How do I plan for AI in my org, when we've never done it?
A simple mental model for sizing up AI investment
It’s planning season, and AI is top of mind. How do we think about planning for AI if we don’t know what we’re doing with it yet, and haven’t done it before?
There are three broad buckets:
(1) You’re covered by off the shelf (OTS) models (e.g. OpenAI): many operational use cases will be handled well by models. You get no upside or competitive advantage, but avoid falling behind. Rent theirs.
(2) A unique blend is feasible: you can create unique value by adding proprietary data to OTS models. You might be able to get to a real product value in a quarter.
(3) Roll your own: what is available off the shelf doesn’t help you, and you’ll need to build from scratch. This is expensive and slow, but possible upside is massive. Pay for experienced help. Ensure there is major business value before you invest (i.e. do your product discovery due diligence).
Start with bucket one, and advance until you get ROI or it’s not worth continuing. Most interesting examples will fall in bucket two. A handful of people should be able to do a feasibility assessment within one quarter.
(If you liked this, go listen to the short (20m), bonus podcast episode released yesterday specifically about thinking about AI in planning season, that I pulled this from. To go much deeper on how to think about incorporating AI into your product, check out the full conversation.)