The story of accelerated delivery was everywhere at hashtag#SnowflakeSummit this week - we all know that story by now. But from my conversations, figuring out what to build is still not streamlined for most SIs. I say this often, but now I feel even more confident: the bottleneck for projects has shifted. It used to be engineering, and now it’s scoping, requirements, and discovery.
One of the most interesting conversations at Snowflake was with an architect at a large SI. Hier team doesn’t write requirements anymore - instead, a solution owner runs some light discovery, the architect builds a prototype, and the client reacts to the prototype.
At first, I thought: that makes it easy, you can bypass requirements altogether and just react until you get to the shape of the thing you want. That’s easier for humans, more intuitive - it’s how I personally iterate with Claude all the time.
But then, the architect continued: the hardest part of her job used to be coming up with the design. Now, it’s educating clients. Clients take the prototype or design, ask their version of ChatGPT to poke holes in it, and then the architect has to explain why they made the choices they did, given the constraints of the project. It sounded exhausting and fundamentally unproductive.
(And don’t forget the timeline expectations: after they finally align on a prototype, the client says, it’s already mostly done, why is it going to take 8 more weeks to be production-ready? 🙄)
The job to be done hasn’t essentially changed. You’re still going back and forth with a client to understand what they want. It still takes time to extract what they actually need versus what they say they want.
From my conversation at Snowflake, at least, it doesn’t sound like using a prototype is much faster or easier than an old fashioned BRD. Maybe the future is requirements that spin up a prototype or a diagram automatically... is anyone doing this today?