PRODUCTHEAD: Evals are just another rabbit hole
» Evals are your way to check your AI is behaving
» Writing evals is a task a PM can do, but shouldn’t
» With AI, “done” becomes a calibration about an acceptable variance in output and behaviour
PRODUCTHEAD is my free curated newsletter of the best articles, videos and podcasts from product leaders and commentators all over the world. All neatly packaged up in a weekly email delivery for your reading, viewing and listening pleasure. Curated by Jock Busuttil (jockbusuttil.com).
» Evals are your way to check your AI is behaving
» Writing evals is a task a PM can do, but shouldn’t
» With AI, “done” becomes a calibration about an acceptable variance in output and behaviour
» ‘Good’ product practice is boring – good and useful progress without drama
» So you can build your own digital services. What’s next?
» Corporate values can be more revealing than might appear
» There’s a tension between what can do and what you should be doing
» We have 4 behavioural defaults: emotion, ego, social and inertia
» AI means product people should focus more on discovery and go-to-market
» AI opens the door for product people to become product builders
(A range of views this week on PRODUCTHEAD)
» The SaaSpocalypse won’t kill off vendors, but they will have to change their pricing strategy
» There is an opportunity to exploit higher usage driven by AI agents and increased automation
» Open and sustainable dialogue across roles leads to greater organisation coherence
» An organisation’s business model solidifies over time
» A new venture with a different model should ideally be pursued separately to the main org
» Understand the needs of your audience to be more influential
» Go-to-market execs won’t hear you unless you speak in currency symbols
» Your manager can’t advocate for you if you don’t brief them
» When a measure becomes a target, it ceases to be a good measure
» Product, market, model and channel must all fit together to be successful
» In complex human systems the constraint is often policy, mindset, or coordination, not a single bottleneck
» Being the “human in the loop” means having genuine authority, time to think, and understanding the bigger picture well enough to question the system
» AI is forcing us to adapt how we work in the face of greater information flow
» However, this doesn’t mean we abandon the fundamental guiding principles of our discipline