AI for surveying firms: organising incoming case information

A new instruction rarely arrives in one piece. The client's email carries two attachments, a forwarded thread from their solicitor adds a third, the access arrangements sit in a phone note in someone else's inbox, and a corrected address turns up two days later as a reply to the wrong message. Before a surveyor can do any professional work, somebody has to turn that into a case file. In many smaller surveying firms, that somebody does it from memory.

Case intake is the stage between an instruction arriving and a surveyor picking up a complete, consistent file. It is a sensible first place for a surveying firm to use AI, because it has a clear start and a clear end and because most of the effort goes on sorting, checking and chasing rather than professional judgement. It only works, though, once the firm has decided what a complete file looks like and who resolves the information that does not agree. This article covers how AI can organise incoming case documents, how missing and conflicting information should be flagged, where the RICS standard on AI fits, and what to measure before automating anything.

Why intake is where surveying firms lose time before the work starts

Intake absorbs time because the information arrives across several channels and the firm has no single written definition of when a case is ready.

The cause is organisational rather than technical. Instructions come in through a shared inbox, individual inboxes, a website form, introducer portals and the phone. Attachments are saved to a shared drive under whatever name the person opening them chooses. Whether a file is complete depends on who assembled it, because the checklist lives in experienced administrators' heads. Chasing happens when someone remembers to chase. Discrepancies, such as an address that differs between the instruction and the title plan, tend to surface at inspection or during report drafting, when they are most expensive to fix.

Hypothetical example, for illustration only.

A firm receives an instruction for a residential building survey by email. The client names the property as "Flat 3"; the attached sales particulars describe "Apartment 3B"; the solicitor's forwarded email gives a different postcode. The administrator files all three, books the inspection against the client's version and moves on. The surveyor finds the mismatch on site. Nobody made a careless decision; the firm simply had no step where the three sources were compared.

Intake ends when a consistent case file is handed to the surveyor. Preparing the report from that file is a separate stage with different risks and different controls, and it deserves separate treatment.

What a consistent case file needs before AI is involved

AI can only flag something as missing if the firm has written down what "complete" means for each type of instruction.

That definition is the most useful piece of work a firm can do before any system is built, and it costs nothing but time from the people who already run intake. For each instruction type the firm handles, it should state the fields that must be present, the documents that must be on file, which source counts as authoritative for each field, and who decides when sources disagree. A building survey, a valuation for secondary lending and a dilapidations instruction will have different lists.

Field or documentWhere it usually comes fromWhat counts as confirmedClient name and contact detailsInstruction email or formMatches the signed terms of engagementProperty addressInstruction, particulars, title documentsSame address across sources, or a person has resolved the differenceInstruction type and scopeInstruction email, quoteMatches the agreed terms of engagementOther parties (seller, agent, solicitor, lender)Instruction, forwarded correspondenceNames recorded and passed to the firm's conflict checkAccess arrangementsEmail, phone noteContact and date recordedSupporting documentsAttachments, client uploadsPresent on file, or requested with a date

The table is a starting template. Each firm's version should come from its own instructions, its own terms of engagement and the questions its surveyors ask most often when a file reaches them incomplete.

How can AI organise incoming surveying case documents?

AI can read each incoming email and attachment, work out what it is, extract the fields the firm has defined, file it against the right case and keep a running checklist of what has and has not arrived.

In practice that is a short sequence. The system watches the agreed entry points, usually a shared inbox and an upload folder. It classifies each item: a new instruction, a document for an existing case, a reply to a request, or something unrelated. It extracts the defined fields and records where each value came from, down to the email or page. It matches the item to an existing case or opens a new one, and it names and files documents consistently. Finally, it compares the case against the checklist for that instruction type and drafts a request for anything outstanding. A member of staff approves the request before it is sent.

The boundaries matter as much as the steps. The system should not accept or decline an instruction, agree scope or fees, decide whether a conflict of interest exists, or give the client any professional view on the property. Those remain decisions for the firm's people. The value of the AI sits in the sorting, matching and checking that currently fill administrators' days, and in making the state of every case visible rather than held in someone's memory.

How should missing or conflicting information be flagged?

Missing information should go on the case checklist with a drafted request; conflicting information should be recorded with both values and their sources and sent to a named person to resolve. Nothing should be overwritten silently.

Every field in the case file then sits in one of four states: confirmed, missing, conflicting, or extracted with low confidence. Confirmed fields need no attention. Missing fields generate a request, and the checklist records when it was sent and when to follow up. Low-confidence extractions, such as a poorly scanned page or a handwritten note, are shown to a person with the source attached. Conflicting fields are the ones that need the most care, because the tempting shortcut is for the system to pick the most recent or most frequent value. That shortcut hides exactly the discrepancy the firm needs to see.

Hypothetical example, for illustration only. The instruction gives the postcode as one value and the attached title plan shows another. The case file records both, links each to its source document and routes the case to the person responsible for intake that day. That person checks with the client, records the answer and the reason, and the field moves to confirmed. The surveyor later sees a confirmed address and, if they want it, the history of how it was resolved.

The same principle applies to the parties involved. Intake is the natural point to collect the names of the seller, agents, solicitors and lenders connected to a case. RICS's professional standard on conflicts of interest is mandatory for RICS professionals and regulated firms (RICS, Conflicts of interest). An intake system can surface the names for the firm's conflict check, but the judgement on whether a conflict exists, and how to manage it, stays with the firm.

Where the RICS standard on AI fits

RICS's professional standard Responsible use of artificial intelligence in surveying practice has been mandatory for all RICS members and regulated firms since 9 March 2026 (RICS, March 2026), and it asks firms to judge whether an AI system has a material impact on the delivery of their services.

The standard covers four areas: governance and risk management, professional judgement and oversight, transparency and client communication, and the responsible development of AI (RICS, March 2026). Its requirements apply to AI with a material impact on the delivery of surveying services, and RICS leaves the materiality question to members' informed professional judgement (RICS, Responsible use of AI). Where AI is material, the obligations include telling clients in writing and in advance when and for what purpose AI will be used, keeping a written register of AI systems with their purpose, date of first use and next review date, and having written decisions on the reliability of outputs prepared by or under the supervision of a named, appropriately qualified surveyor (CMS, 10 March 2026).

An intake system that only files and requests documents may well be judged non-material. One that extracts the address or scope the surveyor relies on is closer to the line. Either way, the firm should make the judgement deliberately, record its reasoning, and take advice from its own compliance lead or RICS guidance where it is unsure. The four-state approach above also helps here, because it leaves a record of what the system extracted, where from, and which person confirmed it.

What should be measured before automating document intake?

Before automating intake, a firm should measure how the current workflow performs on a small number of figures it can count today, and agree the result it wants before anything is built.

Useful measures for intake are practical rather than sophisticated:

• time from an instruction arriving to a complete case file being handed to the surveyor
• number of chasing rounds per case before the file is complete
• number of files returned by surveyors because something was missing
• number of discrepancies found at inspection or later that were visible in the documents at intake
• staff time spent on intake each week, sampled rather than estimated

Two to four weeks of simple logging against the firm's own definition of a complete file is usually enough to produce a baseline. That baseline does two jobs. It shows whether intake is the right workflow to start with at all, and it gives the firm a fixed point to judge the system against after it goes live. If the logging shows that most delay comes from clients returning signed terms slowly, a firm may be better served by fixing that step first.

This is also the point where Carbon's way of working applies. Carbon is an AI transformation firm for Europe's real economy. We help mid-market businesses put AI to work. Every engagement starts from an objective baseline and ends with a measured change in a metric the business already tracks, so the question of whether intake improved is answered with the firm's own numbers.

Starting with one workflow module

Carbon's view is that a surveying firm should start with one recurring workflow, defined against a baseline, automated and measured, before anything firm-wide. Intake suits that approach because its cost is easy to see and its rules belong to the firm: what a complete case file contains, which conflicts need a surveyor's judgement and when a case is ready to book. Those definitions are the firm's own knowledge, and they should stay the firm's whatever system runs them. If case intake is a recurring cost in your firm, the useful next step is a conversation about that one workflow and the result you would want to measure.

Carbon is an AI transformation firm for Europe’s real economy. We transform operations for mid-market businesses and automate recurring workflows for regulated services firms, measured against a baseline the business already tracks.

Own the Build™