What Does It Actually Cost to Automate Admin for a Small Lettings Agency?
For a small UK lettings agency, a single well-scoped automation — one workflow, two or three connected systems — typically lands in the low thousands, and a full operational build across tenancy, maintenance and finance runs into the tens of thousands. The more useful question is payback period, and for most agencies doing 30 or more tenancies a month, a first build pays for itself inside twelve to eighteen months. This article breaks down what drives that number in both directions.
What are the actual price ranges?
Three shapes of engagement cover almost everything a 5–40 person agency needs.
- A diagnostic audit — every manual handoff mapped, timed and priced, with a ranked build sequence — sits at £3,000–£4,000. It is deliberately standalone: you can take the output to any developer.
- A custom build of a real operational system, with multiple connected tools, exception handling and monitoring, runs £18,000–£30,000 depending on scope.
- Ongoing support and iteration once something is live is typically £750–£2,500 a month, driven by how much changes and how critical the workflow is.
Those are consulting-grade numbers for systems that run unattended and are expected to keep running. A single connector between two tools that tolerates occasional failure is a much smaller piece of work, and you should not be quoted a build price for one.
What actually drives the price up or down?
Not the number of systems, and not the number of steps. Three things.
1. Whether your systems have usable APIs
Reapit, Alto and the mainstream accounting packages expose data properly, and that keeps a build cheap. Older or portal-only tools mean working through scheduled exports, email parsing or browser automation — all viable, all more expensive to build and more expensive to keep running when the vendor changes a page layout. If half your stack is portal-only, expect the higher end of any range.
2. How much your data disagrees with itself
The same landlord spelled three ways across two systems is not a cosmetic problem. Before anything can be automated, something has to decide which record is authoritative and what happens when two nearly match. On a messy dataset this reconciliation work can be a third of the project. Agencies that have never merged duplicates are always surprised by this.
3. What happens when it goes wrong
Building the path where everything works is fast. Designing what happens when the accounting API times out mid-write, when a tenancy record arrives with a missing deposit figure, or when the same referral comes in twice — that is most of the real engineering. Retries, idempotency, a review queue for ambiguous records, and an alert when something silently stops. It is the unglamorous part, and it is the difference between a system you trust at 2am and one somebody has to check every morning.
If a quote is notably cheap, this is usually what has been left out. Ask directly what happens on failure.
What is the cost of not doing it?
This is the number most agencies never calculate, and it is usually larger than the build.
Take one task: rekeying a new tenancy across the portal, the CRM and the accounting system. Nine minutes, forty tenancies a month, done by someone on £32,000:
- 9 min × 40/month × 12 = 72 hours/year
- 72 hours × £22/hr fully loaded = £1,584/year
Fully loaded means salary plus employer National Insurance, pension, holiday cover and overhead — roughly 1.25 to 1.4× gross. A £32k administrator costs you about £22 an hour, not £16.
And that is one task. Most agencies have between four and nine. Add the error correction: manual transcription runs around a 1% error rate, so 480 tenancies a year produces about five mistakes, and a single deposit logged against the wrong property can take two days to trace. Add the delay: enquiries arriving after 4pm get handled the next morning, and in a market where the first credible response usually wins the instruction, that is a revenue cost rather than an admin one.
The thing that makes waiting expensive is the asymmetry. The build costs the same in January as it does in November. The manual work does not — it accrues every month, and it grows with your volume.
What does a realistic payback period look like?
Work it out as annual recovered cost divided by one-off build cost.
An agency recovering 72 hours a year on one task plus one avoided two-day incident is recovering roughly £2,300 a year. Against a small connector build that is under a year. Against a £20,000 operational rebuild it is not, which tells you correctly that one task does not justify that project. Three or four tasks, plus the delay cost on new enquiries, usually does.
The rule I use: under twelve months payback is an obvious yes. Twelve to twenty-four months is a yes if the process will still exist in five years, which lettings admin will. Beyond that, fix the process manually first and revisit.
When is it too early to automate?
Two situations, and I have turned down work in both.
The first is when the process itself is wrong. If your team retypes data because the CRM was configured to hold fields it should never have held, automating the retyping locks a bad design into code and makes it harder to change later. Fix the upstream process, then automate what remains. That conversation usually produces a smaller invoice, not a larger one.
The second is low frequency. A task that happens four times a month and takes six minutes costs under £110 a year. That is a checklist, not a project. Automation earns its keep on frequency, not on how irritating the task feels.
Frequently asked questions
Can I just use Zapier or Make instead?
For two systems, modest volume and a tolerant failure mode, often yes, and you should. They become expensive and fragile once you need conditional routing, retry logic, or an audit trail of what happened and why. The switch point is usually a few thousand operations a month, or the first time a compliance requirement lands.
Do we need to replace our CRM?
Usually not, and I would be wary of anyone whose first answer is that you do. Replacing a system of record is a far larger and riskier project than connecting the one you already have.
How long does a build take?
A single clean integration is days of build once scoping is done. A multi-system operational flow with proper exception handling runs to weeks. Scoping quality determines both.
What happens if you are unavailable?
Everything is built on open, self-hostable tooling and documented so another engineer can pick it up. You own the system and the infrastructure it runs on. That is a deliberate design decision, not an afterthought.
Where to start
Before speaking to anyone, including me: list every task where someone reads data out of one system and types it into another, time the top three honestly with a stopwatch, and multiply by real monthly volume from your system of record rather than from memory. If the total comes in under £1,000 a year you have just saved yourself a project. Over £5,000 and you have a business case that will survive a conversation with your accountant.
A Tier 1 Audit does that work properly across your whole operation and returns a ranked build sequence with expected payback for each item, whether or not I build any of it.