Ben Churchill has spent the past two years watching companies get AI transformation wrong in precisely this way. He co-founded ThoughtFox, which helps mid-to-large organisations across Europe. His full framework is worth reading on the Cassidy blog.
The short version: two parts of his framework get skipped more than any others.
One thing up front: PUSH is a UK and European Cassidy partner. We use Cassidy extensively ourselves and recommend it to clients because it helps non-technical teams build useful workflows without waiting for an engineering queue.
Key Takeaways
- AI projects stall when leaders do not use AI themselves and when the process underneath is unclear, slow or broken.
- Start with a business problem, an owner, today's number and a target. A tool without those four things is a subscription looking for a purpose.
- Check seven things before scaling: leadership, strategy, data, governance, skills, process and value. Score each from one to five and fix anything below three.
- ThoughtFox's Five A's give the sequence: Assess, Align, Activate, Accelerate and Amplify.
- Measure every use case against a baseline. Ten measured successes mean you have built a capability.
Why do AI projects go wrong?
Two dimensions of ThoughtFox's framework get skipped more than any others: leadership and operating models.
1. Misled by example
If leaders do not use AI themselves, staff quickly conclude that it is somebody else's project.
Leadership means demonstration. The executive opens the tool, tests the workflow and talks about what worked and what did not. Otherwise, people hear the message behind the message: "This is for them, not for us."
2. Weak operating model
Give AI a process that is unclear, slow or broken and you get that process at higher speed.
A bad approval flow becomes a faster bad approval flow. A vague creative brief produces more vague creative. A reporting process with no owner produces more reports and no decisions.
AI is very good at acceleration. So ask whether the thing you are accelerating is worth doing.
Why is starting with the tool the wrong move?
Starting with the tool puts the solution before the problem. Nobody has said what the business needs, so nobody can say whether the tool delivered it.
Most businesses start with a licence, a workshop and a promising demonstration. Then they ask people to find uses for it.
That is backwards.
Start with the business problem: what is slow, repetitive, expensive or difficult to see today?
Then answer four questions before anything is built:
- What problem are we solving?
- Who owns it?
- What is the number today?
- What number do we want, and by when?
If you cannot answer those questions, you do not yet have an AI use case. You have a software subscription looking for a purpose.
What should you check before scaling AI?
Check seven things before you scale: leadership, strategy, data, governance, skills, process and value. These come from ThoughtFox's seven dimensions of AI readiness. Skip the consultancy vocabulary and ask sensible questions.
Score each one from one to five. Anything below three deserves attention before you scale.
It is less exciting than buying a new tool. It is also the bit that stops the new tool becoming shelfware.
How do you get started with AI transformation?
Work through five phases, starting with a single process. ThoughtFox calls them the Five A's: Assess, Align, Activate, Accelerate and Amplify. The names sound like consultancy language. The sequence works.
- Assess means finding the gaps. PUSH's AI marketing audit is designed to do this for marketing teams.
- Align means agreeing the objective, guardrails and owner before a workflow is launched.
- Activate means selecting one useful use case, building it and measuring it. This is where PUSH's AI marketing adoption work begins.
- Accelerate means scaling what worked and stopping what did not.
- Amplify means making AI a normal part of how the business finds problems, builds solutions and improves them. PUSH's AI training builds that skill inside your team.
Do not try to transform every department at once. Pick one process where the pain is obvious and the result can be measured.
Good first use cases for a marketing team:
- turning weekly channel reports into one clear decision list
- analysing hundreds of ads to find which creative ingredients drive results
- finding wasted search spend across a large account
The goal is one meaningful bit of work made better.
How do you measure whether an AI use case worked?
Set a baseline before you build, then compare the result against it and record the lesson. Building an AI use case, launching it or winning applause in a workshop does not make it successful.
Set a baseline. Run the workflow. Compare the result. Record the lesson.
Every use case needs a short card:
- The problem
- The owner
- Today's number
- The target
- The review date
That is enough. Systems beat experiments. A single successful workflow is useful. Ten successful workflows mean you have built a capability. That is when AI stops being a project and starts becoming an advantage.
For the wider UK picture, see PUSH's AI business transformation statistics for the UK.
How can PUSH help with AI transformation?
PUSH helps marketing teams identify high-value AI use cases, put the right data and guardrails around them, then build and measure workflows that make a commercial difference.
We add AI where it produces better creative decisions, less wasted media spend, faster insight and more time for the work people should be doing.
Read Ben's full framework on the Cassidy blog. Or, if you want to identify your first measurable marketing use case, talk to PUSH.
The first workflow is the easy part. The tenth tells you whether you have changed the way the business works.
Frequently asked questions
What is an AI transformation framework?
An AI transformation framework is a structured plan for building a business's ability to find, build, measure and scale AI use cases. ThoughtFox's version covers seven dimensions of readiness and five delivery phases: Assess, Align, Activate, Accelerate and Amplify.
Why do AI projects fail?
AI projects fail when nobody agrees what success looks like and nobody evaluates the result. Two causes come up most: leaders who do not use AI themselves, and processes that were unclear, slow or broken before AI arrived.
What are the Five A's of AI transformation?
The Five A's are Assess, Align, Activate, Accelerate and Amplify. Assess finds the gaps. Align agrees the objective, guardrails and owner. Activate builds and measures one use case. Accelerate scales what worked. Amplify makes AI a normal part of how the business operates.
What should you check before scaling AI?
Check leadership, strategy, data, governance, skills, process and value. Score each from one to five. Anything below three needs attention before you scale.
How do you measure whether an AI use case worked?
Set a baseline before you build. Run the workflow, compare the result against the baseline and record the lesson. Each use case needs a card with the problem, the owner, today's number, the target and the review date.
