The AI Draft
What deserves to be funded when everything sounds useful?
- Capital allocation
- Prioritization
- Organizational readiness
- Technical capacity
- Opportunity cost
Interactive prototype
Open in a new tab →If the prototype doesn't load here, open it in a new tab.
The situation
Eight departments submit AI proposals in the same planning cycle. The budget is $250K, which is enough for roughly two real efforts and a couple of experiments — not eight.
Three of the proposals quietly depend on the same broken data pipeline. Two of them are the same tool with different names. One of them is a workflow problem wearing an AI costume.
The obvious answer
Fund the proposals with the biggest projected savings, spread a little money across everyone so no team feels ignored, and revisit next quarter.
The question underneath it
This is portfolio construction under uncertainty. You are not picking features — you are deciding which bets can actually be executed by this organization, in this state, with this data, in this window.
What I'm paying attention to
Shared dependencies
Two proposals blocked by the same upstream problem are one proposal.
Readiness, not enthusiasm
The loudest sponsor is rarely the most prepared one.
Overlap
Consolidating three near-identical tools usually beats funding all three.
Reversibility
Cheap-to-unwind bets deserve looser scrutiny than one-way doors.
Capacity
Every funded project consumes engineering attention you already spent.
Opportunity cost
The real cost of a mediocre yes is the excellent thing you didn't fund.
My take
Most AI roadmaps are wishlists with dollar signs attached. The useful exercise isn't ranking ideas by value — it's finding the two or three underlying capabilities that half the list secretly depends on, and funding those first.
"Fix first" is an underrated allocation. So is "avoid."
Related writing
A List of AI Use Cases Is Not an AI Strategy
Substack essay — Coming Soon
Explore another experiment
Build / Buy / Bend
You need the capability. You don't necessarily need to build the software.
Make the Call →AutonomyBlast Radius
How much power should you actually give your AI agent?
Configure the Agent →ReliabilityOne Bad Tuesday
Your AI system worked perfectly for three months. Then Tuesday happened.
Start Tuesday →Office Hours with Chanel
Working through a version of this in your own company?
Private AI office hours for leaders who need to turn messy AI ideas into clear strategy, useful workflows, or safe internal agents — with senior-engineer judgment in the room.