metrics. No charge. You leave with an architecture note and engagement estimate.
— 02
Pilot build
Fixed-price deployment on one narrow workflow. Real output on real work in weeks. Go / no-go before the full build.
— 03
Full build
Agent suite, workflows, sources of truth, QA gates, operator surface. Deployed in your tenant. Auditable.
— 04
Retainer
Monthly retainer: new regulations ingested, agents tuned on live work, new workflows added. We develop the engine; your data, configurations, and outputs stay yours — always portable.