Your business doesn’t need more AI. It needs Workflow Intelligence.
The most useful thing I brought to a recent AI workshop was a pen.
We had three hours together. There were real pitches on the table, a room full of experienced people and plenty of appetite to use AI.
We barely talked about AI.
Instead, we looked at how the work actually happened.
Who decided whether an opportunity was worth pursuing? Where did the information come from? Who owned the next step? Where did hours disappear? What had to be checked by a human, and what was being recreated from scratch every single time?
The client did not really have an AI problem. They had a workflow problem, and AI had exposed it.
This is happening everywhere at the moment. Businesses are buying tools, building agents and asking people to find clever ways to use them. The technology is arriving faster than anyone can sensibly redesign the work around it.
So AI gets bolted onto whatever already exists.
A messy handover becomes an automated messy handover. A vague brief becomes a vague brief generated in four seconds. A sales team produces more activity without becoming any clearer about who deserves their attention.
It is quicker. It is also still a muddle.
That is what I mean by Workflow Intelligence.
Workflow is the structure around the work. The stages, decisions, tools, handovers, responsibilities and rhythms that allow something useful to keep moving.
Intelligence is the human layer. The experience to recognise what matters. The judgement to make a decision when the answer is not obvious. The understanding of how people really behave, rather than how a process map says they should behave.
You need both.
A workflow without intelligence becomes bureaucracy. Intelligence without a workflow stays trapped inside somebody’s head.
Start with the work, not the software
When I look at a business, I am not particularly interested in how impressive its technology stack looks on a slide.
Show me the work.
Show me the opportunity that went quiet because nobody followed it up. Show me the report that takes a day because one experienced person has to correct everything. Show me the founder’s voice note that contains six good decisions and vanishes beneath another 43 WhatsApp messages.
That is where the useful design work begins.
The sequence is usually quite simple:
Look at what really happens.
Find the friction, lost decisions and repeated effort.
Decide where human judgement genuinely matters.
Redesign the workflow.
Use AI and automation where they remove unnecessary load.
Not every task should be automated. Not every decision should be handed to an AI model. The objective is not to engineer people out of the business.
It is to stop wasting them.
AI is very good at searching, sorting, summarising, comparing, drafting and remembering. Automation is good at moving information, creating prompts and making sure the next step appears in the right place.
People are still better at context, trust, taste, ambiguity and the slightly awkward judgement calls on which businesses actually run.
The clever bit is deciding where each belongs.
The brain-friendly version is usually the better version
My own brain has made me unusually interested in weak workflows.
If a system relies on me remembering something at the right moment, maintaining motivation through every stage and returning calmly to a half-finished task three days later, it is not a particularly robust system.
For years, I treated that as a personal failing. I tried harder, bought another notebook, changed task manager and occasionally created a colour-coded dashboard ambitious enough to deserve its own project manager.
None of that changed the underlying design.
A better system externalises the remembering. It makes priorities visible. It reduces the number of tiny decisions needed to restart. It captures useful thinking in the form it naturally arrives, then gives it enough structure to become useful to someone else.
That is neuroinclusive design in practice. It is not lowering the standard or creating a special process off to one side. It is removing unnecessary cognitive load from the main workflow.
That helps people with ADHD, dyslexia and other forms of neurodivergence. It also helps the busy account director, the overloaded founder and the perfectly neurotypical person who has been in meetings since half past eight.
Good ramps help everybody carrying luggage.
What this looks like commercially
The same principle applies to sales and growth.
Most founder-led businesses do not have an information shortage. They have names in the CRM, old conversations, event notes, LinkedIn messages, referrals, half-written proposals and a founder who knows exactly which three relationships matter most.
The problem is that the judgement and momentum live in different places.
A well-designed commercial workflow can use technology to monitor signals, gather context, surface neglected opportunities and prompt the next useful action. But it should not let an AI impersonate the founder or fill the internet with synthetic small talk.
The system does the remembering. The person has the conversation.
That is Workflow Intelligence too.
It can be applied to a sales engine, a leadership team, an agency pitch process or an individual trying to get important work out of their head. The tools will change. The design principle does not.
Get the decisions out of individual heads and into shared systems. Automate the parts that do not require a person. Protect the judgement that does.
Because if important work only moves when one particular person remembers, notices or feels sufficiently motivated to push it forward, you do not yet have a workflow.
You have a dependency.
