How it works

Add the platform as a copy recipient on the email you already send to head office.

The weekly figures your care homes already send to head office, the certificate that came back from the engineer, the insurance schedule, the lease. Next time one of those goes to head office, copy the platform into the email as well. That is the whole of the work. No new software system at the care home, and nobody re-keys anything.

Three steps

  1. It arrives in the module in whatever form it was sent

    Weekly P&Ls, agency invoices, certificates, leases, loan agreements, bank feeds. Spreadsheet, PDF, scan or email, forwarded to one address or dropped in a folder.

  2. Your team approves what was read

    The AI reads the document and places each figure. A person on your team confirms it with one click before it counts anywhere. Nothing enters the record unapproved.

  3. Every figure keeps its route back

    File, page, who sent it, when, who approved it. A missing figure is shown as missing and never filled in. The reading is AI; the arithmetic after it is fixed.

The question everyone is asking this year

"Can we not just put an AI on top of what we have?"

You can, and you should, and we have. Point a good agent at your document store and it will read the leases, the policies and the spreadsheets, work out what they are and tell you what they say. That part works now, and it is cheap. It is worth being exact about the four things it does not do.

It reads now, and then it forgets

An agent layer holds no memory between one question and the next. Ask it again next quarter and it starts from the documents again. But almost everything that matters here is a comparison across time: this month against last, trading against the covenant, the sum insured against what it would cost to rebuild today, the care home that has been under its threshold for three months. A comparison needs something that remembers, and remembering is not a cleverness problem. It is a record.

Accurate is not the same as repeatable

You can teach an agent to run the same twenty calculations every week, and the arithmetic will be identical because it writes the sum once and runs it. The figures going in are the problem. Is this the rent review date or the break date. Is this sum insured the reinstatement figure or the declared value. Those are readings, and a reading can come out differently next time, after which the formula faithfully computes a different answer with exactly the same confidence. What a lender or an auditor asks is never how accurate your model is. It is where this number came from and whether it would come out the same again.

It answers what you thought to ask

A chat window is reactive by construction. The exposure that costs money is the one nobody thought to ask about, and a care home three months under threshold does not announce itself. Raising it unprompted needs a record watched against thresholds on a timetable, which is a system rather than a conversation.

And half of it is in somebody else's building

This is the one that does not go away. Your document store holds your half. The owner has the lease, the loan agreement and the valuation. The operator has the occupancy, the agency hours and the trading. No agent, however good, can read documents the company it belongs to does not hold. The question that matters most, whether a care home is underperforming because of the operator or because of the building, needs both halves at once.

So we are not arguing against any of this. Runko reads with AI, because reading is what it is good at, and that part of the work is now cheap. What sits underneath is a record that remembers, engines that give the same answer twice, and every figure keeping the page it was read from. An agent is only ever as useful as the thing it is allowed to act on.

Your data is messy. Everyone's is.

Weekly P&Ls arrive in different layouts. Rows get renamed, figures go missing, and PDFs are scanned. The platform's AI reads whatever arrives, whether a spreadsheet, PDF or email, and places each figure where it belongs.

A person on your team approves it with a click before it counts.

If a property's documents or spreadsheet layout change, the platform asks rather than guesses. Missing figures are shown as missing, never filled in, so you can see which properties send complete returns and which don't.

Every number traces back to its file, page, sender and date. AI reads the documents; the calculations are fixed arithmetic.

Documents received, Monday 14 September Illustrative
Portfolio Monday morning Asset register Deal pipeline Document assembly Finance Covenants Valuation Returns IFRS 16 EPRA / INREV Tax Operator Performance Benchmarks Admissions Staffing Documents Documents received

Click a document to see what was read

    Select a document to see what was read, what's approved and where it flows.

    Bring your own returns, in whatever layout they arrive.

    Try it

    Five sites or more?

    Start with one module, or one part of the platform, set up around your own reporting cycle rather than a demonstration. Yours, not ours.