Our approach

Design first.
Then rebuild.

We start from what finance must do for any business, not from the department you have today. Then we check the design against the real team and help you put it into practice, one release at a time.

1

Design first, then check.

Start with a clean sheet. Then use today’s work to make sure nothing is missing.

2

Train people before agents.

People learn every week from the first day, and knowledge is written down before any agent depends on it.

3

Decide, don’t just report.

Finance creates value through better, earlier decisions. Time saved only counts when it goes into work the business values.

The book’s blueprint

From a clean sheet
to a working function.

Seven steps build the design, from a clean sheet to a first proven release.

01

Everything finance does

Write down everything finance does.

List every activity finance must perform for any company, not just what your team does today. This becomes the checklist for everything that follows.

02

Where judgment lives

Decide where people must judge.

Go activity by activity. Let agents do the work they do well, and place a person wherever a mistake would be costly, hard to undo, or a matter of trust.

03

The agent roster

Hire the agents.

Give every agent a job description, a set of instructions, and one person who owns it. A few responsibilities can never sit with the same person, for control reasons.

04

The Workbench

Connect the agents to your systems.

Run the agents on a platform IT can secure, connected to the systems you already have, with controls around every agent and a knowledge base they can read.

05

The command center

Give every role the right screen.

Show each person only what changes what they do, from the payables team to the CEO.

06

Amnesty

Find out what the team really does.

Ask everyone what AI they already use, with a promise that nobody gets in trouble. Check every piece of today’s work against the design: it is already there, it should stop, or the design missed it.

07

Prove it

Prove it, then carry it.

Release the first area, purchase-to-pay in the story, and prove it before you trust it. Then carry the same pattern to the next area.

Under every step: people learning every week.

Under every step: knowledge written down.

Timing depends on your business.

Using AI

Agents do the work. People own it.

Let agents do every piece of work they can do well. Keep people responsible for the decisions, and make it normal for anyone to stop something that looks wrong.

Check the answers

Read the logs, and stop what looks wrong.

In the story, a graduate analyst reading the logs every morning finds wrong answers and asks to switch an agent off. The CFO agrees: “mostly works” is not the standard for something that speaks for finance.

Help leaders decide

Answer earlier, before commitments are made.

An agent prices a large bid by year and by risk within an hour. People spend the next two weeks deciding the terms, which is the part an agent cannot do.

Keep data permissions, testing, controls, and the logs part of the work. Plan for an agent that goes wrong or a provider that goes down.

How we help

Rebuild finance with the team you have.

Design what finance should do, decide where people must judge, and put agents to work one tested release at a time.

01

Design finance before buying tools.

Write down everything finance must do, then decide which work an agent does and which a person decides.

02

Leave your people more capable.

Make learning part of every week, write down what the team knows, and show every role where it fits.

03

Prove each change before you trust it.

Release one area at a time, keep controls inside the work, and measure what really changed.

The agents do the volume. The people do the judgment.

People make the calls where a mistake is costly, hard to undo, or a matter of trust, and they own everything the agents produce.

Keep the blueprint working.

Put controls inside the work, review the design on a schedule, and measure what changed and what did not. Time released by AI only counts when it goes into work the business values.

What would you like to change?

Explore the blueprint in the book, or talk with us about where to start.

Scroll to Top

Free companion toolkit

rebuiltfinance.ai

Start Monday with the people and tools you have.

The activity inventory, judgment test worksheet, agent job descriptions and prompt packs, Workbench reference, command center designs, amnesty kit, release playbook, quick wins, control matrix, learning program, cost guide, the numbers behind the story, and the Finance Rebuild Workplan.

Free with the book. Tell us where to send it.

We use your details only to send the toolkit and, if you ask, updates. See our Privacy Policy.

Chapter 1 excerpt

Start with a Clean Sheet

Read an excerpt from Chapter 1.

The board met on the second Thursday of the month, and Mark Bennett had three slides nobody wanted to present.

The first showed the cash forecast finance had given the board at the start of the quarter and the cash the company had actually ended with. The gap was not large enough to threaten anything. It was large enough that Helen Ashby, who had chaired the board for six years and remembered every number anyone had ever given her, had asked about it twice before the meeting started. Behind her question sat a date. The company’s revolving credit line came up for renewal in fifteen months, and the bank had seen the same forecast.

The third slide was the one Mark had argued about with himself for a week. It was a single quotation from the operating leaders’ survey Vivian Cole had commissioned in the spring: “Finance tells us what happened. By the time they tell us, it’s too late to matter, and when we ask for help, the answer is usually no.”

Nobody in the room was surprised by any of it. That was the problem.

Helen let the silence run a little longer than was comfortable. Then she said what she had come to say.

“This is a good company. It’s growing. It’s profitable. The finance team works hard and passes its audit. None of us thinks finance is failing.” She looked at Mark. “But we think it can be something it isn’t yet. We want finance to see problems before they’re problems. We want it to help the business leaders run their businesses, not just score them. And we want the people in it to come out of this more capable than they went in.”

Vivian picked it up. They had clearly rehearsed. “We want you to rebuild finance as an AI-native function. Not a pilot. Not a project. The function.”

Excerpt from Rebuilt Finance. The company and its people are fictional composites.