The most expensive thing in a startup isn't payroll or infrastructure. It's decisions sitting in limbo while the founder tries to find perfect information that doesn't exist.
Every company has a list of these. The pricing page everyone knows needs to change. The hire that's been "almost approved" for six weeks. The product direction that's been deferred so many times it's become a running joke in stand-up.
Open decisions don't just sit there. They bleed — slowing adjacent calls, creating uncertainty the team fills with its own assumptions, burning energy at every touchpoint where the unresolved question resurfaces. The cost is invisible and it compounds.
The fix isn't a better decision-making framework. It's understanding why decisions stall.
#Most decisions are reversible. Founders treat them like they aren't.
Jeff Bezos's two-way door framework is the most useful mental model I've encountered for this. Most decisions are two-way doors — they can be reversed, adjusted, course-corrected if they turn out wrong. But founders treat them like permanent commitments. They wait for more data. They loop in one more person. They sleep on it again.
For reversible decisions, the cost of being wrong is low. The cost of being slow compounds every week the decision sits open.
The decisions that deserve slow, careful treatment — irreversible ones, like a co-founder equity split or a database architecture you'll live with for years — are usually easy to identify once you've trained yourself to ask the reversibility question first. Most aren't in that category.
#Decision fatigue is real and most founders ignore it
Every decision depletes something. By 4pm, after a day of calls, context-switching, and a hundred micro-choices — which Slack message gets a real response, how to word feedback to a contractor, whether to push back on a vendor quote — most founders are making noticeably worse calls than they were at 9am.
Not because they got dumber. Because they made 200 decisions already.
The boring fix works: protect your mornings. One 90-minute block before any meetings, pointed at the single most important decision or hardest piece of thinking you have that week. No meetings before 11am is the rule most high-output founders land on independently — not because they read the same book, but because it's the version that survives contact with a real calendar.
Batch the trivial decisions too. Recurring expenses, meeting formats, stock email responses — turn them into defaults. Every choice you automate into a system is one that stops consuming capacity.
#The one-question shortcut
When a decision feels hard, it's almost always because of one specific missing piece of information — not all information, but one thing.
Ask: "What's the single question whose answer would change my choice?"
Find that answer. Not every answer. That one.
This cuts the research loop in half. Founders get stuck doing broad information-gathering when they actually need a narrow answer to a specific uncertainty. Name the uncertainty first. Then go find out.
#Ownership beats consensus
In teams, decisions stall when nobody is actually responsible for making them. Everyone has input. Nobody commits. The question floats in meeting agendas for three weeks because asking again feels like escalation.
For every significant decision, name one person who owns it. Not the person responsible for gathering input — the person who makes the call once input is gathered. Others advise. One person decides.
This is what Amazon's "disagree and commit" is actually about. You can argue hard for your position and then fully execute a decision that went the other way. The alternative — half-committing while privately undermining — is one of the most corrosive patterns in early-stage companies. Leaders set the tone on whether it's acceptable.
#Two mental models worth ten minutes each
10/10/10: For any significant decision, ask how you'll feel about it in 10 minutes, 10 months, and 10 years. This cuts through two failure modes: short-term anxiety driving bad long-term choices, and sunk cost fallacy keeping you in situations that are clearly wrong when you zoom out.
The pre-mortem: Before a major decision, spend 20 minutes imagining it's 12 months later and the decision was a disaster. What went wrong? This surfaces the failure modes you're not letting yourself see because you're already committed to the outcome. It's more useful than a post-mortem because you can still act on what you find.
#Where AI actually helps with this
Two uses that have changed how I work through hard calls:
Stress-testing reasoning: paste your logic into Claude and ask it to find the flaws. It's less reluctant to challenge you than an advisor who's trying to be encouraging. Ask specifically for the steelman of the opposing view — "argue against this decision as hard as you can."
Synthesis: "Here are four data points about this market. What are the two or three most important implications for a pre-seed founder trying to price a new product?" Getting a clean synthesis in 30 seconds often doesn't make the decision — but it frames it in a way that makes the decision obvious.
AI doesn't decide for you. But it can do the legwork that was previously an excuse to delay.
Every week, find the single decision that's been open longest and force it to closure.
Not because the timing is perfect. Because the cost of leaving it open is almost certainly higher than the cost of being wrong.
Move the stuck things. Speed compounds — but only if you're actually making calls.