Teach the AI to write like your firm.

5–10 reports give a robust profile. Each report extraction takes ~30 seconds; the merged profile builds in another minute. Multi-document merging reduces noise from any single document's quirks. More than 10 doesn't materially improve the fingerprint — diminishing returns set in fast.

Permanently — until you replace it. The profile is smartly cached during auto-mode runs, and the underlying observation records stay in your firm's account and can be rebuilt whenever you upload more samples.

No. Reports are sent only to extract your style fingerprint, and we use AI providers that do not train on customer content by default. The extracted profile is yours, used only by your firm.

The current implementation is one profile per firm. We've found firm-wide consistency is what most clients want — but per-surveyor profiles are on the roadmap if there's demand.

Yes — it's injected into AI Dictation routing too, and into the inline "Rewrite with AI" feature. Anywhere the system is generating text on your behalf, the profile shapes the output.

Yes. The Style Profile dashboard shows every dimension with its extracted value, the confidence score, and example sentences from your sample documents. You can disagree with a dimension and the system will weight it down on the next merge.

Yes — encouraged. The extractor only learns writing patterns; redacted property addresses and client names don't affect the fingerprint quality.

Phrases are discrete reusable text snippets you insert manually. The Style Profile is a broader voice fingerprint — vocabulary preferences, sentence structure, formality, depth. Both feed into Otto auto-mode independently; together they give the AI the most complete picture of your writing style.

Yes — and this is one of its most practical benefits. The profile captures how your firm narrates Condition 2 and Condition 3 observations: the hedging language, the urgency framing, the recommendation timing. "Immediate attention is advised" versus "early remedial action is recommended" — the profile picks up your firm's preferred phrasing and applies it consistently across the D, E and F elements of every report.

Yes. Upload a new set of sample reports and the profile is rebuilt cleanly. The previous fingerprint is replaced — there is no mixing of old and new voices. Many practices retrain after bringing on a new senior surveyor whose reporting style they want to formalise, or after a deliberate decision to modernise the firm's tone.

While the style profile is not a PI tool in itself, consistent language around Condition 2 and Condition 3 observations across all surveyors in a firm removes one source of discrepancy. When every surveyor uses the same urgency framing and hedging conventions — across the D, E, F and I sections — the firm's reports maintain a consistent level of professional diligence. This matters if a report is later scrutinised. The MRICS surveyor's final review remains the safeguard; the profile ensures the first draft already reflects the firm's standard of care.

Yes. The profile captures your firm's general writing approach — vocabulary, hedging level, sentence structure, urgency framing — rather than property-type-specific rules. When you dictate site notes for a pre-1900 terrace you will naturally use different observations from a 1980s detached; Otto applies your firm's house style to whatever content you provide. If your practice specialises in one property type, uploading sample reports from that niche will make the vocabulary fingerprint more precise.

One or two reports are better than nothing: the AI can still extract a useful fingerprint and apply it immediately. However, a single document can skew the profile — if one surveyor was particularly verbose or terse on that day, that quirk will be over-represented. We recommend uploading at least five reports before relying on the profile for client-facing work. You can start with whatever you have and upload more documents to refine the fingerprint over time; each new addition rebuilds the merged profile.

DOCX, PDF and plain text files are all supported. Anonymised reports work perfectly — the extraction reads writing patterns, not client or property details. Most past reports are already in DOCX or PDF format; no conversion is needed before uploading.

Yes. When you upload sample reports for training, the AI detects the RICS Home Survey Standard level from the structure and content of each document automatically — you do not need to label or tag files before uploading. The profile stores separate style guidance for Level 2 and Level 3 outputs, and when Otto generates a report it reads the guidance appropriate to that report type. If your firm produces both levels, upload a mix of sample reports from each; the profile will cover both and apply the correct voice to whichever report type is being drafted.

Yes — and this is one of the most immediate practical benefits for growing RICS-accredited firms. A new associate typically spends months absorbing the firm's drafting conventions through supervised practice. With the style profile active, the first AI-assisted draft they produce already reflects the firm's vocabulary, hedging conventions, condition-rating phrasing and report structure — before senior review. The profile compresses the written side of the RICS reporting learning curve without removing the senior surveyor's review responsibility. Many practices find that associates reach a consistent drafting standard significantly faster when working from a trained style profile rather than a blank template.

No — and it is not designed to. The style profile brings your firm's voice to the first draft, significantly reducing editing time for tone and vocabulary. But a qualified MRICS surveyor must always review and take professional responsibility for the final report. The profile reduces correction effort; it does not remove the surveyor's judgement from the process.

The style profile is optimised for RICS Home Survey Standard reports — Level 2 and Level 3 — where section structure, element codes (D1–D9, E1–E9, F1–F7), and condition ratings 1, 2 and 3 are the primary outputs. For snagging inspections, the AI Dictation tool and dedicated snagging report feature handle voice-to-text capture and issue logging through a separate workflow. If you produce both RICS home surveys and snagging reports, the style profile will shape your survey work while your snagging reports operate independently.

Every section — but with particular attention to the elements that most define your professional voice. The system extracts the exact opening sentence your firm uses for each type of inspection, the verbatim text from your Section B overall property opinion (the most personal section of any RICS report), the precise words you write after a Condition 1, 2 or 3 rating — including the phrase your firm uses when no defect is found — and how you structure defect descriptions from observation through to recommendation. These verbatim anchors, together with the broader 25-dimension fingerprint, are what make the resulting reports sound like your work rather than a generic RICS template.

With five or more reports, the initial training run typically produces a profile your surveyors will recognise immediately — the vocabulary choices, hedging style and condition-rating phrasing are usually clearly representative of your firm. Accuracy improves as more samples are added: each new document is merged using frequency analysis (words and patterns that appear more consistently across documents carry more weight), and the most detailed verbatim examples replace shorter ones wherever the same dimension is covered. The confidence dashboard shows you which dimensions are well-supported by multiple documents and which are based on fewer samples — giving you a clear signal on where to focus additional uploads.