In private beta
Trust Architecture

Trust is not a claim. It is the way Claima is built.

Claima separates live funding from research intelligence, explains every fit signal, and keeps evidence, AI, and human judgement in their proper place.

Source taxonomy

Five categories. Each handled differently.

Every funding record we read falls into one of these categories. The category decides how the record is surfaced — or whether it is surfaced at all.

  1. Live opportunity

    A funder-listed, currently-open application — surfaced as applyable.

  2. Upcoming opportunity

    A scheduled but not-yet-open scheme — surfaced as upcoming with the opening date if known.

  3. Closed opportunity

    An ended scheme — surfaced with its window stamped closed, not as live.

  4. Research intelligence / awarded-project intelligence

    UKRI Gateway to Research and similar — surfaced as research signal only, never as a live application source.

  5. Unverifiable / excluded source

    Sources we cannot trace to a funder — excluded from the catalogue.

Source treatment

Every opportunity traced to its funder source.

Every opportunity Claima surfaces is traced to its official funder source. The catalogue refreshes every 24 hours, and the date of the last refresh is recorded on each record.

Closed schemes and awarded-project records are kept structurally separate from live funding — they live in different parts of the catalogue and are never rendered as applyable. Anything the source taxonomy cannot verify is excluded rather than surfaced as live.

UKRI Gateway to Research is research intelligence — it is never shown as a live application source.

Cautious fit-signal labels

Seven labels. No verdict language.

Every fit signal Claima surfaces carries one of these seven state labels. The vocabulary is deliberately cautious — the funder decides eligibility, not the platform.

  1. Appears to match

    based on published criteria, the company appears to meet the threshold

  2. Evidence needed

    additional evidence is required to assess the criterion

  3. Potential gap

    a known criterion the company may not meet

  4. Source stale

    the underlying source has not been refreshed within the freshness window

  5. Reviewer action required

    the accountant should review before any client discussion

  6. Stack check needed

    interaction with other schemes or tax treatments requires advisor judgement

  7. Not assessed

    the criterion has not been evaluated

Decision logs

Every fit signal carries its source and its evidence state.

Every fit signal carries its source, its evidence state, and its uncertainty — the accountant sees the same trail Claima sees.

Decision log · example row
Scheme
Innovate UK Smart Grants R22
Company
Northwind Robotics Ltd
Signal
Appears to match
Evidence
Evidence needed
Source
Innovate UK Funding Service
Uncertainty
Accountant review required

Synthetic example · no real client identity

Evidence Vault boundaries

Reusable evidence. Accountant-reviewed. Never an auto-submit route.

  • Identity and registration

    company number, registered office, SIC code

  • Financial evidence

    accounts, turnover band, employee headcount

  • R&D activity

    project descriptions, advances claimed, contemporaneous evidence

  • Award history

    prior grant decisions and outcomes

  • Source-linked attachments

    funder-issued PDFs and confirmations, linked to their source

The vault is for reusable evidence, reviewed by the accountant, and is never a route to automatic submission. Claima organises and links what the accountant has gathered — the accountant decides what supports a claim and what does not.

AI-use disclosure

How Claima uses AI.

AI is a tool inside Claima, not the product. It helps with reading and structuring — it does not decide outcomes.

  • AI helps parse, organise, compare, and explain
  • AI does not decide eligibility
  • AI does not replace accountant judgement
  • AI output must be inspectable
  • Source-grounding required
  • Human review where judgement matters

The accountant remains the human in the loop.

Funding-advice boundary

Three roles. Three responsibilities. No overlap.

Funding has three actors and three responsibilities. Claima is one of them — not all three.

Role

Claima

Owns

Readiness · Fit · Evidence · Source · Date

Role

Accountant

Owns

Interpretation · Next action · Client judgement

Role

Funder

Owns

Eligibility decision · Award decision

Claima does not decide outcomes. The funder decides. The accountant interprets.

Human-review boundary

Reviewer-gated where judgement matters.

Claima is reviewer-gated where judgement matters. Accountant verification is required before any client discussion, and nothing on this page authorises any direct-to-customer eligibility verdict. The human stays in the loop.

  • Fit signals are routed to an accountant reviewer before any client-facing surface uses them.
  • Evidence-sufficiency calls sit with the accountant, not the platform.
  • The accountant remains accountable for what is shared with the client.
What Claima will never do

The shape of the commitment is the shape of the product.

  • We will not say you qualify; funders decide.
  • We will not claim every UK grant.
  • We will not show awarded-project intelligence as live funding.
  • We will not hide when evidence is missing.
  • We will not price on grant success.
  • We will not pretend AI replaces accountant judgement.
  • We will not treat drafting as v1 if it is not live.
  • We will not turn internal audit logs into customer-facing exports without a separate decision.
  • We will not expose reviewer notes or internal judgement fields in public outputs.
CLAIMA.AI / PRIVATE BETA
In private beta — invitations limited

Want to see how Claima earns trust structurally?

We’re speaking with a small number of accountants and advisers before opening access more widely.

See also: About · Privacy · Security · Terms