A decision that isn't recorded gets asked again
I kept answering the same question in different AI sessions, and once an AI's guess got filed as a decision I never made.
One of the most annoying failures in my AI workflow had nothing to do with model intelligence.
I’d make a decision. The session understood it, and work continued. Later another session asked me the same question again. My answer existed in a conversation, but it never became durable state.
The opposite happened too. An early AI session suggested a paid service as part of a plan. That suggestion got written down as a decision awaiting my approval and sat on my blocker list for weeks. I had never approved it, and nothing had been spent. It was an AI assumption that looked like a decision because it had been recorded. I rescoped the work to local tools with no new vendor.
So decisions became their own data type. Each one records:
- what was decided
- who decided it, and when
- whether it’s still binding
Open-work views check those records before showing something as unresolved. And my blocker tool now refuses to create a record without evidence that the problem exists.
“Remember this” is too vague for an operational system. Memory helps a conversation. Recorded decisions help a system.
Written from my work building internal systems for a commercial solar contractor. Details are generalized; no company, client, or financial data.