Your team can use ChatGPT, Claude, Copilot, agents and automations every day — while the same important work still comes back to the same people.
Because making individuals faster is not the same as making the organisation more capable. AI becomes capacity when human expertise, well-designed work, systems and AI combine so valuable work can move reliably to standard.
How much AI are we using?
What can the business do now that it couldn’t reliably do before?
Your team may already be using AI well. Documents arrive faster. Research that took hours takes minutes. Meetings summarise themselves. Someone has built an agent. There may even be a respectable little zoo of automations quietly doing useful things in the background.
And yet…
You still add the client context that makes it right.
You still decide what matters, what looks wrong and what happens next.
You are still the person who knows whether the result is good enough.
The firm’s capacity hasn’t followed it.
The real dependency sits somewhere else.
Imagine the process says: Prepare the monthly client report.
It sounds perfectly clear. Until someone else — human or AI — actually tries to do it.
For years, you may have supplied the missing answers almost without noticing. That’s the invisible job specification. The process described the task. You supplied the judgement that made the task work.
Nobody had needed to make them explicit while the founder was quietly supplying them. So before asking “What can we automate?” ask “What does this job actually require?”
A folder full of SOPs can be useful. It can also become a beautifully organised record of everything the founder still needs to approve. Real transfer involves four things.
Documentation matters. Ownership matters more.
These are examples of the kinds of recurring jobs where the distinction becomes obvious.
The meeting happens. AI produces immaculate notes. Then everyone gets busy. A week later: “What did we actually decide?” The founder becomes the memory system.
The workflow turns the meeting into:
AI helps extract the work. A named person owns what happens next.
The output stopped being meeting notes. It became an owned job.
The information exists everywhere. Email. CRM. Projects. Messages. People’s heads. So the founder reconstructs the state of the business from six conversations and a strong cup of coffee.
Agreed information feeds a consistent brief. AI prepares the synthesis. A named person checks it against defined standards. The founder sees the few things that genuinely require attention.
Information became usable operating context.
The team creates 80%. Then the founder adds the client context, commercial nuance and the bit that makes it sound like the firm rather than a competent stranger.
Relevant context, strong examples, quality criteria and decision boundaries become part of the workflow. Routine correction disappears. Commercial exceptions still escalate.
The team gained access to more of the judgement behind the work.
Someone produces the numbers. The founder explains what they mean and tells everyone what to do.
The workflow includes interpretation criteria, thresholds, exception triggers, ownership, decision rights and escalation. AI can surface patterns. People remain responsible for decisions.
The work moved from reporting information to supporting action.
You probably do.
Choose recurring work worth moving.
Capture the context, judgement, standards, examples, exceptions and red flags it actually depends on.
Decide ownership, human/AI roles, decision rights, boundaries and escalation.
Use the simplest useful combination of people, AI, automation, knowledge and tools.
Run real work. Find the failures. Fix the design.
Move ownership beyond the founder and check whether it holds.
A three-agent orchestra is not automatically better than one well-designed workflow and one competent human. No medals are awarded for unnecessary orchestration.
Responsible AI does not mean letting software wander around the business like a Labrador with the office keys. It also doesn’t mean the founder approves everything forever. That just recreates the original problem with nicer technology.
For bounded, repeatable work with clear standards and controlled risk. A person remains accountable for the system. They don’t need to inspect every routine output.
Where judgement, relationships, interpretation or meaningful commercial risk remains. AI removes preparation work. A person owns the decision.
Where the situation is highly sensitive, genuinely novel or consequential. Leadership. Culture. Critical relationships. Material commercial or reputational decisions.
I came to AI from the organisational side. For more than 25 years, I’ve worked with strategy, leadership, decisions, accountability and execution across organisations including BHP, Microsoft, Vodafone and LexisNexis and with founder-led businesses.
The technology is new. The questions underneath it aren’t:
AI makes those gaps dramatically more visible. And when the work is designed well, it gives us a powerful new instrument for addressing them.
I’m not interested in getting your business to use more AI for the sake of it. I’m interested in what the business can genuinely do better because AI is there.
“Genuinely seeks to understand and design processes that suit the client’s business — not a one-size-fits-all approach like so many other consultants.”
Not every company needs an “AI transformation”. Thank heavens. There are two practical routes.
Three recurring jobs become firm-owned in four weeks. For established founder-led firms with three suitable recurring jobs that still rely too heavily on founder context, judgement or correction.
Real work. Real owners. Live testing. A 30-day durability check.
Redesign the work. Then implement the AI. For a more complex workflow or broader implementation challenge.
This can involve workflow redesign, knowledge architecture, agents, automation, integrations, permissions, testing, handover and adoption. The scope depends on the work. But the philosophy does not:
Start with the work. Choose the technology second.
Not sure which applies? Tell me what work keeps coming back →
If three recurring jobs immediately spring to mind.