Most organisations today say they are doing something with AI, but privately, many leaders will admit they are unsure whether any of it will stick. The data backs that up:

• 70% of AI initiatives fail to deliver business value

• 89% stall because teams do not understand the operating model changes required

• 54% fade because people feel unsafe speaking up as the risk increases

These are not technology problems. They are leadership, system and learning problems.

And paradoxically, in this moment where AI is accelerating at an exponential pace, the answer is not to buy more platforms. It is to return to the fundamentals we once learned, then often forgot: human centred, iterative learning; clarity of purpose; and the ability to build safe, adaptive systems around people.

In other words, the foundations of Agile are not outdated. They are more relevant now than at any time in the last two decades.

Why the vendor-led model is failing us

The speed of AI means vendors can now produce compelling demos, proofs and prototypes in minutes. But beneath the excitement sits a predictable trap: the business becomes the passenger; the vendor becomes the driver.

When that happens:

• AI is done to people, not built with them

• Teams do not understand the why behind decisions

• Capability does not transfer

• Leaders are asked to govern systems they did not help design

• Trust breaks down

AI succeeds should be helping organisations learn faster than the pace of change. I'm not sure it's there yet.

The real bottleneck is not the tech but the learning speed

Over the last 18 months at Source Agility, we have seen something interesting. AI is obliterating the gap between idea and prototype. What used to take cross-functional teams weeks can now be generated in an afternoon.

But here is the uncomfortable truth many organisations are now discovering:

AI accelerates product iteration but exposes (or amplifies) organisational dysfunction faster than ever.

This is why Agile fundamentals matter more now: fast feedback, safe experimentation, co-design, working with evidence rather than assumptions, iterative delivery, and learning faster than competitors.

AI has made the speed of organisational learning the new competitive advantage.

Our own shift: from Agile coaching to delivery performance to AI adoption

This is exactly why Source Agility shifted direction.

We did not abandon Agile. We doubled down on the parts of it that matter.

When we began, we were a coaching-led practice focused on helping delivery teams move faster, align better and meet commitments. Then we moved into delivery performance, helping organisations create predictable, high confidence systems for technology delivery.

But over the last few years, something became clear: every conversation about delivery now includes a conversation about AI.

Not hype. Not maturity models. But real questions:

• How do we use AI to reduce waste without overwhelming our people?

• Where are the opportunities hidden in our operating model?

• How do we move at speed but responsibly?

• How do we scale a good AI idea?

Our About page tells the honest story of our journey. We did not start in boardrooms or strategy labs. We started in the trenches, watching teams wrestle with complexity, burnt out leaders trying to keep promises, and organisations looking for a better way to work.

AI did not change that mission. It sharpened it.

Our work in AI adoption and scaling is the next natural step: helping organisations make promises they can keep with systems that actually work.

If AI is the accelerator, Agile is the steering

AI generates more options, faster. Agile provides the structural rails so that speed does not become chaos.

AI makes iteration cheap. Agile makes iteration meaningful, grounding it in human needs rather than vendor promises.

AI can rapidly build solutions. Agile ensures we design with the people those solutions are meant to serve.

AI can run. Agile tells us whether it is running in the right direction.

And perhaps most importantly, AI makes it tempting to push ahead because the prototype looks convincing. Agile gives us the discipline to test assumptions early and to stop an AI idea quickly when the value is not real.

Stopping something early is not failure. It is responsible leadership. Especially now that AI makes it easy to produce solutions that look polished but sit on invalid assumptions.

AI expands possibilities. Agile protects organisations from wasting time on the wrong ones.

A way forward

This moment demands something simple but difficult: slowing down enough to make sense of what’s real, what matters, and what can be safely ignored. It requires clarity about value, not more noise. It requires conversations that are honest rather than optimistic. And it requires the ability to stop work as easily as we start it.

If you are trying to navigate all of this, and you want a place to step back from the pressure and get your bearings, we are running a session called AI Reality Check. It is designed to help leaders cut through the hype and reset direction for 2026. Learn more here.