MO | Edition 66... Every Build Is a Lease on Your Future Attention

Intro / Hook

Starting something new with AI feels free.

You get an idea on a Tuesday night, open Claude or ChatGPT, and by midnight there's a working tool on your screen. No budget meeting. No developer. No invoice.

The owning is where it costs you, and that part never shows up on a bill.

Everything you start is a lease on your future attention. Experiments are short leases you can walk away from. Once one proves out, you've signed a long one, and the building is the down payment.

Two stories keep coming up when founders talk about their AI builds. Both are composites, and neither names anyone. They end the same way.

That's why this matters right now.

Why This Matters Now

Building got cheap. Owning didn't.

S&P Global Market Intelligence found that 42% of companies abandoned most of their AI initiatives before they reached production. The year before, that number was 17% (S&P Global Market Intelligence, 2025). The tools got better and the abandon rate went up. So the build was never the hard part.

Story one: the tool in the founder's own account.

An owner gets excited about a new AI tool. They build a project in their own environment, loaded with their own context, and it works beautifully for them.

The team can't get in.

So the fixes start. Someone exports the outputs and emails them around. Someone pastes the instructions into a second account. A third person gets a seat, but it can't see the files. Each fix opens a new limit, and each limit gets its own fix.

Months later the company has a "tool" that works poorly, took a huge amount of effort, and caps how far anyone can take it. Systems start as a promise and end up as a patch. Nobody decided to sign that lease. It piled up one workaround at a time.

Story two: the frenzy build.

A different founder gets a radical idea on a Sunday and drops everything. For days, nothing else matters. They build an AI tool at full speed and it's impressive.

Then reality comes back. A client fire. A hiring question. A team member who needs ten minutes that turns into a day. Over the next few weeks they get pulled in every direction.

The tool gets no new improvements and no maintenance. Questions about it go unanswered. People stop opening it, and eventually it gets dropped.

The part that stings is that nobody ever found out whether it was worth anything. The value was never seen, because nobody stayed long enough to measure it.

The pattern shows up in governance data too. Retool's 2026 AI Governance Report, covered by Cybersecurity Insiders in July 2026, surveyed tech leaders at companies with 50 to 1,000+ employees. Only 8% rated their governance of AI-built internal tools as strong, and 44% had no clear default owner when one of those tools caused an incident (Retool 2026 AI Governance Report via Cybersecurity Insiders, Jul 14, 2026). Those companies are bigger than yours. I'd bet the pattern scales down.

The Systems Cycle

Think about renting space for a business. A month-to-month rental lets you try a location. A five-year lease means you inspect the building, check the plumbing, and know who fixes the roof. Nobody signs the five-year lease on a hunch.

There's a third founder, and this one works on the business instead of in it. They spot an area where an AI-built tool could carry real weight. Before touching anything, they spend time up front on architecture: what resources it needs, how it affects the whole company, and whether they can commit to it long term. They start slower and then they keep getting faster.

That founder runs a cycle. I call it The Systems Cycle, and it has six moves:

Spot → Cost → Experiment → Prove → Own → Improve

🔹 Spot. Name the problem and what it costs today. Hours, errors, delays, dollars. If you can't say it in two sentences, you're not ready to build.

🔹 Cost. Before you build, write down the upkeep: who owns it, hours per week, who else touches it, and what it replaces. ISBSG's 2024 paper says maintenance and support run 65% to 85% of a software application's total cost of ownership (ISBSG, 2024). That covers software in general, not AI alone, but it tells you where the bill usually lands.

🔹 Experiment. Build a small, time-boxed MVP in a shared environment from day one, never a personal account. Experiments are short leases: cheap on purpose, with an end date on the calendar. Killing one that doesn't prove out is a win. You paid a little to learn a lot.

⚠️ Prove. This is the gate. You decide: kill it, keep it as an experiment, or commit. Centercode (July 2026) describes a simple version of this decision: one week, three real users, one usage number, then invest, hand off, or kill (Centercode, Jul 15, 2026). Long-term thinking becomes required here, because this is where you sign the long lease.

🔹 Own. Name a champion. Document it so someone else can run it. Set up a feedback loop on what works and what doesn't, so the tool gets better from real use.

🔹 Improve. Review it on a cadence so it compounds instead of decaying. Over time it becomes a core part of how the business runs, and its value lifts the whole company.

🧠 How to tell a build that will bury you from one worth committing to.

A build that will bury you has no named owner, lives in a personal account, has no end date, and only its builder uses it. Nobody has written down the weekly upkeep.

A build worth committing to solves a problem you can put a cost on. People other than the builder used it without being reminded. One person owns it with real hours attached, and it would survive that person taking a week off.

Run the two stories through the cycle and you can see where each one broke. Story one skipped the shared environment in Experiment, so the patches began on day one. Story two skipped Cost and Prove, so a long lease got signed by default and then abandoned.

A fair objection: fast solo shippers win, and they do. The reason matters. They're the only owner. There's no team to lock out, no handoff, and nobody to confuse. If that's your situation, ship away.

The cycle asks for experiments that stay cheap and one deliberate decision before the lease gets long. Speed stays.

How to Apply This (This Week)

1. This week, list every AI build in your company today, including the ones in personal accounts. Next to each, write one word: experiment, committed, or orphan.

2. This week, pick one orphan or one new idea and write the Cost note: owner, hours per week, who else touches it, and what it replaces.

3. This week, move anything living in a personal account into a shared environment, or put an end date on it.

4. This week, set a Prove date for every experiment. Kill, keep as an experiment, or commit. Put the date on your calendar.

5. This week, for anything you commit to, name the champion and book the first review.

Quick checklist:

Final Thoughts

Everything you start leases your future attention. Keep the leases short until a build earns a long one, then read the terms and sign on purpose. The building was only the down payment.

See you next week,

Damon Flowers

Founder, Modern Operators

This is Issue 66 of Modern Operators. We help founder-led businesses install the operating infrastructure that lets their team run without the founder in the middle of everything.