Examples

Things worth building.

Each of these is a real, buildable system rather than a concept. For every one, the honest question is the same: what does a rule do, what does AI do, and what stays with a person?

These are examples, not case studies.

They describe work I can build and how I would approach it. None of them describes a completed client project, and there are no client results, percentages or testimonials on this site. When there are real case studies they will be labelled as such, with the client's permission and their numbers. Thelive demo is the first of these, built and running.

Inbound enquiry triage

Low risk

The problem

A shared inbox mixing viewing requests, repair reports, landlord questions, contractor updates and cold sales email. Someone reads all of it and decides what happens next.

The approach

Classify each message, pull out the contact details and requested times, match it to the property record, set a priority and prepare a reply for approval.

AI does
Optional rephrasing of the prepared reply into house tone.
A person does
Approves or edits every reply before it sends. Confirms anything urgent.

Compliance deadline watch

Low risk

The problem

Gas safety records, EICRs and EPCs expire across a portfolio. Tracking sits in a spreadsheet that only gets checked when someone remembers.

The approach

A daily check across the portfolio that produces one list: what has expired, what expires in 30 days, and what to book this week.

AI does
None. This is a date comparison and should never be probabilistic.
A person does
Books the renewals. Owns the exemption decisions.

Property details from a landlord pack

Low risk

The problem

A new instruction arrives as a folder of PDFs. Someone retypes the details into the CRM and the portal listing.

The approach

Extract the fields, validate them against format rules, and present them side by side with the source document for a one-screen check.

AI does
Reading unstructured documents where layout varies.
A person does
Confirms each extracted field against the source before it saves.

Viewing feedback chase

Medium risk

The problem

Feedback after viewings is chased by hand, inconsistently, and landlords ring to ask why they have not heard anything.

The approach

Scheduled requests after each viewing, replies collected into a per-property summary, landlord update drafted weekly.

AI does
Summarising free-text feedback into themes.
A person does
Reads the summary and sends the landlord update.

Arrears position, every morning

Low risk

The problem

Working out who is behind and by how much means exporting from the accounts system and rebuilding the same spreadsheet.

The approach

A scheduled report producing the arrears position, movements since yesterday and a suggested contact list.

AI does
None. Financial positions must be exact and reproducible.
A person does
Decides who to contact and what to say.

Maintenance job pack for contractors

Medium risk

The problem

Every job means retyping the address, access notes, tenant contact and issue description into an email.

The approach

Assemble the job pack from the property record and the original report, ready to send with one check.

AI does
Turning a rambling tenant message into a clear fault description.
A person does
Approves the instruction before it goes to a contractor.

Meeting and viewing notes into actions

Low risk

The problem

Notes stay in a notebook. Actions get remembered or they do not.

The approach

Turn a recording or rough notes into a structured summary with owners and dates, dropped into the property record.

AI does
Summarisation and action extraction.
A person does
Confirms the actions are right before they are assigned.

Internal answer assistant

Low risk

The problem

New staff ask the same twenty questions about process, and senior people answer them again.

The approach

A search assistant over your own written procedures that answers with a citation to the source document.

AI does
Retrieval and phrasing, always with the source shown.
A person does
Owns the underlying documents. Checks anything consequential.

Beyond lettings

The starting focus is letting and property management agencies, because the same handful of processes repeat at every one and I can build them properly rather than approximately.

The underlying work is not property-specific though. Sorting an inbox, pulling fields out of documents, tracking dates that carry a penalty, and turning a system export into a report someone actually reads — those exist in recruitment, in wholesale, in professional services and in any business where a person is the integration between two systems. If you are not a letting agent but recognise the description,the call is still free.

Which of these is your Monday morning?

Tell me which one you recognised and I will tell you, honestly, whether it is worth automating at your volume.