AI adoption isn’t a technology challenge. It’s a mental model challenge.
Integrating LLMs into my work over the last few months has reinforced two critical shifts organisations need to make.
From Micromanagement to Delegation. When working with AI, give broad, contextual instructions on the expected outcome and trust the model to execute. Shift from managing tasks to managing outcomes.
From Process to Outcome — the JTBD shift. My brain naturally defaults to SIPOC: inputs, steps, hand-offs. But forcing this process-first framework onto GenAI is exactly what trips people up.
Consider creating a presentation. The SIPOC mindset: you act as a slide factory manager, deeply embedded in the mechanics of how. The JTBD mindset: you prompt the ultimate outcome — “Make me look prepared and persuasive for tomorrow’s board meeting.” The prompt is a declaration of intent, not a step-by-step manual.
Mapping LLMs onto old, sequential process flows gives incremental gains at best — and massive frustration at worst.
Stop asking: “How do we automate the specific steps in our current workflow?”
Start asking: “What is the core progress our user is trying to make, and what outcome are we hiring this intelligence to deliver?”
SIPOC ensures you execute steps correctly. JTBD ensures you build the right thing. AI excels when we feed it the why and let it figure out the how.