AI automation in an accounting firm, with staff approving every output
We built an automation system for an accounting office: client requests are sorted and priced, P&L statements are drafted, and staff review and approve every output before it goes out. The system now runs under the firm's own engineer.
Start with the work that repeats
Good candidates for automation arrive the same way every week. Sort those requests by type, price them by the firm's own rules, and draft the first version of the work. Unusual cases can stay with people.
What the system does
- Sorts incoming client requests by type
- Prices them by the firm's rules
- Drafts P&L and recurring statements for review
- Puts every output in a queue where staff approve or correct it
Why a person approves everything
In accounting, a wrong number with the firm's name on it costs more than the time saved. So the rule here has no exceptions: nothing leaves the office until a person has approved it.
The review queue is also where the system meets the staff. They stay responsible for the work and spend their time checking it rather than preparing it.
Hand-over was part of the plan
The firm did not want to depend on us forever, and it should not have to. We documented the system and handed it to the firm's own engineer, who runs and maintains it now. Whoever builds your system, ask how that hand-over would work before you start.
Questions
- How can an accounting firm automate client intake?
- Start with requests that arrive the same way every week: sort them, price them by your rules, and draft the first version of the work, with a person approving before anything leaves the office.
- Can our own engineer run the system?
- Yes. This firm's engineer runs and maintains it today.