Teach it your standard.
Every firm has a standard — it lives in its best reports and its most experienced heads. Upload the reports, press Train, and the AI learns how your firm reports a defect, words a condition rating and phrases a recommendation. Every auto-mode draft starts from it.
How training works
Your seniors' red pen, encoded.
When a senior surveyor marks up a junior's draft, they're teaching a standard: what a rating 2 sounds like, when "monitor" becomes "before exchange", how long a section should run. Training does the same job on the AI — from your documents, on your command.
Upload the reports you're proud of
Pick your firm's best recent work — PDF, DOCX or DOC, up to 25MB each, ten slots per firm. Upload each document exactly as it was delivered to the client; there's no template and no reformatting. Tag it with its report type so Level 2 and Level 3 train separately, and choose whether it teaches the shared firm-wide profile or a personal one.
Press Train — nothing happens before that
Uploading a document does nothing on its own. You decide when the AI reads — one document at a time or Train All in one go — and each run reports its progress live, section by section, so compute is only ever spent deliberately. A failed read (a scanned PDF with no text layer, say) is flagged with a plain-language reason.
The engine reads like an apprentice
The extraction step doesn't skim — it studies. It records 25 writing dimensions across five categories (structure, tone, vocabulary, patterns, formatting), lifts verbatim examples of how you write a condition rating 1, 2 and 3, captures your standard no-defect phrase, keeps exemplar passages per RICS section and measures how long your sections actually run. Every run is logged with the sections it detected and a confidence score.
The learning merges into your firm's profile
Each trained document contributes one set of observations, and the observations merge into a living profile for your organisation — level-aware, so your Level 3 standard never dilutes your Level 2 style, and scope-aware, so a personal profile can sit alongside the firm's. The same corpus powers your Style Profile, the voice side of the equation.
Every draft starts from your standard
From then on, the profile is injected into every Otto auto-mode run and every AI dictation session for your firm. Defect language, rating caveats, recommendation phrasing, section length — the first draft arrives already shaped like your work, whichever surveyor recorded the inspection.
The apprenticeship problem
A firm's standard lives in three fragile places.
On shelves, in margins, in heads. It takes years to build and one retirement to lose. Training moves it somewhere more durable.
1 · On the shelf
Years of judgement, bound
Every report your firm has ever delivered is a worked example of its standard. Filed, finished — and teaching nobody.
2 · In the margins
The red pen, one draft at a time
Seniors pass the standard on by correcting drafts — the slowest, most expensive teaching hour in the practice.
3 · In one head
Carried, until it isn't
The sharpest version of the standard usually belongs to one principal. Train it into the platform and it stops being a single point of failure.
Nothing is a black box
Visible learning. Checked writing.
You can inspect what each document taught, and every draft written with your standard passes a reviewer step before it saves. Both ends of the loop leave a record.
“Slipped slates were noted to the front elevation…”
“Obtain quotations from a suitably qualified contractor…”
Logged in the observations panel for this run
What one document teaches
Every trained report contributes verbatim exemplars — defect passages per RICS element, your condition-rating language for 1, 2 and 3, your recommendation style and your measured section length. The run is logged with sections detected and a confidence score, so a weak read is visible, not silent.
Draft
“Severe widespread damp throughout the whole wall.”
Corrected · matches what was dictated
“Elevated moisture readings were recorded at low level to the rear wall.”
Before/after logged · you still review and sign off
Every draft is checked against your words
After routing, a reviewer step re-reads each auto-mode draft against the original dictation — coverage, routing and fidelity — and applies corrections before the report saves, with a before/after log. It keeps drafts honest to what you actually said; your trained standard changes only when you press Train.
The training loop
Five stations. One turning wheel.
Training isn't a one-off setup step — it's a loop your firm turns whenever the standard moves on. Scroll through the stations and the wheel turns with you.
The version counter is illustrative — in the product your profile simply rebuilds from the current corpus each time you train.
Station 1 · Upload
Feed it your best work
Ten slots per firm for the reports that define your standard — PDF, DOCX or DOC, up to 25MB each, tagged Level 2 or Level 3, firm-wide or personal. Nothing is read yet; the shelf is just stocked.
Station 2 · Extract
Press Train — it studies the documents
On your command — per document or Train All — the engine reads each report and records 25 writing dimensions, verbatim condition-rating examples for 1, 2 and 3, your no-defect phrase, per-section exemplar passages and measured lengths. Progress streams live while it works.
Station 3 · Merge
Observations become one profile
Each document's observations merge into a living profile for your organisation. Level 2 and Level 3 stay separate so a Building Survey never dilutes your HomeBuyer voice, and personal profiles sit alongside the firm's with their own privacy.
Station 4 · Apply
Every draft starts from it
The profile is injected into every Otto auto-mode run and AI dictation session — so defect language, rating caveats, recommendation phrasing and section length follow your standard from the first draft, whoever recorded the inspection.
Station 5 · Review & go again
Refine the corpus, close the loop
Check the observations log, generate a sample section beside the passage it learned from, and tune the adherence dial. When the standard moves on, add newer reports, delete stale ones, press Train — and the wheel comes back round to the top.
The version counter is illustrative — your profile simply rebuilds from the current corpus each time you train.
Judgement is not voice
Substance, voice, language — one standard.
Voice makes a draft sound like your firm. Substance makes it say what your firm would say — which defects get named, how a rating 3 escalates, when "monitor" becomes "before exchange". Three tools carry the standard; this page is the substance.
Voice · how it sounds
Style Profile
The writing fingerprint — a 25-dimension read of structure, tone, vocabulary, patterns and formatting, built from the same documents you train here.
Language · the exact words
Standard Phrases
Your pre-approved phrase library — element descriptions and recommendation wording inserted verbatim, never paraphrased.
Substance · what it says
AI Training
The training corpus and the judgement inside it — defect language, ratings philosophy and recommendation style, managed, reviewed and cleared from one screen.
All three are injected into every Otto auto-mode run — which is why firms that train arrive at review-ready drafts, not generic ones.
The control panel
Training you can see and steer.
Everything the AI takes from your reports is visible, adjustable and removable — from one screen in your surveyor account.
Two training scopes
Train a shared firm-wide profile, or let individual surveyors train personal ones. Personal documents stay private to their owner — colleagues can't open them or delete them.
Train on your terms
Uploading a document does nothing until you press Train — per document or Train All — and each run shows its progress live, so compute is only ever spent deliberately.
Observations log
Every training run is recorded — which RICS sections were detected in the document and how confident the AI is in what it read — so you can see exactly what each report contributed.
A sample in your standard
Generate a fresh report section from a few generic notes and read it beside the verbatim passage the AI learned from — same structure, same measured length. Proof, not promises.
Adherence controls
Set how tightly the AI follows the trained standard — low, medium or high — choose whether a personal or the firm profile takes priority, and leave standing notes of up to 300 characters applied to every run.
Yours to remove
Download the original of any document, delete documents individually to free their slots, or clear the whole profile in one click. Clearing is immediate — and it's your call, always.
Why train at all
A trained draft vs an out-of-the-box one.
Untrained, the AI writes competent generic surveyor prose. Trained, it writes yours. The difference shows up in every element comment you'd otherwise reword by hand.
| What you'll notice | Trained on your firm | Out-of-the-box AI |
|---|---|---|
| Defect language | Your shape — observation, cause, action — in your firm's phrasing, guided by verbatim exemplars per element | Competent generic surveyor prose |
| Condition-rating caveats | A rating 2 or 3 carries the caveats your firm always attaches, learned from real examples of each rating | Standard boilerplate caveats |
| Recommendation wording | Escalates the way you do — monitor, budget for, obtain quotations before exchange | One-size-fits-all recommendations |
| The no-defect sentence | Your standard phrase, every time it applies | A different phrasing every report |
| Section length | Runs to your measured length, taken from your documents | The model's default length |
| Review experience | Drafts arrive closer to sign-off — less rewording of style, more attention on judgement | More time restyling in the editor |
Qualitative comparison of the same drafting engine with and without a trained profile — no accuracy figures implied.
What makes it different
Ten things a generic AI never learns about you.
Verbatim exemplars, not vibes
Training keeps real passages from your reports — how you actually wrote a rating 1, 2 and 3, your no-defect phrase, and a good and defect example per RICS section. Generated text is guided by the real thing, not a summary of it.
Level-aware profiles
Level 2 and Level 3 are trained and stored separately. Your concise HomeBuyer style and your fuller Building Survey style never blur into one another — and the log records the level the AI detected, so a mis-tagged upload is visible.
Two scopes, real privacy
Firm-wide documents teach everyone's drafts; personal documents teach one surveyor's and are private to their owner. Colleagues can't open or delete them.
Nothing trains until you press Train
Upload is inert. Extraction runs per document or via Train All, on your command only — so you always know exactly what the profile was built from.
Live extraction progress
Each run streams its progress with plain-language status messages while the engine works through the document, so training is something you watch, not a spinner you trust.
A log for every run
Sections detected, extraction confidence, and a failed-run reason when a document can't be read. The observations log is the training corpus's paper trail.
Proof by sample
One click writes a fresh section from fixed neutral notes and shows it beside the verbatim passage it learned from. If the sample doesn't read like your firm, retrain before it ever touches a client report.
An adherence dial
Low, medium or high — you choose how tightly generated text follows the trained standard, and whether a personal or the firm profile takes priority when both exist.
Standing notes
Up to 300 characters of house rules applied to every run — the instruction your training documents don't quite capture, written once instead of repeated in every review.
One-click clear
Delete any document to free its slot, download originals whenever you like, or clear the entire profile instantly. The corpus is yours — including the right to remove it.
The spec sheet
AI Training at a glance
Who trains
Three firms, three reasons.
Multi-surveyor practices
Five surveyors, one letterhead. Train the firm-wide profile on the principal's best work and every draft — whoever dictated it — starts from the same defect language and the same rating caveats. Consistency without the Friday review pile.
Sole traders
Your reports are your reputation. Train on the five you're proudest of and the AI drafts in that register from the first sentence — so the polish you're known for survives the busiest weeks of the year.
Firms growing the team
A new starter's drafts arrive already shaped like the house standard — the senior's red pen is in the profile, not just in their diary. Reviews become teaching moments about judgement, not rounds of restyling.
Getting started
Trained before your next inspection.
AI Training is included with every verified Home.co.uk surveyor listing. No separate subscription, no setup fee — the slowest step is choosing which reports you're proudest of.
Claim your firm
Most RICS-regulated firms are already listed — claim yours free, or register it in the same panel. The toolkit activates once your membership is verified.
Open AI Training
It lives in your surveyor account alongside your reports — one screen for documents, settings, observations and the sample generator.
Upload & tag
Add five to ten of your best reports, tag each Level 2 or Level 3, and pick a scope — firm-wide or personal. Nothing processes yet.
Train & check the sample
Press Train All, watch the runs complete, then generate a sample beside its source passage. Happy? Your next auto-mode draft starts from your standard.
Part of the toolkit
Training feeds the whole report suite.
The corpus you build here shapes what every drafting tool writes. These are the tools on the other end of the loop.
Otto — AI RICS Report Writer
The auto-mode engine your trained profile is injected into. Dictate the inspection, and the finished Level 2 or Level 3 draft arrives written to your standard in 5–15 minutes.
AI Dictation
Field-by-field dictation inside the editor — generated text follows your trained profile and adherence settings, the same as a full auto-mode run.
Firm Style Profile
The voice half of the equation — the 25-dimension writing fingerprint built from the same corpus you manage here. Substance says what; voice says how.
Standard Phrases Library
Pre-approved wording inserted verbatim where only the exact words will do — the complement to trained exemplars, which guide rather than dictate.
RICS Standards Check
Your standard covers style and judgement; the Standards Check covers compliance — reviewing any text block against current RICS guidance before you sign off.
Level 2 Report Writer
The editor where trained drafts land — in-place editing, AI rewrite and branded PDF export, with your Level 2 profile behind every generated sentence.
FAQ
Common questions about AI Training
Getting started
What does AI Training cost?
AI Training is included in the surveyor toolkit — there is no per-document or per-run charge. Costs stay controlled by design: uploads are capped at ten documents per firm, over-long documents are trimmed before processing, and nothing is processed until you press Train, so you decide exactly when compute is spent.
What should I upload?
Completed survey reports teach the AI most — upload your firm's best recent Level 2 and Level 3 reports as PDF, DOCX or DOC files, up to 25MB each. Tag each document with the report type it represents: Level 2 and Level 3 profiles are trained separately, so a Building Survey teaches your Level 3 standard without diluting your HomeBuyer style.
How many documents do I need?
A handful of strong reports is enough to build a working profile — five to ten is plenty. Each organisation has ten training slots, pooled across the firm-wide profile and any personal profiles your surveyors set up. You can delete a document to free its slot at any time, and what the AI has learned builds up as more documents are trained.
Do I need to prepare my reports before uploading?
No. Upload reports exactly as they were delivered to clients — PDF or Word. The extraction step reads the document as it stands, works out which RICS sections it contains, and over-long documents are trimmed automatically before processing. There is no template to fill in and no reformatting to do.
What it learns
How is AI Training different from Style Profile?
They are two lenses on the same training corpus, managed from the same screen. Style Profile is the voice — a 25-dimension fingerprint of how your firm sounds: structure, tone, vocabulary, patterns and formatting. AI Training is the substance — the judgement your reports encode: verbatim examples of how you write a rating 1, 2 and 3, your no-defect phrase, per-section defect passages and how your recommendations escalate. Voice makes a draft sound like you; substance makes it say what you would say.
What exactly does the AI extract when I press Train?
Each training run reads the document and records 25 writing dimensions across five categories (structure, tone, vocabulary, patterns, formatting), verbatim condition-rating examples for ratings 1, 2 and 3 plus your standard no-defect phrase, exemplar passages per RICS section, and measured section lengths. The run is logged with the number of sections detected and a confidence score, so you can see exactly what each report contributed.
Does the AI copy passages from my old reports into new ones?
No. The exemplars guide shape, phrasing and length — the content of a new report comes only from what you dictated, typed or photographed on that job. The AI does not invent observations and does not recycle findings from training documents into unrelated reports. You remain the surveyor of record on every output.
How do I know what the AI actually learned?
Two ways. The observations log records every training run — which RICS sections were detected in the document and how confident the extraction was. And the sample generator writes a fresh report section from a few generic notes, shown beside the verbatim passage it learned from — same structure, same measured length. Proof, not promises.
Control, scopes & levels
What is the difference between firm-wide and personal training?
Every document is uploaded to a scope. Firm-wide documents teach a shared profile that applies across the practice. Personal documents teach a profile for one surveyor and stay private to their owner — colleagues cannot open or delete them. A profile priority setting decides which wins when both exist: automatic, firm first, or personal first.
Are Level 2 and Level 3 reports trained separately?
Yes. Each document is tagged with the report type it represents, and Level 2 and Level 3 profiles are built separately — so a Building Survey teaches your Level 3 standard without diluting your HomeBuyer style. The extraction also records the level it detected in the document, so a mis-tagged upload is visible in the log.
Can I control how strongly the AI follows the trained standard?
Yes. A style adherence setting (low, medium or high) controls how tightly generated text follows the trained profile, a priority setting chooses between personal and firm profiles, and you can leave standing notes of up to 300 characters that are applied to every run — for example, a house rule the training documents don't fully capture.
Does training happen automatically as I write reports?
No. Nothing is processed until you press Train — per document or Train All — so your standard only changes when you choose. Separately, every auto-mode draft passes through a reviewer step that checks coverage, routing and fidelity against your dictation and corrects the draft before it saves; that check runs per report and does not alter your trained profile.
Data, security & removal
Is my data shared or used to train anyone else's AI?
No. Training documents are stored privately against your organisation, and only members of your firm can see them — personal-scope documents are visible only to the surveyor who uploaded them. The profile built from your documents belongs to your firm and shapes your firm's reports only. The AI providers used to read your documents do not train on customer content by default, and you can download or delete any document — or clear the whole profile — at any time.
What happens when I delete a training document?
The original file is removed from storage and its slot is freed immediately. Retrain your remaining documents to rebuild the profile without it, or clear the whole profile in one click if you want to start again. Deletion is your call, always — there is no retention period.
What if a training run fails or reads a document badly?
A failed run is marked with a plain-language reason so you can fix and re-upload — a scanned PDF with no readable text is the usual culprit. A weak read shows up in the observations log as a low confidence score or fewer sections detected than you expected. Delete the document, free the slot, and train a cleaner copy.
Which reports are written with the trained standard?
Every Otto auto-mode run and AI dictation session for your firm. The trained profile is injected at the drafting step, so defect language, rating caveats, recommendation phrasing and section length follow your standard from the first draft — whichever surveyor recorded the inspection.
More tools
Explore the rest of the surveyor toolkit
Otto
AI auto-mode RICS report writer
Dictate on-site, upload photos, get a finished, fully-routed Level 2 or Level 3 report in minutes.
Level 2 Report Writer
RICS Home Survey Level 2
Field-by-field RICS Level 2 reports with inline AI suggestions, image gallery, comparables and one-click branded PDF.
Level 3 Report Writer
RICS Building Survey Level 3
Comprehensive Level 3 Building Survey reports with deep technical sections, custom rating scales and AI legal-adviser notes.
Surveyor Comparable Tool
Sold + rental comps in one click
Auto-fetch sold and rental comparables within a 1-mile radius, weighted valuation range, branded PDF appendix for any survey.
Snagging Report Builder
New-build defect tracker
Build professional new-build snagging lists with location-based defect tracking, photos, urgency, status and PDF export.
AI Dictation
Voice-to-fields routing
Speak observations on-site; AI routes every sentence to the correct RICS field with confidence scores and replayable audio.
Image AI
Auto-categorise survey photos
Upload up to 200 photos; vision AI tags each to the right RICS element, rates condition, extracts defects and picks the best 35 for the report.
Style Profile
Your firm's writing fingerprint
Teach the AI to write like your firm — upload past reports, the AI extracts a 25-dimension writing fingerprint and applies it to every future auto-mode report.
Property Viewer
Pre-survey property intelligence
Look up any UK property and instantly see EPC, energy renovation costs, broadband coverage, environmental risks (flood, radon, mining, landfill), invasive plants, sales history and listing status — all in one dashboard.
Property Data Drawer
EPC + risks + comps in one panel
Open the property data drawer in any report to see EPC, environmental risks, listing history and comparables — copy-paste ready.
Standard Phrases
Reusable language library
Personal, firm-wide and industry-preset phrase libraries — insert with one click and use to train AI routing for voice consistency.
RICS Standards Check
Home Survey Standard compliance pass
One-click AI compliance pass against the RICS Home Survey Standard — flags missing content, weak ratings and structural gaps before you sign the report.
Home Payments
Paid faster, no invoicing
Home Payments mission control — customers pay at booking through a Stripe-verified checkout, payouts land on a schedule you can see, payment links cover your private work, and there is no invoice chasing ever again.
Bookings CRM
Every job, one pipeline
The surveyor CRM built into the platform — Homemove jobs land pre-paid with customer and property attached, enquiries convert to bookings, and the whole pipeline runs from one screen.
Calendars & Requests
A diary that fills itself
Per-surveyor diaries with live availability — keep your dates current and we book real jobs straight in, with inbound availability requests answered in a tap.
The firm inbox that knows the job
A full mail client scoped to your firm — @home.uk purpose mailboxes and per-user addresses, with every thread landing against the booking it belongs to.
Documents & e-sign
Signed before you set foot on site
Terms of engagement sent with the booking, tracked from sent to signed, chased automatically and stored against the job — no more chasing clients to sign.
Telephony
The firm number that never misses a job
A phone system built into the platform — take calls in the browser, forward to mobiles, ring the whole team at once, and keep a shared call log with optional recording.
Customer portal
Your client, never in the dark
A white-label portal under your firm's branding and domain — booking status, reminders and secure report delivery, so customers stop asking where things are.
Website builder
Built by asking, not designing
A firm website built and edited by talking to Inigo — brand and theme sync, a real section library, custom domains via CNAME, and enquiries that land in your pipeline.
Campaigns
Win the next job from the last one
Firm email campaigns with per-recipient tracking — sends, opens, clicks and read time — built on consented lists with one-click unsubscribe handled for you.
Reviews & badges
Proof that wins the next booking
Reviews earned on real completed jobs — scored by the homemover, shown on your directory profile — plus embeddable trust badges that carry the proof to your own website.
Performance
See the pipeline, beat the clock
Firm analytics without the spreadsheets — pipeline conversion, team performance and your own turnaround speed, the fourth lever of your rate.
Inigo
The AI assistant for surveyors
Ask in chat, dictate by voice, or click any element for contextual help — grounded answers on valuations, sales history and market data, inside the platform.
Environmental Report
Every risk, one report
A one-click environmental risk report for any property — flood, ground stability, environmental factors and planning constraints, delivered as a PDF.
Free directory listing. Pay-as-you-go toolkit.
Every RICS-regulated surveying firm gets a verified profile, direct bookings from home.co.uk consumers and the surveyor toolkit included with the account. No subscription. No pay-per-lead. No commission on the job. Pay only when you use the premium AI tools.
2 minute setup · Verified against the RICS register · Trusted UK property platform since 1995