03 / AI Operating Model

Your best way of working, built into the AI your whole team uses.

A few people on your team get good results from AI, each in their own way, with their own prompts and their own version of the brief. Everyone else gets something different. An AI operating model turns your unique processes into skills and automations the whole team runs from the same inputs, so the quality doesn't depend on who's asking. It builds the proficiency to run them and puts the software in your business rather than in ours, so the way you do things scales past the people who invented it. This is where your AI gets its rocket fuel, and where you start building IP out of how your business works.

Book a call

6 to 12 weeks a buildFrom £8,000

The man showing the robot how he does a task, while the robot builds a row of small machines that each do it the same way.
crm email project board accounts team chat shared drive what thebusinessknows campaign brief blog draft social posts from long form newsletter competitor watch campaign agent account research call prep proposal first draft follow up email crm tidy deal desk agent monthly client report meeting notes to actions onboarding pack renewal risk flag new client setup client health agent supplier checks policy answers project status summary invoice chase month end pack ops desk agent board pack summary weekly pulse decision log all hands notes board prep agent marketingsalesclient servicesoperationsleadership

hover or tap anything for detail. pick a team to zoom in.

your tools what the business knows your teams one release, everyone updated skill automation
The operating model, as we draw it.

How the four parts work together

This is the hierarchy, top to bottom: who decides, what starts the work, what does it, and what it all stands on.

People

Set the goals, build the parts and sign off what goes out.

direct set up

Agents

Work towards a goal and pick the skills.

Automations

Start on a time or an event, by themselves.

use call

Skills

One job, done the same way every time.

  • account research
  • blog draft
  • monthly client report
  • supplier checks
read from

What the business knows

positioning, ideal customer, messaging, tone of voice

What we do in AI Operating Model

Your processes, written as skills

Most teams we meet have the same skill built three times by three people, with copies passed around and changed without the others knowing, most of them one file of instructions. We build one library where skills work together in a hierarchy: your PowerPoint template is one skill that every skill making slides calls, so a change to it reaches them all.

Produces One library the whole team works from: everyone gets the same best version, and nobody builds their own from scratch or keeps a copy to themselves.

Pilot programme design and delivery

We take the first skills off the backlog and pilot them with the team best placed to run them, in your environment, on your data, with success agreed before it starts.

Produces A pilot of the first skills, with success agreed up front.

Phased rollout management

Going from the pilot to every team is where new edge cases turn up. We handle the integration, the adjustments and the curveballs as each team goes live.

Produces Every team live, with the integration and the adjustments handled.

Capability transfer

Every skill comes with documentation your team can follow and a handover that leaves them able to change it without us. If you'd rather we ran it for you, we can, on a monthly retainer.

Produces Documentation for each skill, and a team who can change it.

You leave with

  • Live workflows inside your own stack
  • Documentation for each one
  • A team who can change them without ringing us

Operating models we've built

Ice Blue Sky

An operating model of 29 skills so far, each one built to make the agency's work faster and its output better and more consistent. Here's what two of them did.

29
skills in their operating model so far
5 hours to 20 minutes
on pitch proposals
Over 50%
less time to make branded slides

B2B analytics firm

A manual keyword matching process that took nine days, replaced by a script that handles the matching in 12 seconds. One day instead of nine, with the team confident in how to use it and when to override it. For the same client, a data enrichment workflow automated and running on $100 a year in API credits.

9 days to 12 seconds
keyword matching, one day end to end instead of nine
90%
reduction in processing time on data enrichment
$100 a year
in API credits to run it

What this won't do

Build you a tool of its own. When the answer is a piece of software rather than a set of skills and workflows, that's the next rung.

Read about Software Build next