AI agents

AI agents, built one workflow at a time. Your people stay in charge.

Agents come after the War Room, not before it. First you see where the business slips. Then we map how the work gets done, decide step by step what an agent should do, and build the most valuable agent first.

Why the War Room comes first

Agents built on guesses get ignored.

Most AI projects start with a tool and look for a problem. We start the other way round. The War Room runs your rules on live data and shows where the business is slipping, every day, by department.

That tells us which work matters. Mapping then tells us where the hours go. An agent is worth building where the two meet: a workflow that keeps causing flags and eats your team’s week.

A War Room flag with the facts, why it happened and the plan to fix it
Step 1: a flag, with what, why and how. Illustrative.
Step 2 · The mapping template

Filled in with the person who does the work.

MAP

The steps

What starts it, each step as a verb, who does it and their hours a week, systems touched, how often, and what goes wrong.

SCORE

Value and ease

Value is hours a week. Ease is how ready the data is and how rule-based the work is, less a point if it crosses three or more systems.

DECIDE

Each step

Agent, agent with approval, or stays with your team. Anything costly to get wrong keeps an approval.

HOURS

Time handed back

Hours a week returned to each person, labelled PROJECTED until the agent is live and measured. Shown in hours, never money.

Today
A workflow as it runs today, every step done by a named person
With agents
The same workflow with each step assigned to an agent, an agent with approval, or the team
One workflow, before and after the decision on each step. Illustrative.
Where to build next

Value against ease decides the order.

Every mapped workflow lands on one chart. We build from the top right, take quick wins in between, and flag the workflows whose data needs fixing first.

Build firstHigh value, easy to automate.
Quick winsEasy, fewer hours.
Fix the data firstHigh value, data not ready.
Leave for nowLow value, hard to automate.
Where to build next: workflows placed by hours a week against ease to automate, in four quadrants
Illustrative workflows for a sample manufacturer.
Step 3 · Agents at work

They prepare the work. A named person approves it.

Mapped→Scored→Approved→Building→Live
  • Agents read portals, compare schedules, chase suppliers, draft reports and post entries.
  • Anything that commits money, stock or a customer waits for a person’s OK.
  • Every run is logged, with what the agent did and who approved it.
95%

An agent keeps its approval step until 95% of its work is approved without edits over 50 runs. The bar is set per workflow, and you can raise it.

The Agents section: agents live by stage, runs today, approved without edits, hours returned projected, and today's agent actions
Preview of the Agents section. Workflows, people and figures are illustrative.
Where agents start

The first agents we look at, by industry.

Every business is mapped from scratch. These are the workflows that usually come up first.

Manufacturing

Orders, plant, quality, purchase

  • Customer schedule intake from OEM portals
  • Raw material PO chasing
  • 8D report drafting
  • Tool, die and machine selection
AI for manufacturers →
Automotive dealerships

Sales, service, used cars

  • Lead follow-up
  • Service reminders
  • Valuation chasing
  • Stock clearance plans
AI for dealerships →

Agents give your people their time back.

Agents take the copying between systems, the chasing and the checking. Your team keeps the judgement calls, the customer conversations and the final say. We report the hours handed back to each person.

Agents run inside the same boundary as the War Room. Your data stays under your control. Read how.

For owners and CEOs

Find your first agent.

A 30-minute walkthrough, with your industry in mind.