AI systems for property, lettings and estates teams
Property management is a document business wearing a people business's clothes. Certificates expire, tenancies renew, evidence has to be served and proven, and every one of those things has its own date and its own penalty. Managing that across three hundred units is not a job anyone should be doing from a spreadsheet.
The admin that carries a penalty
In most sectors, forgetting a renewal is embarrassing. In property it's a fine, an unenforceable notice, or a tenant living somewhere unsafe.
Compliance certificates across a managed portfolio
Gas safety annually. Electrical inspection every five years. EPC validity. Alarm compliance. Legionella assessments. Licences where they apply. Each with its own cycle, its own document, its own penalty — and the operational burden sits with the agent even when the legal duty sits with the landlord.
The highest-value workflow to automate in property managementMaterial information nobody can find
Tenure, council tax band, construction type, services, parking, flood risk, covenants, rights of way. It lives in a solicitor's file, a survey, an EPC or the seller's head. The listing waits while somebody chases it.
Maintenance triage, twenty times a day
“The boiler's making a noise.” Decide urgency, decide whose responsibility, check the contractor rota, get authorisation if it's over the landlord's limit, book it, chase it, close it, log it. Five to fifteen minutes of coordination each, all day.
Sales progression held together by one person's memory
Ringing solicitors, brokers, surveyors and other agents; getting voicemail; leaving notes; repeating tomorrow. There's no shared system of record across a chain, so the whole process depends on a progressor and a spreadsheet.
What we'd build for a property business
These are document-and-deadline problems, which is exactly the shape of work AI handles well — provided the judgement calls stay with your people.
Compliance expiry and chase worker
Reads certificates, extracts issue and expiry dates, monitors them against your thresholds, raises the booking request to the contractor, chases the landlord for authorisation, files the returned certificate and serves it with proof of service recorded.
Document in, document out, hard dates. The strongest fit in the sector.Listing preparation and gap list
Reads the EPC, title register, lease, survey and seller questionnaire, populates what it can evidence, and produces a list of every field it could not. The gap list is the useful part.
Every field carries a source reference.Maintenance triage
Classifies inbound reports by urgency and responsibility, checks them against the tenancy agreement and any statutory timescales, drafts the tenant acknowledgement and raises the work order — escalating hazards to a person immediately.
AML file assembly
Collates identity documents, screening results and source-of-funds evidence into a structured file with a checklist of what's still missing, and chases it. The screening stays deterministic and the decision stays with your nominated officer.
Assembles and chases. Never decides.Sales progression agent
Maintains a live state model of each chain, generates the daily chase list, drafts the chase emails, logs the replies and produces a plain-English weekly update for the vendor — which removes most of the “any news?” calls.
Leaseholder query answering
Answers “why has my service charge gone up” from the lease, the budget and the actual invoice ledger, with citations to the documents — instead of a property manager digging through invoices for an hour.
A regulatory picture that is genuinely moving
Property is mid-reform on several fronts at once. Some widely-circulated guidance has been withdrawn, and commencement dates have moved more than once. We track it, because you shouldn't have to.
- <strong>Consumer protection law was replaced, not amended.</strong> The regime governing misleading actions and omissions in property marketing changed on 6 April 2025, when Parts 3 and 4 of the Digital Markets, Competition and Consumers Act 2024 replaced the Consumer Protection from Unfair Trading Regulations 2008. A lot of “AI for estate agents” material still cites the superseded rules.
- <strong>Material information guidance is in transition.</strong> The previous published guidance has been withdrawn pending replacement. The underlying duty not to mislead by omission is unchanged and enforceable right now — which is why our builds evidence every field rather than infer it.
- <strong>Anti-money-laundering duties apply to both sides.</strong> Once an offer is accepted, due diligence is required on buyer and seller, not just your instructing client. The commonest enforcement failing is administrative, not sophisticated.
- <strong>Lettings reform is phased and dates have shifted.</strong> We build systems that can absorb a changed tenancy regime rather than hard-coding today's rules into a workflow that breaks on commencement day.
- <strong>Surveying work carries its own AI rules.</strong> Professional standards now require recorded judgement on the reliability of AI outputs and disclosure of AI use in terms of engagement. Any build for a surveying practice accounts for that.
What we won't automate in property
Property decisions affect where people live and who gets to live there. Several of these are refusals we'd hold to even if asked.
Tenant selection, referencing or affordability decisions
This is an automated decision with significant effects on a person, and it carries discrimination risk under equality law. We'll structure and present the information. A human decides who gets the tenancy, and the applicant can contest it.
Generating material information that isn't evidenced
A model inferring “freehold” or “no flood risk” because a property looks typical has created a misleading statement with legal consequences for you. Every field in anything we build either carries a source document or is marked “not yet established”.
Clearing an AML risk, sanctions or PEP hit
Screening runs deterministically against official lists. Hits are adjudicated by your nominated officer. The system gathers, chases and assembles — it does not judge.
Closing out a hazard report, or serving notice
Damp, mould, electrical faults, a smell of gas, structural concern — instant human escalation, every time. Notice validity is technical, dated and litigated; get it wrong and possession fails.
Producing a valuation or a condition rating
A model output can be an input to a valuation. It can never be the valuation. Professional judgement is the product and it belongs to the surveyor.
Common questions.
Can AI do AML checks for estate agents?
+
It can do the administrative majority of the work, but not the decision. A system can collect identity documents, chase what's missing, run screening against official lists, assemble the evidence into a structured file and flag gaps before a deal stalls. What it must not do is judge the risk rating, clear a sanctions or politically-exposed-person hit, or sign off source of funds — those are attributed to your nominated officer and can't be delegated to software. If you're looking for AI to make AML go away, it won't. If you're looking to stop losing hours to chasing, it will.
How can letting agents keep on top of compliance certificates?
+
By treating it as a data problem rather than a diary problem. A system reads each certificate as it arrives, extracts the issue and expiry dates, and monitors the whole portfolio against your notice thresholds — then acts: raising the contractor booking, chasing the landlord for authorisation, filing the returned certificate and recording proof of service on the tenant. The failure mode in most agencies isn't that nobody knows the dates; it's that knowing them and acting on them are two different jobs and only one of them is anybody's.
Can AI write property listings legally?
+
Only if it is structurally unable to invent an attribute. Estate agency is precisely where enforcement lives, and the rules on misleading actions and omissions apply to every descriptive claim you publish. We build listing tools that can only assemble copy from evidenced fields — so the system can say a property is freehold when the title register says so, and cannot say it when nobody has checked. Descriptive language still gets a human read before it goes out.
Does this work with our CRM or property management platform?
+
Yes — we build around whichever platform you already run rather than replacing it. Most property builds read from email and shared document folders on the way in, and write into your CRM or management platform on the way out, so negotiators and property managers carry on working where they already work.
How long does a property AI build take?
+
Typically 4–6 weeks from signed scope to live for a single workflow. Compliance tracking is usually the quickest to prove because the inputs are structured documents with dates on them. Anything touching AML or tenancy decisions takes longer, because the human checkpoints need designing properly rather than bolting on.
Start with the certificates.
It's the workflow with the hardest dates and the clearest penalties, which makes it the easiest to prove value on. Tell us how many units you manage.