Not every workflow step needs AI — selective judgment is the key
If you have a 10-step workflow, it might be that the entire workflow should not be changed by AI — it could be too risky, not high enough ROI, or already sufficiently automated. Even within a workflow worth touching, it's often only three of those ten steps that actually require judgment. Lead categorization in a CRM tool, for example, is non-deterministic enough to warrant an LLM. But the remaining steps can be solved with if-then-else statements and API calls. Knowing which steps need judgment and which don't is a far more complex decision than it sounds.
Go deeper with Greg Isenberg
Ask anything about this, and Greg Isenberg answers from everything they've shared.
More to learn
See everything in Learn
MembersHaving an 'AI center of excellence' is not enough — a radical reset is required
The assumption most companies are operating under is that they're doing some things with AI — they have an AI center of excellence, maybe an AI feature — and that's sufficient. It's not. The wake-up call from investors to their portfolio co
VC memo parable: the knife salesman who bet everything on Google AdWords in 2003
A benchmark partner sent a wake-up call memo to portfolio companies urging them to radically rethink how they operate with AI — immediately. The memo ended with a parable: imagine two friends in roughly 2003, both selling knives door to doo
Why every enterprise CEO is writing $100M AI checks right now
The game theory for large company incumbents is brutally simple: the CEO's choice is either pay a massive check — say $100 million or more — to an AI consulting or transformation firm, risk wasting the money, and maybe get fired over it, or