Sometimes we push AI into places where we don’t even need it yet. You also pointed out that starting from the worst-case scenario is actually a great approach. If everything goes really really wrong, what does that look like? The consequences of that wrong execution?
On another note, in the end of the day, it’s usually a lot less AI and a lot more “humans being humans” than any real system architecture situation. But that's a whole different topic, I know. Thanks for putting this together, Joel!!
For sure, and you are completely right. I can tell that this is something you have experience with. Starting with the worst-case scenario is something I do with my human teams all the time before launching a project. The question I ask them is: if this initiative was to fail completely in six months, what are the top two or three reasons you can think of that caused it to fail?
Nihal, that worst-case question works even better once you make it about the agent specifically. Not what could go wrong in general, but which single action this agent is allowed to take on its own, the one where a wrong classification actually ships instead of getting caught.
In the story above it was one routing decision, a misclassified email sent straight to billing. Naming that one action ahead of time is what tells you where the human escalation actually needs to sit, not just that one should exist somewhere.
The "framework is the easy 20%" line is the whole article in one sentence, and it's why so many build-vs-buy debates go nowhere, everyone argues about the 20% because it's the part you can screenshot. The shipment escalation misroute costing three enterprise customers in 48 hours is a great example precisely because it didn't require smarter AI, it required someone to have already decided where billing questions shouldn't go. Out of the five checklist questions, which one do you find leadership teams skip most often before they ship?
Fully agree, @Elaine! I think the last step is the one we so often miss, reviewing and learning from outcomes. That’s why so many things fail over and over and new things are shipped with the same issues. What to you think?
Sometimes we push AI into places where we don’t even need it yet. You also pointed out that starting from the worst-case scenario is actually a great approach. If everything goes really really wrong, what does that look like? The consequences of that wrong execution?
On another note, in the end of the day, it’s usually a lot less AI and a lot more “humans being humans” than any real system architecture situation. But that's a whole different topic, I know. Thanks for putting this together, Joel!!
For sure, and you are completely right. I can tell that this is something you have experience with. Starting with the worst-case scenario is something I do with my human teams all the time before launching a project. The question I ask them is: if this initiative was to fail completely in six months, what are the top two or three reasons you can think of that caused it to fail?
Nihal, that worst-case question works even better once you make it about the agent specifically. Not what could go wrong in general, but which single action this agent is allowed to take on its own, the one where a wrong classification actually ships instead of getting caught.
In the story above it was one routing decision, a misclassified email sent straight to billing. Naming that one action ahead of time is what tells you where the human escalation actually needs to sit, not just that one should exist somewhere.
The "framework is the easy 20%" line is the whole article in one sentence, and it's why so many build-vs-buy debates go nowhere, everyone argues about the 20% because it's the part you can screenshot. The shipment escalation misroute costing three enterprise customers in 48 hours is a great example precisely because it didn't require smarter AI, it required someone to have already decided where billing questions shouldn't go. Out of the five checklist questions, which one do you find leadership teams skip most often before they ship?
Fully agree, @Elaine! I think the last step is the one we so often miss, reviewing and learning from outcomes. That’s why so many things fail over and over and new things are shipped with the same issues. What to you think?