There was a time when working on a strategy meant having time to think. Enough to understand the problem, talk to the client, challenge the brief and argue with the creatives.
Then deadlines got shorter. Teams got leaner. Strategists were stretched across more accounts. Despite all the technology meant to make life easier, we kept starting from scratch. The big pitch got the senior strategist. The smaller account got whoever had time.
People still cared about good strategy. The conditions for producing it were getting worse. And then AI arrived.
We've been here before.
Programmatic advertising promised precise targeting and efficient delivery at enormous scale. Useful things. Somewhere along the way, the industry began confusing efficient delivery with effective advertising.
David Abbott put the problem rather elegantly:
“Shit that arrives at the speed of light is still shit.”
Quoted in Mark Ritson’s CIM Levitt Group Annual Lecture, 2019 (p. 27)
The programmatic advertising comparison is ours.
With AI, the same fascination with speed has reached production. More headlines, concepts, variations and presentations. Much of the debate revolved around recreating the Coca-Cola Christmas ad or insisting that AI could never be creative.
Why was everyone starting with creativity?
Strategy has methods: questions to answer, evidence to examine, relationships between insights and choices that give creative work a direction. It also needs judgement, imagination and the occasional leap no framework can manufacture.
If AI could work with information, recognise patterns and follow structured reasoning, strategy looked like an interesting place to apply it. The methods of good strategic work could become available across more assignments, with strategists responsible for the decisions.
That was the idea behind Mirel. We thought it made sense. We underestimated just how difficult it would be.
01 / Workflows in n8n
Our first attempt was pretty terrible.
We started with workflows built in n8n, a tool for connecting software and automating tasks.
On paper, the idea was straightforward. Break strategic work into defined steps. Use AI to help complete them. Connect the results into a larger process.
In practice, we discovered that a technically structured workflow and a strategically useful workflow are not necessarily the same thing.
Getting a meaningful insight from one stage into the next was surprisingly difficult. The system followed technical processes, but strategic reasoning doesn't always fit neatly into them.
Worse, the user had to know exactly what they wanted to do.
If they didn't, they could end up somewhere entirely different from where they intended.
We had taken the flexibility of tools like ChatGPT and Claude, restricted it, and created something that was in many ways less useful.
We were a long way from the breakthrough we'd imagined.
So we changed the technical foundation and built something more flexible.
02 / The chat interface
Then we built a chatbot. That was better. And still wrong.
The new system was a considerable improvement.
We could offer specific strategic tasks, connect more of the work and give people considerably more freedom to explore.
But there was a problem we hadn't properly understood.
A chat interface invites a conversation. A strategic assignment needs a process.
Put an empty chat field in front of someone and you're inviting them to ask a question, follow an interesting thought, try another prompt, explore another angle.
There's nothing wrong with any of that. Exploration is an important part of strategy.
The rest of the assignment still needs doing.
At some point, you have to stop and ask whether you've actually solved the problem in front of you.
Is the audience insight useful? Does the positioning hold up? Have we established something worth building on? Or are we just producing another plausible-sounding answer?
We discovered how easy it was to generate something that looked useful, get pulled into another conversational rabbit hole, and lose sight of whether the previous step was actually finished.
And there was a deeper problem.
Even with predefined workflows and buttons to start specific tasks, the user still had to know which work needed doing.
Think about that for a moment.
We wanted to make experienced strategic methodology more accessible. Yet our product still expected people to bring much of that expertise with them.
We were asking users to design their own strategic process before Mirel could help them execute it.
We had improved the technology. But we hadn't fully solved the problem.
03 / Starting with the assignment
The process had to become the product.
That realisation changed how we thought about Mirel.
We decided to start with the assignment.
Give Mirel a brief. Let it examine what needs to be done and propose a strategic process.
Then let the strategist review that proposal, change it where necessary, and work through the assignment with a clear sense of where they're going.
The process would have to adapt to each problem.
It would also need a clear direction, so each new answer helped the work move forward.
A structured, adaptable way of working that connects the thinking, makes decisions visible and gives people the opportunity to challenge the work before moving forward.
In other words, something closer to how an experienced strategist actually approaches an assignment.
The AI helps organise the work and develop the thinking. The human decides what is good enough, what needs another look and what deserves to become the foundation for the next step.
Exploration still matters. Progress requires us to decide what to do with it.
And we stopped expecting the user to do all the work of figuring out what the system should do next.
Faster was never the point.
Of course, Mirel can make parts of strategic work faster.
It would be rather strange if it couldn't.
But speed is a poor measure of strategic quality.
A weak positioning developed in ten minutes is still a weak positioning. A beautifully formatted presentation doesn't make the argument inside it any better. And a hundred plausible audience insights aren't worth much if nobody stops to ask which one actually matters.
We judge the work by whether it helps someone arrive at a better decision.
The strategist stays responsible for developing and approving the strategy.
Mirel helps people do the work properly, with more structure, more continuity and more room for the judgement that makes strategy valuable.
Strategists need room to think.
We've worked with enough talented people in agencies to know what they're capable of.
We've worked with enough brilliant strategists to know what a difference they can make. We've also seen how often their time gets consumed by starting over, rebuilding context, managing the process and doing work that leaves too little room for the part they're actually brilliant at.
Meanwhile, smaller accounts still need strategic thinking. So do the everyday assignments that will never become the agency's biggest pitch.
The quality of that work shouldn't depend entirely on whether the most experienced strategist happens to have a free afternoon.
We want Mirel to help agencies raise the standard of strategic work across more of their business.
To give less experienced strategists a stronger foundation to work from. To let senior strategists spend more time challenging ideas, guiding clients and collaborating with creatives. To make good strategic discipline available even when time and resources are limited.
And perhaps most importantly, to give strategists more opportunity to do the work that made them want to become strategists in the first place.
Not just finding a good answer.
Finding the brilliant one.
