Every week I talk to a founder who’s under pressure — usually self-imposed — to “do something with AI.” They’ve read the articles. They’ve watched competitors announce new capabilities. They feel like they’re falling behind.
So they buy a tool. Or they tell their team to “start using AI wherever it makes sense.” Or they hire a consultant who delivers a 40-page AI strategy that sits in a folder and gets implemented almost nowhere.
None of that is wrong exactly. But it’s all starting in the wrong place.
The question you should be asking first
Before you evaluate tools, before you write policy, before you hire anyone with “AI” in their title — you need to answer one question: where is your team’s time actually going?
Not where you think it’s going. Not where it should be going. Where it actually goes, empirically, over the course of a normal week.
The reason this matters is that AI is not a strategy. It’s leverage. And leverage applied to the wrong place — to activities that shouldn’t exist at all, or that don’t compound, or that don’t touch your real constraints — produces a faster version of mediocrity.
The companies getting the most out of AI right now aren’t the ones who adopted it earliest. They’re the ones who adopted it most deliberately — who started with a clear view of their operational bottlenecks and then asked where AI could move the needle on those specific constraints.
A simple time audit
Here’s how I approach this with the founder-led companies I work with. Spend two weeks tracking where your team’s collective time actually goes. Not self-reported, idealized time — actual time, traced backward from outputs.
Group activities into four buckets:
- High-value, hard to replicate. Deep thinking, relationship work, judgment calls, creative strategy. This is where humans have durable advantage.
- Repetitive execution. Work that follows a template, rule, or pattern — and produces the same output regardless of who does it.
- Coordination and communication. Internal meetings, status updates, handoffs, context-setting. Often necessary, often overdone.
- Non-value-adding. Formatting, reformatting, finding information, waiting for approvals. Things that exist because of process friction, not because they produce output.
The second and fourth buckets are where AI has the most immediate impact. Not because the work is trivial — it often isn’t — but because the bottleneck isn’t judgment, it’s throughput.
Where I see it working in founder-led companies right now
The applications that consistently produce ROI in the $3M–$15M range:
First-draft generation. Proposals, SOWs, client reports, job postings, campaign briefs. The time isn’t in writing — it’s in editing. AI handles the blank page; humans handle the substance.
Research and synthesis. Competitive intel, market research, prep for client meetings. An hour of reading compressed to ten minutes of insight.
Internal documentation. Meeting notes, process documentation, onboarding materials. The most neglected category in founder-led companies, and one of the highest-leverage investments you can make in operational resilience.
Customer communication at scale. Not replacing relationship communication — augmenting the parts that don’t require a human: follow-ups, summaries, FAQ responses, renewal outreach.
Strategy generation. Senior-level judgment. Anything that requires real context about your specific business, customers, or relationships. AI produces confident-sounding output in these areas — which is exactly what makes it dangerous. The companies that get burned by AI are the ones who let it make decisions it doesn't have the context to make.
The governance question nobody asks
Most AI adoption conversations in small companies are entirely about capability: what can this tool do? The question that gets skipped is governance: who decides how it’s used, where its outputs can be trusted without review, and what happens when it’s wrong?
This isn’t bureaucratic overhead. In a founder-led company with 15 people, one person’s AI-assisted output can propagate across the whole organization before anyone catches an error. Having a clear — even lightweight — set of norms about where AI is in the loop and what oversight looks like is what separates teams that scale with AI from teams that create new problems as fast as they solve old ones.
Start with the time audit. Figure out where AI can do the most good in your specific business. Then build the norms that let you scale it safely. That sequence matters. Most founders do it backwards.
