Most brands don’t have an AI problem. They have an operating-model problem that AI made visible.
We redesign how marketing work moves — what you own, what you rent, who decides, and how fast an idea becomes something live.
The shift
The gap isn’t capability.
It’s readiness.
Almost every marketing leader now says becoming good at this is critical. Far fewer say their organisation is actually ready. That gap is not a tooling problem — the tools are cheap and anyone can rent them. It’s a question of decision rights, approval chains, and who is allowed to ship something without asking three people first.
The failure pattern is well established by now. Projects get cancelled for three reasons, in this order: the cost climbed, nobody could point to the value, and the risk controls weren’t there. Most of that damage is done at procurement, before a single piece of work starts — because a great deal of what gets sold as an autonomous system is a chatbot with a new name on it.
We start on the other side of that. Not “what could AI do here,” but “which repeated decision in your marketing is slow, expensive or wrong, and what would have to be true for a machine to help with it.”
Client challenges
SOUND FAMILIAR?“We’ve run pilots for eighteen months and can’t point to a P&L line.”
We start with the decision, not the tool. Which repeated decision is costing you, what a better version would look like, and what would have to be in place for a machine to make it. Then we build backwards — read-only first, anything that writes or spends sitting behind an approval gate.
“Every team is using different tools and nobody knows what we’re exposed to.”
A tool inventory, a data-flow map, and a rights and consent audit. More often than not the answer is fewer tools with proper gates, rather than more tools with a policy document nobody reads.
“Our agency got faster and our costs didn’t move.”
We map where the efficiency actually went and rebuild the commercial relationship around what gets delivered rather than how many people were on it. Sometimes the answer is a different partner. Sometimes it’s a different contract. We’ll tell you which.
“We want to bring this in-house but we don’t know what ‘this’ is.”
A build-versus-rent map. Which capabilities should become assets you own — the brand model, the first-party data, the measurement — and which stay rented. Including the exit question nobody asks until they’re already leaving.
“Data protection is on the risk register and marketing owns it.”
Consent architecture, notice requirements across the languages you operate in, data governance, and a sequenced plan that clears the enforcement dates rather than arriving just after them. This one has a clock on it, so it goes first.
What we build
The way in to this capability is the Diagnostic. Everything below is what happens after it.
Marketing Operating Model Design
A redesign of how work moves from brief to live. Decision rights, approval architecture, the in-house and partner split, cycle-time targets, and the specific points where a human has to choose.
Build-vs-Rent Strategy
Which capabilities become owned assets and which stay rented. Includes what happens to each of them if the relationship ends.
AI Adoption Roadmap
A sequenced, costed plan for putting AI into a marketing function without walking into the usual failure modes. Read-only first. Anything that writes or spends behind a gate. An evaluation layer from day one, so you can tell whether it’s working before the budget conversation.
Growth Strategy
Where the next increment of growth actually comes from. Portfolio, audience, channel economics, category position. The classic work, done with better inputs.
Agency & Partner Model Review
Where the marketing spend actually goes, what each partner is really delivering, and whether the commercial model still makes sense now that hours have compressed. We don’t buy media and we don’t want to, which is what makes this review worth having from us.
FAQ
Frequently asked
questions.
Yes, and it’s often worth more there. A team of three with the right operating model outperforms a team of fifteen with the wrong one. The work is smaller, not shallower.
Sometimes. We’ll also tell you where we’re not the right people for part of it. That goes in the report, not into a follow-up call.
We build the thing afterwards. The recommendation and the delivery are the same team, which means the plan has to be one we can actually execute. That’s a useful limit on how ambitious a plan gets.
The Diagnostic is two to three weeks. Operating model work is six to ten. Neither asks you to stop what you’re currently doing.
Start with
the diagnosis.
Before we redesign anything, we’d rather show you what the system is actually doing today.