R&D credit co-pilot
Worth building — if a CPA will sign it
The pain is real and recurring: seed-stage teams leave the §41 payroll offset on the table because the paperwork costs more attention than the refund feels worth. The wedge is narrow and defensible for about eighteen months. What the score is missing is whether a licensed preparer will attach their name to a narrative a model drafted — that single answer moves this number more than any feature you could add.
Where the score comes from
The composite weights the first two axes most heavily. Regulatory exposure caps the score — it cannot be offset by strength elsewhere.
Strongest signal
Problem is real, at 88 — the highest any of your seven ideas has scored on that axis.
The question that decides this
Will a licensed CPA put their name on a §41 narrative that a model wrote? Ask five. If four say no without a human review step, the product is a workflow tool for preparers, not a filing co-pilot for founders.
Review details
What to build first
Four steps, in orderGet one CPA to co-sign a manual filing
Do the whole thing by hand for a single startup. If no preparer will attach their name, nothing downstream matters — and you will have learned it in a week rather than a quarter.
Payroll ingestion for two providers
Gusto and Rippling cover most of the seed-stage market. Pull employee records, wage totals and role titles — that is the raw material of the qualified-expense calculation.
Narrative drafting with a mandatory review gate
Generate the technical narrative, then block submission until a named preparer approves it. The gate is the product's legal posture, not a nice-to-have.
Form 6765 output and hand-off
Emit a filled 6765 plus a substantiation pack. Do not attempt e-filing in v1 — hand the pack to whoever already files the return.
How efficient it is
Model estimateMost of the cost is not the model — it is payroll integrations and the review workflow. Every week you spend on narrative quality before a preparer has agreed to sign is a week spent on the wrong axis.
Who else is doing this
See full landscapeOverlap score, 0–100 — a model estimate of how much of your wedge each company already covers.
MainStreet
Closest to the wedge. Sells the R&D payroll offset directly to early-stage teams and has done the payroll-integration work you would be starting.
Neo.Tax
Automation-first positioning that is the nearest thing to your stated pitch. Also the clearest evidence that a human review step is standard rather than optional.
TaxTaker
Advisory-led rather than product-led. Slower to serve each customer, which is exactly the gap a software wedge is supposed to exploit.
Boast.AI
Strong in Canadian SR&ED with a US practice attached. Overlap is partial because the buyer profile skews later-stage than seed.
Clarus R+D
Software-plus-service model aimed at small and mid-size businesses. Not startup-native, which is the positioning space left open.
Gusto
The platform risk. Already holds the payroll data and offers the credit as an add-on — if they deepen it, your integration advantage disappears overnight.
Brief roadmap
3 of 12 doneProve someone will sign
No product work until this is answered.
Make the data flow
Two payroll providers cover most of the market.
Draft, review, file
The review gate is the product, not a checkbox.
Sell it to the second customer
The first one is a favour. The second is a business.
This report is model-generated from a short description and public information. Scores, overlap figures and effort estimates are estimates. Verify anything you'd stake money on — this is not tax or investment advice.