Letting Agent Software With AI: A Checklist Before You Switch

Before switching letting software, ask each vendor five things: can you export all your data in a usable format, how are client-money records and approvals protected, what the AI features actually do and can be switched off, what the price will be for your exact property count, and who does the migration. Then make them show it with your own data.

Lettings software is sold on feature lists, and AI has become the newest line on every one: an assistant that drafts tenant replies, summarises inspection notes, reads documents. The risk is that a shiny assistant distracts from the boring questions that decide whether the move goes well. The checklist below puts the boring questions first and treats AI as something to test, not take on trust.

Follow me on Instagram@sagnikteaches

Decide whether you are switching or adding

A switch is a heavy project: tenancies, deposits, landlord statements, documents and every standing instruction move across. Adding AI to your current setup, with a separate tool or an integration, is usually far lighter. Write down the answer to one question before any demo: what is the problem the current system cannot solve? If it is "no AI", read how AI handles tenant queries and consider a tool alongside your current software first. If it is unreliable statements, poor support or a price rise, a switch may be right, and AI becomes a bonus you check carefully.

Connect on LinkedInSagnik Bhattacharya

Part one: getting your data out and in

Lock-in is the question vendors least like. Ask it of the new supplier and, just as importantly, work out what the old one will give you.

Subscribe on YouTube@sagnikteaches

1. What can I export, and in what formats? You want properties, landlords, tenants, tenancies, rent schedules, ledgers, deposits, certificates and their expiry dates, maintenance history, notes and documents. CSV or spreadsheet for records, original files for documents.
2. Is the export self-service or a paid request? Ask for the fee and the turnaround in writing. Some vendors charge per export, some limit how often you can request one.
3. Will you import my data, and who checks it? Find out whether the vendor does the migration, a partner does, or you do. Ask for a mapping document showing where each old field lands.
4. What is lost in migration? History of changes, notes attached to tenancies, communication logs and document versions are often dropped. Ask for the list.
5. What happens to my data if I leave you? Look for an export right on termination, a deletion timetable, and no extra charge to receive your own records.
6. Do you have an open interface (API)? If you want accounting, referencing or marketing tools to talk to the software, you need one. What is an API? Why it matters when you buy software explains what to look for.

Part two: client money, the question that is not about AI

Client money is the part of the system where an error is costly and a regulator may care. Rules on how you hold, reconcile and report client funds differ by market, so check yours and ask the vendor how the software supports them; do not assume a feature meets your rules because the brochure says "compliant".

1. Does the software keep client funds separate in its ledger from your own funds, and show balances per landlord and per tenancy?
2. Can I produce a reconciliation report on any date, and does it tie to my bank statement?
3. Who can approve a payment out to a landlord or contractor, and can I require two approvers above a limit?
4. Is every change to a ledger entry logged with user, time and old value?
5. How are deposits tracked, including registration or protection references and expiry or release dates?
6. Can the software reverse a mistaken entry without deleting history?
7. What is the process if a bank feed fails or duplicates a transaction?

Take a printed list of those seven questions to the demo and ask the presenter to show each one on screen. If an answer is "our accounts team can help with that", the feature is not in the product.

Part three: what the AI actually does

AI features vary from a thin wrapper that drafts an email to something genuinely wired into your records. These questions separate the two.

1. List every AI feature, by name. Ask for a plain list: drafts replies, summarises inspection notes, extracts details from documents, answers tenant calls, suggests maintenance triage, flags overdue rent.
2. Which model or provider sits behind it? You do not need the technical detail, but you need to know whether your data goes to a third-party model provider and where its terms are published.
3. Is my data used to train anything? Get the answer in the contract or the published privacy terms, not from the presenter.
4. What can it do on its own, and what only suggests? Anything that sends a message to a tenant, changes a ledger or ticks a compliance box should need a person's click. Ask for the list.
5. Can I switch each AI feature off, per user or per branch? You may want drafting on for negotiators and off for the accounts team.
6. Is AI activity logged? You should be able to see what was generated, who approved it and what was sent.
7. Does it read my documents correctly? Ask to upload three of your own: a tenancy agreement, a gas certificate and a messy scanned inventory. Do not accept the vendor's polished sample.
8. What does it cost? Some AI features are included, some are credits or add-on modules billed per use or per property. Get each in writing.
9. What happens when the AI is wrong? Ask for the stated process, any accuracy claim and whether the claim is backed by a document you can read.

For inspection and inventory features, AI inventory and check-out reports explains what accurate output looks like, and for document-checking claims, tenant referencing with AI sets out what a person must still sign off.

Part four: per-property pricing without surprises

Letting software is typically priced per property under management, per user, per module, or a mix, and the headline number rarely matches the bill. Most lettings platforms quote on request, so ask each vendor for a written quote built from your own profile. Two property-management platforms that do publish dollar list prices show why the profile matters. Neither is a recommendation; they are here only because their pricing pages are public. These are the figures their pages showed when this tutorial was checked in November 2026, so look again before you rely on them.

Now run a quick sum for a forty-property agency with about twenty-five new tenancies a year. On AppFolio Core, 40 units at $1.49 looks like $59.60 a month, but the 50-unit minimum makes it at least $74.50, and the minimum spend the page mentions could lift it further, so that minimum is the first number to get in writing. On Buildium Essential, the $65 starting fee is smaller than the payment fees if all forty tenants pay rent electronically and the agency absorbs the charge: 40 times $2.35 is $94 a month. Add one signed tenancy agreement for each new tenancy, 25 at $5, or about $10 a month, and the bill is roughly $169 a month before any unit-based increase. On Growth the same profile comes to about $202 plus $54 in payment fees plus $2 in signatures, roughly $258, but with lower fees per transaction. The cheapest headline is not the cheapest bill, and the deciding numbers are your own payment and tenancy volumes.

Also ask what the price does when you grow from 40 to 80 properties, whether there is a price lock, whether payment fees can be passed to tenants, and the notice period and any minimum term. On a per-unit model the growth is easy to see: 80 units on AppFolio Core at the published rate is $119.20 a month. Price rises matter just as much. A $169 monthly bill is $10,140 over five years if it never changes, and about $12,380 if it rises ten per cent a year; per-seat fees compared over five years shows how to lay out that calculation.

Part five: support, security and the contract

• Support: hours, channels, target response time, and whether a named person helps during migration.
• Uptime and backups: stated availability, backup frequency, and the process for restoring a single record.
• Access control: role-based permissions, two-step sign-in, and the ability to remove a leaver at once.
• Security evidence: a published security page or an independent certification you can read, not a verbal assurance.
• Data location and sub-processors: where data sits and who else handles it, especially the AI provider.
• Contract: term, notice, price-change clause, what happens on a change of ownership of the vendor, and data return on exit.
• Reference customers: two agencies of your size, whom you call yourself.

A demo script you can run with every vendor

Vendors rehearse their demos with tidy data. Send the same script to each one in advance, with a request to use your anonymised sample files, and score the same things.

1. Add a landlord, a property and a tenancy from scratch. Time it. (Target: under ten minutes for a first-timer.)
2. Take rent and pay the landlord after a management fee, then show the ledger and the reconciliation report.
3. Register a deposit and show its record through to release.
4. Log a maintenance request from a tenant's message, assign a contractor, approve a quote above your limit and show the audit trail.
5. Upload your messy gas certificate and show what the software extracts and what it gets wrong.
6. Let the AI draft a reply to this tenant message: "Heating has been off for two days and I have a baby. Can someone come today?" Check tone, speed and whether the draft promises anything it should not.
7. Show the export: run a full export while we watch, and open it in a spreadsheet.
8. Remove a user and show that their access ends and their history stays.

Score each step from zero to three, and add a column for "had to be shown a workaround". Anything that required a workaround in a demo will need one every week in real use.

What the heating message teaches you

Step six deserves a closer look because it exposes the quality of the AI. A poor draft says "We will send an engineer today" because it wants to sound helpful, though nothing in your records says an engineer is free. A good draft acknowledges the urgency, asks for the property address if the record is ambiguous, says a member of staff will confirm timing, and flags the message as a priority for a person. Ask the vendor whether the AI can see your contractor availability; if not, it must not promise. That small test shows the guardrails in about a minute.

A scoring sheet and the decision

Put the results in one table with weights, because the final choice should come from numbers that do not depend on the best presenter.

With those scores, Vendor X totals 2.15 out of 3 (0.75 + 0.5 + 0.4 + 0.1 + 0.1 + 0.3) and Vendor Y 2.35 (0.5 + 0.75 + 0.4 + 0.2 + 0.3 + 0.2). The sums are illustrative; the point is that AI carries a ten per cent weight here, which reflects its real importance to most agencies compared with money handling and getting your data out.

Audit your own data before anyone migrates it

The quality of a migration is capped by the data you start with, so spend an afternoon on a sample before the first demo. Export your current tenancy list and pick ten rows at random. For each one, check that the tenancy start date, rent amount, deposit reference, landlord and certificate expiry dates are present and consistent. Say you find that of ten sampled tenancies, three have no deposit reference, two have a rent figure that differs from the last rent review letter, and one certificate has no expiry date. Scale that up: on a forty-property book, roughly a third of records may need fixing, which is twelve to thirteen files at perhaps fifteen minutes each, about three hours. That is cheap to do before the move and expensive to discover in the new system during your first rent run.

Share the audit result with each vendor and ask how their import handles missing or inconsistent fields: rejected, imported blank, or flagged for review. A tool that refuses an incomplete row and tells you why is safer than one that quietly imports a blank.

Testing document extraction with your own files

Document-reading is the AI feature most likely to impress in a demo and disappoint on a Monday. When you upload your own messy gas certificate, look at the output field by field. A realistic, illustrative result might read: "Document type: gas safety record; inspection date: 14 March; next due: 14 March next year; engineer: [name]; appliances: boiler, hob; result: pass". Now compare to the paper. Perhaps the scan said the hob failed a flue check, and the extraction summarised "pass" because the boiler passed. Or the date read "14/03" and the model assumed the wrong year. Record every field that differs, count it, and ask the vendor whether the extraction shows a confidence level or a link back to the source line on the page. Without a way to see the source, a busy negotiator will trust the summary, and the one wrong field is the one that matters.

Calling a reference customer

Ask for two agencies of roughly your size, and call them yourself. A five-minute script gets more than a general chat:

• How long did the migration take compared with what you were promised, and what was lost?
• When did you last have a problem with your ledger or reconciliation, and how fast was it fixed?
• Which AI features do your staff use daily, and which did they stop using?
• Has the price changed since you signed, and how much notice did you get?
• Would you choose them again, and what would you check if you were buying today?

The answer to the third question is often the most valuable line in the whole procurement exercise, because it separates features the vendor shows from features customers keep.

Red flags that should end the conversation

• The vendor will not let you run an export during the trial.
• An AI feature sends messages to tenants without a person approving them and cannot be switched off.
• No audit trail for ledger changes.
• Pricing is "on request" until after you have signed a contract term.
• The privacy terms are silent on whether your data trains a model.
• The migration plan is "we will sort it once you are live".

A cut-over plan for the weeks after you choose

If the vendor passes, protect the move. Pick a start date away from your busiest rent run. Do a trial import of one landlord's portfolio first and compare every number to the old system. Run both systems side by side for one full rent cycle, with the old one as the record until balances tie. Keep the old contract alive until you have verified a full export and one month-end reconciliation in the new system, and tell landlords and tenants what is changing and when. Then decide which AI features go live first: usually drafting replies, with a person approving every message, before anything that touches money or compliance records. For the compliance side, the landlord compliance tracker tutorial shows the checks worth keeping in the new system.

If you would rather have someone test the shortlist against your real workflow, that is part of what my AI implementation consultation covers.

Questions agencies ask when comparing letting software

How long does switching letting software usually take?

Ask each vendor for a dated plan rather than a headline figure. Small portfolios with clean data can move in weeks, but messy tenancy records, client-money balances and landlord statements often stretch it to a few months. Budget for a period where both systems run side by side.

Is AI a reason to switch on its own?

Rarely. If your current system holds your data safely and reconciles your accounts, a new chatbot or drafting button seldom outweighs the cost and risk of migration. Switch for reliability, price or missing core features, and treat AI as a tiebreaker.

What should I do with my old data after migrating?

Keep a full export, archived read-only, for as long as your record-keeping rules require. Do not let the old contract end until you have checked balances, tenancy dates, deposit records and documents in the new system against that export.

Can AI features in letting software handle client money?

They should not move or allocate money without a person's approval. Ask the vendor which actions the AI can take, which it can only suggest, and whether every AI-assisted entry is recorded in the audit trail alongside the user who approved it.

How do I compare per-property pricing fairly?

Build a sum from your own numbers: units under management, tenancies started a year, number of users, add-on modules and messaging credits. Ask for the quote in writing for that exact profile, including what happens to the price when you add or lose properties.

Further reads

• Landlord Lead Generation for Letting Agents: AI Outreach, Not Spam — Winning landlords once the new system is live.
• Tenant Complaint Response Letters When ChatGPT Wrote the Complaint — Handling complaints that arrive written by AI.
• How to Evaluate an AI Implementation Proposal or Quote — Judging a vendor's or consultant's quote.
• Restaurant AI Tools: 12 Questions to Ask Before You Sign Up — A similar question list from another sector.
• Estate Agent AML Checks: Organise Due Diligence Files With AI — Keeping due-diligence files organised.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — Check your own readiness before buying.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: AppFolio and Buildium pricing pages (checked 4 November 2026) for the published pricing structures; the rest is general software-procurement practice.