Summary
Entrepreneur Liam Ottley spends four days at Thomas and Co., a fast-growing UK accountancy firm, attempting to automate its most time-consuming manual process: producing “working papers” that back up every number in a client’s accounts. The firm’s 25 accountants were each spending over 20 hours a week on this manual data entry, and Ottley sets out to build an AI system to fix it using no prior accounting experience.
The video documents the real build process: auditing all 17 types of working papers to find the best automation target, choosing VAT reconciliation as the starting point, and using a platform called Kilon to create AI agents that read the firm’s business context, pull data from spreadsheets into a shared database, and flag discrepancies for human review rather than replacing accountants outright. Firm owner Josh Thomas and technical lead Jonathan discuss the pressure of adopting AI in a regulated industry where a human must still legally sign off on the work.
By the end, the team has a working agent capable of running VAT reconciliation checks and surfacing issues for accountant review, with a roadmap to extend the same system to the other 16 working paper types. The video offers a candid, warts-and-all look at applying AI agents inside a real, regulated small business rather than a hypothetical demo.
📺 Source: Liam Ottley · Published September 21, 2026
🏷️ Format: Workflow Case Study
