This website uses cookies

Read our Privacy policy and Terms of use for more information.

Hey, thanks for coming to my ProductTank Oxford talk.

As I mentioned, S&P Global's latest research shows 42% of companies scrapped most AI initiatives in 2025—up from 17% last year. These aren't just statistics. They reveal the core problem: teams chase AI trends rather than address actual user pain points.

Finding the AI Sweet Spot

Successful AI implementations start with: "Where do poorly addressed human needs meet new machine capabilities?" As I shared, the sweet spot isn't where AI is most advanced, but where it's most reliable. Reliability comes from strong patterns solving real problems people care about. Focus on suggestions and recommendations over answers and judgment.

A presentation slide titled 'Identifying Your AI Opportunity Sweet Spots' featuring a three-circle Venn diagram. The circles represent: 'Real Human Needs' (pink, top), 'Recommendation Over Answers' (blue, bottom left), and 'Well Defined & Documented Problems' (green, bottom right). The intersection of all three circles is labeled 'AI Sweet Spot' in the centre. Key guidance points on the left state: 'Start with underserved needs—the problem, not the tech', 'Map intersecting user pain points and available AI capabilities', and 'Ask: Where does machine ability support human judgement?' The slide is branded as ProductTank Oxford, September 2025, with attribution to Saielle DaSilva at saielle.beehiiv.com."

Key Takeaways

  • Leverage whatever UX (not UI) skills you can get your hands on to help

  • Map your customer journeys to understand real human issues

  • Use lightweight scoring for impact, feasibility, and repeatability

  • Build cross-functional squads with clear ownership for data quality, feedback loops, and model updates

  • Decouple models from core capabilities through abstraction layers

  • Start with underserved needs, not technology capabilities

Slide titled "Impact First AI Adoption" with four main sections: 1. "Keep AI Adoption Near Business Goals" - Advises setting team AI adoption goals no more than one step removed from company goals like increasing revenue, decreasing costs, or increasing margin. 2. "The Low Impact Death Spiral" with "Estimate impact early and often" - Recommends evaluating AI value by assessing human impact, costs, and key levers AI can drive. 3. A highlighted quote: "Avoid the 'Low-Impact Death Spiral'—shipping outputs without clear goals. Start with capability gaps: 'What work consumes strategic time that could be better spent elsewhere?'" Attribution to Matt LeMay. 4. "Keep Impact Visible and Focused Throughout" - Emphasises filtering strategy, discovery, problem definition, and scoping through an impact lens. 5. "Avoid over-complication in goals and metrics" - Advocates for simple, transparent goals over complex metric systems.

Worth Your Time: What I’m Reading this Week

  • Measuring Developer Productivity | Pragmatic Engineer
    Kent Beck and Gergely Orosz dismantle McKinsey's approach to measuring dev productivity. Their framework on effort vs. output vs. impact directly applies to AI measurement challenges.

  • AI Governance Predictions for 2025 | Enterprise AI Governance
    Oliver Patel's predictions on agentic AI challenging current frameworks and the rise of "train your own AI" policies. Essential reading for governance planning.

Thanks for making it to my talk. I’d love to know what’s sparking your curiosity this week. Until next time, keep learning fast and building well.

-Saielle

Reply

Avatar

or to participate

Keep Reading

View more
caret-right
© 2026 Intentionaut Ltd..
beehiivPowered by beehiiv