Why your Otto reports sound like you: training on your writing style

The most common complaint surveyors raise about AI-drafted reports is that they all sound the same: generic, slightly formal, and unmistakably not theirs. An experienced surveyor has a voice built over years of practice, and a report that does not reflect that voice feels like something to rewrite rather than review.

We built StyleProfileService to fix that. Here is what it does and how it works.

The problem with generic AI output

Most AI writing models have a default register: clear, competent, slightly verbose. For RICS reports, that produces text that is technically passable but does not match how any particular surveyor writes.

Some surveyors are direct and use short sentences. Others prefer fuller explanations with more context for the client. Some consistently recommend specific tradespeople for certain repair types; others refer to "a suitably qualified contractor" throughout. Some abbreviate in particular ways; others do not. These are not arbitrary preferences. They are the fingerprint of a professional who has refined their communication over time.

If a surveyor has to rewrite AI-drafted text to match their voice, the time saving is eroded. The point is to reduce editing, not replace one type of editing with another. It is the same logic behind the way Otto v2 handles auto-mode drafting: the report should arrive closer to finished, not just faster.

What StyleProfileService does

StyleProfileService builds a writing profile from a surveyor's existing completed reports. When a surveyor connects past reports to Otto (uploaded as documents), the service analyses them to extract:

• Sentence length and structure preferences
• Vocabulary choices for common defect descriptions
• How the surveyor phrases recommendations and urgent items
• Patterns in how they describe specific construction types or defects they encounter frequently
• Consistent terminology they use across reports

That profile is then applied when Otto generates draft text. Rather than drawing on a generic register, Otto shapes its output to match the surveyor's learned patterns.

How it changes the editing experience

The practical effect is that the draft reads more like a first pass from an experienced colleague who writes the way you do, rather than a template that happens to contain your observations.

Surveyors who have used Otto with a trained style profile report spending less time on phrasing and more time on the substantive review: checking that the observations are accurate, that the condition ratings are correct, and that the advice given to the client is complete. That is where the professional judgement sits, and that is where the surveyor's time should go.

What training does not mean

Style training does not alter condition ratings, change the observations the surveyor dictated, or introduce new content. The profile shapes how existing observations are expressed, not what they contain.

It also does not train on sensitive client data in a way that is shared across surveyors. Each surveyor's profile is their own, which matters for the same reason we built a rating fidelity guarantee into Otto: AI should shape expression, never touch the substance of what a surveyor found.

Getting started

Style training works best with a handful of completed reports that represent your typical work. The more variety in construction type and defect complexity the better, since Otto can learn from a broader range of your phrasing rather than overfitting to one property type.

If you are new to Otto and do not yet have reports to upload, the default output is still useful. The style profile makes it better over time. For the reasoning behind why we built an AI report writer for chartered surveyors in the first place, see Otto v1: why we built an AI report writer for chartered surveyors.

Learn more at RICS surveyors on Home.

Further reading: RICS on home surveys.