Read more insights
Want to get started?
Book a 1:1 session with our team to see how Cassidy can support your goals.
Book demoWhat Good AI Transformation Actually Looks Like: A Framework for Getting It Right
AI transformation isn't a technology problem. Assume the technology can do the job. The harder question, the one most organizations aren't asking clearly enough, is whether the organization is built to absorb it.
Ben Churchill, co-founder of ThoughtFox, an enterprise AI transformation consultancy working with mid-to-large organizations across Europe, has spent the past two years watching companies get this wrong in predictable ways. The tools aren't the bottleneck. The organizational scaffolding is. "Enterprise transformation isn't a technology problem," Churchill says. "It's more about organizational scaffolding and capability."
ThoughtFox has developed a structured approach to AI transformation built around two parallel tracks: building organizational capability across seven dimensions, and delivering measurable use cases that create the momentum to keep going. Neither works without the other.
Why AI Projects Stall
The failure pattern Churchill sees most often isn't dramatic. There's no crisis moment, no public failure.
Projects just quietly stop mattering.
Someone had an idea, tried something, and nobody could say whether it worked or not. It fizzled. The root cause is almost always the same: no clear reason to do it in the first place, no definition of what success looks like, and no evaluation afterward. Without those three things, even a genuinely useful AI deployment can't build momentum.
Marie Toft, ThoughtFox's AI adoption lead, describes a version of this she sees repeatedly: organizations treating AI as a solution in search of a problem. "There's this race to say, let's just get AI," she notes. "And if you're not strategic about it, if you're not saying, what business problem is this actually solving, where are we now, what are our KPIs, where do we want to be, then that's where an awful lot of people are trying things and the pilot isn't working."
The fix isn't more technology. It's more clarity before the technology gets deployed.
The Seven Dimensions of AI Transformation
ThoughtFox structures enterprise AI transformation across seven dimensions. The reason for this breadth is deliberate: siloing AI in IT or legal is one of the most reliable ways to kill adoption before it starts. Real transformation touches every part of how an organization operates.
The seven dimensions are:
- Leadership — Are executives visibly using AI, not just sponsoring it?
- Strategic alignment — Does the AI strategy connect directly to business outcomes?
- Technology infrastructure and data — Is the underlying data clean, accessible, and governed?
- Governance and ethics — Are there clear boundaries for what AI can and can't do?
- People, culture and skills — Do employees have the knowledge and confidence to use AI well?
- Operating models and processes — Are the processes AI is being applied to actually working?
- Value realization — Is anyone measuring whether this is working?
Each one matters. But Churchill is particularly pointed about two that get skipped most often.
Leadership isn't just about sign-off. It's about demonstration. "Leadership demonstrating that they are using AI as well, then you're far more likely to get pick-up within the organization," Churchill says. An executive who talks about AI transformation but isn't visibly using the tools themselves sends a signal that adoption is someone else's job.
Operating models and processes is the dimension that trips up even well-resourced organizations. Before layering AI onto a process, the question has to be: does this process actually work? "There's no point in slapping an AI solution on a process that isn't working," Toft observes. AI can accelerate a good process. It will accelerate a broken one too, just toward worse outcomes, faster.
The Five A's: A Phased Approach to Capability Building
ThoughtFox guides organizations through AI transformation in five phases, which they call the Five A's. The framework is designed around a core insight: you build organizational capability and deliver results at the same time. Results fund more capability. Capability enables bigger results. That's the force multiplier.
The five phases move from assessment through to full organizational embedding:
- Phase 1 — Assess. A maturity assessment across all seven dimensions. Where is the organization today? What are the gaps? What's the realistic starting point?
- Phase 2 — Align. Leadership, strategy, and governance brought into alignment before deployment begins. No skipping this.
- Phase 3 — Activate. First use cases selected, built, and measured. This is where momentum either starts or doesn't.
- Phase 4 — Accelerate. Proven use cases scaled. Capability expanding across teams and departments.
- Phase 5 — Amplify. AI embedded into every process. Every team aligned. The organization can find, build, and scale workflows continuously on its own.
Getting from phase one to five isn't a single project. It's a sustained program, with each stage reinforcing the next.
The PRIME framework sits underneath this, guiding how individual use cases get selected, measured, and evaluated. Churchill is direct about why measurement matters so much: "You create momentum by delivering use cases that are measured to be successful." Without measurement, there's no momentum. Without momentum, there's no transformation. Just a series of disconnected pilots that never add up to anything.
What This Looks Like in Practice
Churchill's work with enterprise clients using Cassidy illustrates how the framework plays out in practice. One of the clearest examples from his consulting work at Ocuco, a leading supplier of software to the optical industry, : the decision to deploy Cassidy wasn't just about finding a capable AI platform. It was about finding one that non-technical people could actually use.
"The fact that non-technical people could deploy usable workflows in Cassidy was the game changer for us," Churchill says. When automation requires engineering resources to build and maintain, it stays in IT. When anyone on the team can describe what they want and have a working Workflow running in hours, transformation stops being a project and starts being a practice.
The other factor Churchill cites consistently is security. Enterprise clients handling sensitive data, PII, financial records, health information, need to know the governance piece is handled before they'll engage at scale. Cassidy's SOC 2 Type II certification, GDPR compliance, and role-based access controls meant that conversation didn't have to slow the deployment down. It was already answered.
The Question of "Done"
Churchill's answer to whether AI transformation is ever finished is simple: it isn't.
The technology keeps changing. The use cases that weren't possible six months ago become standard practice. The organizations that treat transformation as a destination, a project with an end date, miss the point entirely.
The ones that treat it as an operating model do not.
Continuous, measured, culturally embedded — that's what compounding advantage looks like in practice. The gains from any single workflow are real. The gains from an organization that knows how to find, build, and scale workflows continuously are a different order of magnitude.
That's the real argument for doing this deliberately. Not just to deploy AI faster, but to build the organizational capability to keep deploying it well.
ThoughtFox works with organizations from initial maturity assessment through to full deployment and adoption programs across all seven dimensions. To see how Cassidy supports the workflow and automation layer of that transformation, book a demo and walk through what your first high-value Workflows could look like.
Build AI Agents that operate like your team.
Cassidy Inc. © 2026