AI promises the answer to a question that purpose-led organisations have been grappling with for years: ‘how do we do more with less?’
But easy access to LLMs has not given way to easy implementation.
Momentum behind promising pilots has stalled, impressive demos have been hard to operationalise and activity has fragmented across the organisation.
Leaders now find themselves stuck: sensing big wins are available but with no obvious route to realising them.
So what does it actually take to get AI working for the sector?
Breaking the AI deadlock
We’ve seen firsthand how the more organisations push on AI, the faster they run into their existing constraints.
Challenges like low data readiness and siloed teams reduce organisational effectiveness generally and limit what’s possible with AI specifically.
This creates a deadlock. Running isolated, low-impact AI experiments won’t generate the momentum or evidence necessary to justify further effort. But pausing on AI to prioritise long-term, readiness-building initiatives isn’t an option either.
To break the deadlock, we see an opportunity in a different AI constraint: today's models are far more capable than they are reliable. We use reliability here to cover consistency, robustness, calibration and safety – the cluster of properties that Kapoor and Narayanan argue are under-weighted in how organisations currently evaluate AI.
Larger commercial entities can absorb the gap between AI capability and reliability. They have the digital foundations, the budgets and the risk appetite to run high-impact experiments at scale and learn from inconsistent results. MIT NANDA’s State of AI in Business Report estimated enterprise investment into generative AI at $30 to 40 billion in 2025. Yet 95% of organisations saw no measurable financial impact.
Purpose-led organisations don’t have the same luxury. But by reframing the AI capability-reliability gap as an opportunity, they can make the case for digital transformation more broadly: capturing what’s reliable with AI now to build organisational confidence, while investing in the data foundations to unlock more transformative potential later.
What we learnt developing an AI roadmap with Blood Cancer UK
We recently piloted a new, ten-week AI strategy sprint programme with Blood Cancer UK.
Blood Cancer UK had real ambition, a strong innovation culture, and a clear sense that AI could be significant for their mission. What they needed was a strategic approach to testing what works, building capability and confidence and scaling with intent.
Here are three lessons from our experience:
-
Why, not what: start with your mission
Introducing AI for its own sake typically leads to stalled pilots, demos that do not work in practice and fragmented activity.
At Blood Cancer UK, we started with fundamental questions about the organisation:
-
Where does your mission stall?
-
Where do your people spend time on work that does not require their expertise?
-
Where do the people you serve experience friction, delay or gaps in support?
The themes in the answers to those questions became the map for understanding where AI might be a strong candidate to add value, alongside other technical solutions.
This approach shifts the AI conversation to outcomes. By framing AI in terms of what it enables – for example, more people reached, faster response times, deeper insight into user needs – it’s easier to build the internal case for experimentation and investment.
-
Make space for conversation and hold competing perspectives
For organisations held to high standards of social and environmental accountability, you need to confront ethical concerns and trade-offs early and often.
By using structured activities to surface and discuss competing perspectives, we bring everyone into the conversation. With Blood Cancer UK, we saw an opportunity for staff more sceptical about AI to help pressure-test decision making; helping to identify a wider range of consequences and mitigations. This approach may take longer, but it makes the thinking more thorough and experimentation safer. This matters especially in a sector where trust is everything.
-
Identify experiments you can run in-house
Rather than pursuing transformative AI solutions that aren’t yet reliable enough, we helped Blood Cancer UK identify experiments that they could run internally.
Through the collaborative process, we explored where value might be realised, what foundations more sophisticated experiments will need and what evidence would tell us when to move faster. A setup guide for each experiment also helped assess the level of in-house readiness. This bridges the gap between strategy and actual activity, and demonstrates value to the organisation within months rather than years.
Find the full Blood Cancer UK AI Wayfinder case study here
This thinking has informed AI Wayfinder: Manifesto’s new, ten-week programme that delivers an AI strategy and roadmap. Read more about AI Wayfinder.
Getting AI adoption right is a long-term effort and a good start matters. Organisations that get it right will steadily build skills and confidence, until AI becomes just another part of how they deliver the mission. True to its name, AI Wayfinder helps you identify where you’re going with AI and how to get there.
We'd welcome a conversation if any of the challenges here sound familiar. Thank you to Blood Cancer UK for being ambitious partners and helping us refine our thinking.