Guide · Published November 14, 2026 · Reviewed November 14, 2026 · Ingenious AI team · Not legal advice
ADMT Notices in Plain Language: A Writing Guide for Notices, Explanations, and Answers
The short answer: Plain language is not a style preference in these regimes — it is in the duty text. California's pre-use notice is a plain-language notice (section 7220); Colorado's adverse-outcome document is a plain-language statement of the principal reasons; Connecticut's interaction notice is a plain-language notice. The kit's own standard for its explanations is 'short sentences a non-lawyer can act on: the reasons named concretely, the data summarised honestly, and the next steps stated as actions with a contact'. This guide collects the techniques that achieve that — without diluting the legal content the duties require.
Check your duties and dates — free Build your first notice — free
What is plain language, operationally? It is writing whose intended reader — a person who did not choose to receive the document, mid-process, often on a phone — can answer four questions after one read: what happened, why, what data, and what can I do. Every technique below is tested against those four questions. The documents it applies to are the ones this series dissects: the pre-use notice and the adverse-outcome explanation.
Technique one: the reader's verbs for rights, the business's nouns for facts
State facts in stable nouns (the decision, the tool, the date) and rights in the reader's verbs ('you can ask what data was used; you can ask us to correct it; you can ask for a person to review the decision'). Verb-phrased rights convert legal content into instructions the reader can execute; noun-phrased rights ('correction and review mechanisms are available') describe a building the reader cannot enter.
Apply the conversion test to any sentence carrying a right: can the reader do something after reading it, and do they know how (the contact, the route)? If either half fails, the sentence is information about rights, not a statement of them — and the duties ask for the kind a person can use.
Technique two: one fact per sentence, reasons before labels
Long compound sentences are where notices hide their content. Split them: one fact, one sentence, in the order the reader needs them — what happened, then why, then what next. Inside reasons, lead with the fact and demote the label: 'the rent-to-income ratio on the verified figures was below the threshold for this property' carries its label ('affordability criteria not met') inside it as a shadow; the label alone carries nothing the reader can check.
This is the technique that makes corrections possible: a person corrects facts, and facts live in sentences that state them. A reasons section written in labels is a reasons section that generates confused phone calls instead of usable corrections — measurable, if you log contact reasons, within a month of sending explanations either way.
Worked example: Mile High Property Management
Example — fictional business, real duty structure. Mile High, the fictional Colorado property manager from the clock guide, applied the five techniques to its explanation template. One before-and-after pair shows the programme:
- Before (reasons paragraph): 'Your application was evaluated pursuant to our standard screening criteria, including but not limited to rental history factors and financial qualification metrics, and was determined not to meet the requisite thresholds at this time.' — 0 correctable facts; the paragraph generated phone calls asking what it meant.
- After: 'We decided not to offer the tenancy. The main reasons: an eviction filing from 2023 attributed to your file, and verified income below the stated threshold for this property. The screening tool produced a recommendation from your application and a credit file; a member of our team made the decision.' — 2 correctable facts, the tool's role, the human's role.
- Reader actions after the rewrite: correction contacts shift from 'what does this mean?' calls to specific corrections — in the following quarter, 14 of 17 correction contacts name the exact fact at issue, against 3 of 15 before.
- Four-question test: a colleague from the maintenance team, reading cold, answers all four questions correctly after one read — the acceptance standard the template now has to meet on every revision.
- Template governance: the rewritten text becomes the version of record; the read-aloud and four-question tests are added to the template-change process, with the footer left verbatim.
Technique three: numbers and dates stated as the reader experiences them
Dates in these documents are promises: 'within 30 days of the decision' should appear with the decision's actual date beside it, so the reader can verify the promise from the document. Numbers should be the reader's numbers — the income figure used, the threshold applied — not internal codes whose meanings live in another system. Where a figure comes from a source, name the source category ('from the credit file dated…') so the reader knows where a correction would be aimed.
Avoid the two characteristic failures: relative time ('recently', 'in due course') where a date exists, and precision theatre (a score quoted to three decimals) where the decision turned on a band. State what decided, at the resolution it decided at.
Technique four: the tool described as it behaves, not as it is marketed
Tool descriptions drift toward brochure language under no pressure at all: 'our advanced AI platform leverages hundreds of signals'. The plain substitute is behavioural: what the tool does ('scores applications using…'), on what ('…the information in your application and a credit file'), producing what ('…a recommendation a person reviews'). Three clauses, each checkable by the reader against their experience of the process.
This technique has a compliance edge: behavioural description is what California's logic element and Connecticut's purpose-and-nature fields actually ask for. A marketing description fails those fields while sounding richer; the behavioural one satisfies them while sounding plain. Plainness here is not the absence of content — it is the presence of only content.
Technique five: read it aloud, to the person it describes
The final test is performative: read the finished document aloud, once, at speaking pace. Sentences that run out of breath are too long; passages you stumble on will be passages the reader abandons; the point where you hear yourself become a lawyer is the point the reader stops trusting the document. Then run the four-question test with a colleague outside the team as the reader: what happened, why, what data, what can I do.
Keep the kit's boundary sentences intact while you plainen everything else: the footer (templates and records tool, not legal advice, does not judge fairness) is legal content whose precision is the point. Plain language is a discipline applied to the document's working parts, not a licence to paraphrase its disclaimers. Generate a document in the free generator and apply the five techniques to its structure — the builder already writes plainly, which makes its output the right practice text.
- Convert every right into the reader's verbs, each with its contact or route attached
- Split to one fact per sentence; lead reasons with the fact and demote the label
- Print real dates beside every clock promise and the reader's own numbers beside every threshold
- Describe the tool behaviourally — what it does, on what data, producing what — never in brochure language
- Finish with the read-aloud pass and a cold reader's four questions (what happened, why, what data, what next) — and leave the legal footer verbatim
Sources, as carried in the kit's own data: the duty summaries in app/rules.py and the checker data in app/checker_data.py, verified against primary or authoritative sources on October 28, 2026. Run the free duties checker for your jurisdiction, decision type, and the tool's role, and compare the kit's plans for the record, builder, and review queue. More from Ingenious AI is on the hub.
Related guides
- Colorado ADMT Pre-Use Notice Template: What the Notice Must Say, Line by Line
- How to Write an Adverse-Outcome Explanation: Reasons, Data, and Next Steps
- Colorado AI Act Automated Decision Notices: What to Send
- Adverse-Outcome Explanations in Plain Language
See all ADMT guides for the full set.
Frequently asked questions
Is plain language actually required, or just preferred?
It is in the duty text: California's is a 'plain-language notice' (section 7220), Colorado's a plain-language statement of principal reasons, Connecticut's interaction notice a plain-language notice, per the kit's data.
Does plain language weaken legal precision?
Applied correctly, no: it removes obscurity, not content. The required contents (tool, decision, reasons, data, contacts, dates) all survive; what goes is label-only reasons, compound sentences, and boilerplate. The documents' legal footer stays verbatim.
How do we test whether a notice is plain enough?
The four-question test: a reader outside the team, after one read, can say what happened, why, what data was used, and what they can do next. Plus the read-aloud test for sentence length and trust.
What is the most common plain-language failure?
Rights stated as nouns without contacts ('review mechanisms are available') and reasons stated as labels without facts ('criteria not met'). Both read smoothly and communicate nothing actionable.
Who should do the plain-language pass?
Someone who did not draft the legal content, reading as the recipient. The kit's templates are written to the standard already; businesses adapting them should re-run the tests on every adaptation.