resourcing genai initiatives
originally posted to linkedin
don’t resource genAI projects. resource learning.
it used to be that dev resources were the bottleneck. writing detailed product specs was costly but necessary to mitigate even more expensive dev failure. you needed a high degree of confidence the product would achieve its business goals before starting development.
things have changed. mvp cost has plummeted. you can now build working prototypes faster than teams used to design mockups.
but genAI projects are risky. the tech is immature, use cases are still being discovered, and what works for early adopters doesn’t always work for later users.
instead of design → develop, think explore → exploit
in this environment we need to balance:
— validating concepts & use cases — controlling risk
explore phase is about proving out the use case and minimizing conceptual risk.
small, time-boxed teams pilot the concept with real users. if it works, advance. if not, kill it fast, learn from it, and redeploy resources.
exploit phase is about scaling the product and minimizing execution risk.
greater investment is required to make it production-ready and drive adoption.
learning is a first-order priority at both stages. postmortems for both successful and retired projects feed directly back into idea sourcing and prioritization.
operationalizing genAI is more than models and math
it’s designing business systems for learning velocity, decision making, and efficient capital allocation. it’s moving projects from intake to scoped pilots to scaled products, with clear success criteria, stage gates, and separate resource pools at each stage. it’s systematically feeding what you learned back into idea sourcing and prioritization.
many teams fund genAI as if it were a traditional project that is guaranteed to succeed with proper execution.
the best teams know better. they understand that many won’t succeed and design to be successful anyway.
how do you resource genAI projects?
one big bet… or a portfolio of small, time-boxed bets designed to learn and systematically fund the winners?
david crowe - reducibl.com - gatewaystack.com