Agile Yorkshire – May 2026 notes
An event of two halves. I was only there for the first as it's been a long few days and I needed some downtime afterwards. Some notes on the first half.
Intro
Dave West, CEO of Scrum.org, on What happens to Scrum in an AI world? With organisations simultaneously cutting agile investments whilst AI tools are making delivery faster than ever, Dave used his vantage point - constantly talking to CTOs and senior leaders globally - to explore whether Scrum's foundational assumptions still hold, and how the framework might need to evolve.
Context and main argument
- We're living in an increasingly uncertain world where change is measured in hours, not quarters
- Most organisations have responded by freezing agile investments and returning to command & control (the opposite of what they should do)
- AI is exposing that delivery has never been the bottleneck – governance, innovation, and deciding what to build are the real constraints
The speed paradox
- AI dramatically accelerates individual output (code, content, analysis)
- But organisations aren't seeing corresponding increases in value delivery
- Developers are burning out from the sheer volume of AI-generated output that needs reviewing and coordinating
- This is classic local optimisation – we're accelerating the wrong part of the system
Scrum as governance, not delivery
- Dave's key provocation: Scrum might be evolving from a delivery framework into a governance framework
- With AI dramatically accelerating what teams can build, Scrum's ceremonies become governance mechanisms:
- Sprint goals as guardrails to prevent teams doing everything they're now capable of
- Definition of done as quality gates for AI output
- Daily scrums to coordinate the volume of AI-generated work
- Sprint reviews as checkpoints to ensure direction is still correct
- Only 23% of organisations are actively thinking about how to govern AI
- This is where Scrum could provide the most value in an AI-augmented world
Scrum as an infinite versus a finite game
- Most organisations treat agile as a finite game with clear wins and endings
- Scrum should be an infinite game – the goal is to keep playing, not to "win"
- Sprint goals, empiricism, and regular cadence become even more important with AI
How AI challenges agile's foundational assumptions
- Do we still need cross-functional teams if AI can provide all the skills?
- Is working software still the primary measure of progress when AI can generate massive amounts of code?
- Example: Rabobank consolidated from 12-14 teams down to 2-3 per value stream by using AI to reduce coordination overhead
New roles in an AI-augmented world
Scrum masters become:
- AI governance managers (who owns the output?)
- AI collaboration coaches and fluency builders for teams
- System designers for human + AI hybrid teams
Product owners become:
- Context managers (AI needs explicit context that humans take for granted)
- Discovery specialists (AI enables rapid option analysis)
- Relationship builders (still can't be automated)
What organisations should do
- Invest in AI fluency for teams, not just individual tools
- Build shared prompt libraries and context models
- Define clear governance and accountabilities for AI outputs
- Focus relentlessly on outcomes, not output (story points/velocity are now completely meaningless)
What individuals should do
- Develop empirical judgement – your job is to provide wisdom, not just data
- Build team cohesion: people are disconnected and anxious about AI
- Treat your career as a product backlog with a clear product goal
- Keep learning: use AI to help you learn, ironically
- Help evolve Scrum – share what's working and what needs to change
Dave's core message
Technology changes are catalysts for human change. AI will fundamentally reshape work. We can either let a small group of tech lords drive that change.
Or we can use agile principles to harness AI for good: building better teams, delivering more value, and keeping humans at the centre.
Postscript: relevance to my work
Several threads from Dave's talk resonated with my product and user-centred design leadership hats on:
Context as competitive advantage
Dave's point product owners become "context managers" in an AI world seems critical. AI can generate solutions rapidly, but it can't know the nuanced context of NHS services, clinical pathways, user needs, or organisational constraints unless someone actively manages that context. A person's deep understanding of user needs and service design becomes more valuable, not less.
Discovery elevation
Dave argued discovery should become a "first class citizen" in Scrum, with AI enabling rapid option analysis and hypothesis testing. How does this align with user-centred design principles? AI could accelerate the ability to prototype and test multiple solutions – but the judgement about which problems to solve and which user needs to prioritise remains fundamentally human.
The governance bottleneck
Dave made a central argument that delivery is no longer the constraint, governance is. Which sounds familiar in my NHS work. The real bottleneck isn't building the integration or the interface – it's the decision-making across ICBs, trusts, and national stakeholders. AI doesn't fix that. Does it make the governance and prioritisation decisions even more critical?
Infinite game thinking
NHS transformation work shouldn't be a finite project with a clear "done" state (and let's avoid the use of BAU for now too – work is work). Dave's emphasis on treating the work as an infinite game – the goal is to keep learning, adapting, and delivering value – maps directly to how I want to approach ongoing service improvement through research, analysis and then service iteration.
Outcomes over output
I especially liked Dave's dismissal of velocity and story points as "completely meaningless" in an AI world. A hopscotch of logic takes me to this reinforcing core product and UCD principles: what matters is the value delivered to users and the organisation, not the volume of features shipped.