Access Request Screening

Access-to-information, ATIP, and FOI requests are all variations on the same problem: a requester defines a scope, and the public body has to find every record that falls inside it -- no more, no less. Get the scope too wide and you over-disclose, miss the deadline, or both. Get it too narrow and you risk a complaint and a re-do. The work itself is unglamorous: a coordinator (or a small team) reads thousands of records against a single paragraph of scope language.

Claira takes that paragraph, applies it consistently to every document in the collection, and returns a responsiveness call backed by the excerpts that drove it. Reviewers stay in the loop on the partials and the close calls -- which is where the judgment actually matters.

How Claira helps

• Applies the scope consistently. The same request scope is read against every document, eliminating the drift that creeps in when several reviewers interpret the language differently over a long review.
• Returns a structured call. Each document is tagged Responsive, Partially Responsive, or Not Responsive, with the scope element that matched and a short justification.
• Surfaces the partials. Mixed-content documents -- where some passages fall inside the scope and others do not -- are flagged for the redaction and exemption pass that comes next.
• Grounds every call in excerpts. The supporting text is quoted directly so the access coordinator can verify the call without re-reading the document end-to-end.

When to use this

• First-pass responsiveness review on a freshly assembled records collection
• Validating reviewer responsiveness calls before the file is finalised
• Triaging the gray-zone partial matches that need a coordinator's judgment
• Preparing a clean responsive set as the input for exemption review

A hypothetical request

The sample prompt below is built around a worked example so you can see what a useful scope statement looks like. Adapt it to your actual matter.

Hypothetical organisation. Regional Transit Authority (RTA) -- a public transit body responsible for bus, light rail, and paratransit service across a metropolitan area.

Hypothetical request.

"All records concerning the decision to discontinue Route 47 bus service, dated between January 1, 2024 and the date of this request, including internal emails and meeting records, ridership and financial analyses, communications with the regional or municipal authority, and consultation records with affected community groups, employee unions, or accessibility advocates."

The request is narrow enough to be tractable but broad enough that responsiveness calls require judgment: a board agenda that lists Route 47 alongside thirty other routes is not substantively about the discontinuation decision; a single sentence inside a long operations report might be. That ambiguity is exactly what the prompt has to handle.

Sample prompt

This prompt is structured to be lifted into any access regime -- ATIP, federal or provincial FOI, FOIA, or an equivalent statute. Replace the request scope with your actual request language and keep the rest of the structure intact.

Tips for better results

• Anchor the scope with concrete examples of what is out. The sample prompt explicitly excludes other routes, unrelated capital projects, and general budgeting. Negative examples cut the false-positive rate more than any positive guidance.
• Use Case Context for organisational shorthand. Internal acronyms, route codes, project names, committee names, and the names of relevant decision-makers all help Claira recognise responsive content that does not contain the obvious keywords.
• Treat Partially Responsive as the working bucket. That is where the redaction, severance, and exemption work lives. Filter to it after the bulk scan and feed it into the next stage of review.
• Run exemption review as a separate workflow. Personal-information, advice-and-recommendations, and privileged content are exemption questions, not responsiveness questions. Mixing them produces muddy calls. Use Privilege Review and PII Identification as follow-on passes on the responsive set.
• Pressure-test the prompt on edge cases first. Pick five documents you already know the answer for -- ideally one obvious responsive, one obvious not-responsive, and three close calls -- and run the prompt on those before scaling. If the close calls do not come out the way your coordinator would call them, the scope language needs sharpening.

Need help? Contact support@claira.to