AutoSettle: How an AI Agent Now Chases Debt Better Than Any Credit Controller

Of the seven categories Autothink is targeting, credit control is the most embarrassingly obvious. Every other category requires you to argue against an entrenched incumbent and an entrenched buyer preference. Credit control requires you to argue against a workflow that no one in the building actually wants to do.

Ask any UK SME finance lead what their least-favourite weekly task is. The answer is the chase. Ask them what the highest-anxiety cashflow item is. The answer is the chase. Ask them how much they are paying a credit controller, internally or externally, to do the chase. The answer is more than they would like. Now ask whether the chase is genuinely creative judgement work, or whether it is rules-based escalation dressed up as judgement. The answer, when they're honest, is the second.

1. The £20bn that nobody is defending

The Federation of Small Businesses' 2025 Late Payment Tracker put outstanding overdue invoices to UK SMEs at £32.4bn. Roughly £20bn of that is the structural "always-late" core — invoices that roll month-on-month, businesses that habitually pay 30, 60, 90 days late, a slow-moving liquidity drag that costs the UK SME economy more than energy, more than insurance, more than commercial rent.

No one is defending it well. Xero sends the invoice. Chaser, Satago, Float and the rest of the credit-control SaaS layer offer better templates, better sequences, better dashboards — but they all leave the human in the chair. The economic question — why is the chair still occupied at all? — has not been answered until now, because the technology to answer it did not exist.

2. Why credit control is the textbook agent target

Five tests determine whether a workflow is agentic-ready. Credit control passes all five:

• Rules-based escalation. The chase pattern is substantially deterministic — chase day 7, escalate day 14, call day 21, formal dispute day 30. Every credit-control playbook in Britain is a variation on this.
• Repetitive written comms. The agent writes thousands of slightly different chase emails in the brand voice. This is the cheapest, most-improved capability in modern AI.
• Complete written record. Every interaction is logged in the inbox or the ledger. The agent has perfect memory.
• Hard outcome. Cash either landed or it did not. The agent is judged on £, not on emails sent.
• Bounded judgement. The agent does not need to invent strategy — the strategy is "collect the cash within terms". The judgement is tactical.

3. What AutoSettle actually does

AutoSettle is connected to Xero, QuickBooks or Sage as the system of record. It pulls the receivables ledger, applies the configured sensitivity tier per client, and runs the chase. The chase has four phases — soft chase, firm chase, escalation, formal dispute — each with its own tone register and decision tree.

The agent writes every email in the brand voice (trained on the client's existing finance correspondence). It books calls into the finance lead's calendar when the escalation tier requires human voice. It logs every interaction back to the ledger. It updates the cash forecast as it works. The finance lead's week becomes: Monday morning, review the exception screen — the seven invoices that escalated, the three disputes the agent flagged, the one VIP account it deliberately held back. Twenty minutes of work where there used to be six hours.

4. The numbers from the first cohort

Across the first 22 UK SMEs deployed, AutoSettle reduced credit- control operator hours by an average of 14 per month per £100k of monthly receivables. Days Sales Outstanding (DSO) fell by an average of 9 days within 90 days of deployment. Payback against subscription cost: under six weeks.

The interesting number is not the time saved. It is the consistency. Human credit controllers chase brilliantly when they have time and patchily when they do not. The agent chases identically every day, on schedule, in the right tone, without decision fatigue. Most of the cashflow improvement comes not from the agent being smarter than a controller, but from the agent never being too busy to send the chase.

5. The category position

The category is "agentic credit control". Until 2026 it did not exist as a label. Owning the label is cheaper now than it will ever be again. By 2028 there will be three credible UK names in the space and a dozen also-rans. AutoSettle is built to be the first of the three and the longest-distributed, because every Creative Agent client with receivables is a natural deployment site.

This is the post-SaaS thesis applied to one of its most defensible workflows. Software does not "help the controller". Software is the controller. The buyer is the CFO. The line item is cash collected. The seat licence dies on contact.

6. What you should do, depending on who you are

If you run a UK SME — measure your DSO this month, measure your credit-controller cost (internal or external), and compare to a 6-week payback on agentic credit control. If the maths works, the conversation worth having is whether the function exists in 2027.

If you are a finance leader — the strategic work isn't going anywhere. The chase is. The role of "credit controller" shrinks; the role of "exceptions and disputes lead" grows. Hire and train accordingly.

If you are a founder — credit control is the lowest-resistance agentic build in the country right now. Pick a different category. The market is going to be loud.

To apply this to your own business, book a call. To go deeper on the operating philosophy behind AutoSettle and the rest of the portfolio, read Systems, Not Features.