- Graeme King – AI Governance Thought Leader
- Ahmed Syed – Business Transformation Specialist
- Paul Birkin – CTO at Homely
The theme was simple:
Where and how AI fails — and what we learn from it.
What surprised me wasn’t the failure stories.
It was how different the delivery approaches were across organisations.
In a space still lacking recognised best practice, we’re seeing a wide spectrum of methods for:
- Turning AI opportunity into initiative
- Delivering capability
- Industrialising it
- Making it scalable and sustainable
And it reminded me of something important.
The Continuum Most Organisations Don’t Acknowledge
There is a delivery continuum in AI.
At one end: 🔥 The AI Firestorm
- Minimal prescription
- Bottom-up experimentation
- Rapid iteration
- Self-organising teams
- Strategy evolving through learning
- Maximum development velocity
It’s not new. It echoes the Agile Manifesto and the early 2000s software movement — when companies like Thoughtworks championed self-assembling teams over rigid project governance.
And it works.
Start-ups and digitally native firms can unlock extraordinary energy this way. Homely is a strong example of how relatively unbound AI development can drive real success.
But now move to the other end.
⚛️ Building an Virtual AI Doctor
- Formal governance at the deliver level
- Ethics , Compliance Governance
- Quality Assurance Governance
- Professional Medical Oversight
- AI Data Readiness Platforms – data lakes
- Strong Inference Frameworks for model development, deployment and maintenance.
- Tight detailed blueprints
- Data Quality, Lineage and control
- Compliance and ethics frameworks
- Risk discipline
- Budget governance
- Control gates
- Auditability
- Transparency
Above is a subset of a bigger list to make sure AI does not cause harm and is better than human clinicians. However the other could be characterised by big pharma, insurance carriers, public health services to name a few.
This is continuum should always be in your mind where AI failure failure has systemic consequences and where delivery success in terms of software delivery and its operationalisation needs top be optimised .
Very different environments. Same continuum.
Here’s the Problem
Too many organisation pretend this continuum doesn’t exist.
They copy a model that worked somewhere else.
And they place the pointer in the wrong position.
There Is No Universal AI Delivery Framework
This was the one point everyone agreed on last night:
There will never be a one-size-fits-all AI method.
Where the pointer sits depends entirely on context.
In my experience, it moves toward the left of the continuum when you have:
- Large multi-country footprints
- Deeply layered (often undocumented) processes
- Heavy regulatory oversight
- High decision criticality
- Significant adoption resistance from staff or customers
In those environments, “let the firestorm run” is not a strategy.
It’s a headline waiting to happen.
But equally:
Over-governing early experimentation suffocates innovation before value is even discovered.
Why So Many AI Initiatives Are Failing
There are countless studies claiming AI failure rates are high.
Some say the numbers are exaggerated. Some say success is simply about unleashing bottom-up experimentation.
My view?
The failure rate has less to do with model performance and more to do with judgement.
- Too much control where speed is required.
- Too little control where systemic coherence is essential.
It is not a technology failure.
It is a leadership judgement failure.
Experience Still Matters
When I assess an AI initiative, I instinctively ask:
- Is this a productivity-led innovation opportunity?
- Or does this require structured governance and delivery coherence?
That judgement is contextual. It’s nuanced. It’s based on experience.
And for now, that’s something AI itself won’t replace or I am out of a job.
The Real Discipline
AI delivery is not about chaos versus control.
It’s about:
Understanding where your organisation sits on the continuum. Deliberately placing the pointer. And adjusting it as reality unfolds and this stretches from technical delivery to data quality management and the often all important change management aspects of AI delivery and embedding and operationalising in large organisations
Get it wrong and initiatives stall, fragment, or collapse.
Get it right and AI becomes a compounding strategic advantage.
That, in my view, is where real leadership lies.
Thanks Virturi for letting me attend
2 Comments
Neural Foundry
The continuum framing here is genuinely one of the clearest ways I have seen AI delivery complexity articulated. You are so right that the biggest failure mode is organizations copying a model that worked elsewhere without asking why it worked in that specific context. The distinction between startups running a firestorm approach vs. heavily regulated sectors needing structured governance is critical and often overlooked. The leadership judgement point at the end realy ties it all together.
Neural Foundry
The observation about how AI failure is less about the technology and more about the wildly varied delivery approaches is really insightful. When there is no recognized best practice yet, organizations are essentially improvising, and that creates such different paths to both success and failure. The framing around industrializing AI and making it scalable and sustainable is exactly the right framing, and one that many orgs are still figuring out. Great teaser for what must be a fasciating event.