AI is the new Agile Transformation

Not so long ago… in a tech hub not so far away… the mantra was everywhere: “We need to adopt Agile. We need to be more agile. We need to go quicker and respond to volatile market conditions.”

Some of those transformations are still happening kind of. Like every large‑scale transformation, they start, they implement an operating model, the consultants roll off and kablam: You are “Agile”.

But are you Agile? Well… not really.

That Target Operating Model (TOM) was probably sound and most likely copied and pasted from the last Exec Steer Co at [Insert blue‑chip org]. But it’s still just an operating model. The real implementation work, the embedding, the behavioural change required to actually embrace agility never came from that TOM.

And those same exec members who bought into it? They never changed. The sales pitch was irresistible: “We can help your teams execute faster, handle change more dynamically, and make sure you can have your cake and eat it.” So it’s sold! The answer to all organisational woes.

Then comes the post‑consultant reality. Shareholders want better performance. Execs set impossible challenges, compress timescales, remove empowerment. And the customer? Well, they just pay money to be inconvenienced, right?

It sounds crazy, but this pattern has repeated for years. It’s been highly effective for many consultancies, big or small.

Most Agile transformations have subsequently crashed and burned. Early SAFe adopters are no longer shouting about its benefits. At best, agility survives in small pockets of tech teams — places where exec oversight hasn’t yet spotted that culture and morale still exist. (I am joking!).

So what has this got to do with AI?

In 2026, there’s a growing pressure in organisations: “We need to adopt AI.”

It’s the Agile rush all over again… Transformation as a badge of honour, not a strategic decision. No real buy‑in. No commitment to the change required to truly embrace it.

The benefits being sold are phenomenal and in equal measure, wildly overstated. Balance sheets are filled with attractive numbers: multi‑million £ / $ savings, better ROI, reduced employment risk, fewer tax and NI implications. It’s compelling. I get it.

And this is where the consultancies circle, like sharks with fresh chum in the water. The model is familiar: hire a visionary, snake‑oil salesperson who sounds convincing. They’ve done some vibe‑coding, knocked out a few apps with AI. Great.

Business execs continue to buy into the idea of bringing in the “best and brightest” consultants, most of whom have no real experience implementing the solution, nor truly understand it.

No one stops to ask: “What outcome are we trying to create that we can’t achieve today?”

Instead, organisations go all‑in. The stakes are too high. Everyone needs to save face. Saying no becomes a dirty word. “Yes, and…” becomes the new theatre.

The uncomfortable truth

There’s a long‑standing saying in IT and Data: rubbish in / rubbish out.

LLMs learn from human interactions, inputs, and behaviours. Dirty data is still a risk. So is “good” data used in the wrong context.

Off‑the‑shelf integrations with mainstream LLMs are already producing wild outcomes and driving huge amounts of misinformation. You could blame a rogue prompt, but users are not easy to replicate. Ask any seasoned test engineer.

That user who clicked the button 15 times for no reason and crashed the system? That was never a test case. But it happens.

And this is the truth about implementing these tools: a human can recognise and adjust for the unexpected.

Can an LLM handle those weird scenarios? Maybe. Maybe not. And when it gets it wrong, you get: “Yes, you are right, that wasn’t the right answer” without offering the right answer. Ask again: “We are having difficulties answering this question right now, please try again later.”

Infuriation ensues. CSAT scores tank. Trustpilot gets a fresh batch of juicy reviews.

Before adopting AI, leaders should ask:

If the answer is unclear, AI won’t fix it. If the problem is cultural, AI won’t fix it. If the process is broken, AI will only accelerate the chaos.

Because leadership isn’t about chasing the next methodology or technology. Leadership is the discipline of choosing what not to do.

AI can be extraordinary. But only when it’s intentional, not performative.

Sometimes the bravest decision is to pause, think, and choose the path that actually serves your people and most importantly… your customers.


Author:
David Massa, Senior Programme Manager at Sky, specialising in enterprise technology recovery and large-scale transformation, bringing complex, high-risk programmes back under control and delivering measurable outcomes.