In early 2025 the people around me at Social Habit Marketing started pushing us to build our own software. We ran the company on subscriptions. Notion held our documents and projects. Zoho ran most of the rest, and Zoho is a software conglomerate in its own right, with a feature for nearly everything a business does. We used a thin slice of each and paid for all of it. The case for building our own was obvious. I agreed with it, and I said not yet.
I knew the Social Habit App was coming. A tool built for our company would do what we needed and nothing else, at a fraction of what we paid the vendors. A software vendor builds for the broadest customer it can find, so every customer pays for features built for someone else. That argument held the whole time I waited. The cost of building is what changed.
Saying not yet to people who are right is harder than saying no to people who are wrong. The people pushing to build saw the same waste I saw, and they were correct about it. A wrong idea loses the argument on its merits. A right idea at the wrong time wins every argument about merit, and the only answer to it is the calendar.
I hold myself to one rule. If I am going to pour resources into something, my strategic intuition has to be close to perfect, and the plan has to be feasible with the people and money we have. The size of the company sets how strict that rule is. A large company can spread a bad bet across many good ones. A small company puts a large share of what it has into each bet, so it gets fewer chances to be wrong. In early 2025 the app passed the first test and failed the second. The direction was right. The tools to get there with a small team did not exist yet.
A wrong-timed bet costs everything. You can build the right product at the wrong time and lose the same way you lose with the wrong product. The money goes, the months go, and nobody remembers that the idea was right.
I watch direction more than volume. The right direction saves effort, and most of the effort I see companies spend goes into making up for a wrong one: the rework, the workarounds, the migration off the system they built too early. A team heading the right way can move at half the pace and still arrive first.
So I watched. Through 2025 I tracked how fast AI development tools improved, and the change was one of kind, not degree. Claude Code and Codex arrived, and development cycles that used to take a team months started to shrink to weeks. Every release moved the same question. The question was not whether AI would write our software. The question was when it could write enough of it for a company our size to ship.
My bar was a better-than-even chance of deploying the app successfully within the year. Even odds sound low for a company that cannot afford a miss, but the bar applied to the whole project: a working app, integrated into how the company runs, inside twelve months. Below that line, most of the effort goes into fighting the tools. Above it, the tools carry part of the load. And the tools improved every few months, so a project that started at even odds would get better tools halfway through. The odds improve while you work.
Then the question changed shape. Claude Cowork appeared, an agent that works inside your files and does the task itself, and I had to ask whether the kind of software we planned still made sense. An app full of menus, forms and dashboards is a bet that people keep clicking. If an agent does the work, does anyone still need the screen? I did not want to build the last generation of software with the first generation of tools that could replace it. I decided the screen still earns its place in two jobs: showing people the state of the business at a glance, and letting people check and approve what the agents did. We built everything behind the screen for agents to read and write as easily as people do.
The signal to start came from inside the company. Another senior executive came to me with problems that were growing bigger than we expected. Waiting cost us something from the first day, but until then that cost stayed smaller than the risk of building. That conversation reversed the order, and by then the tools cleared my bar. We started in April 2026. In September the Social Habit App launched, and we integrated it into the company in place of the subscriptions we built it to replace.
An on-time bet returns more than the tool it produces. The subscriptions were the visible saving. The larger gain is the next stage of the business: a company that runs on software it owns, shaped to the way it works, with room to grow on top of it. A wrong-timed bet can cost a small company everything it has. A right-timed one moves it forward.
Timing is not patience. Patience waits until you feel ready. Timing watches two costs: the cost of building, which the tools kept pushing down, and the cost of waiting, which the business kept pushing up. You move where they cross. In early 2025 building the app cost more than living with our subscriptions. By April 2026 it cost less. The idea did not change in between. The tools changed, and the problems grew.
The right product at the wrong time is a wrong bet. Get the direction right early, then wait until the odds pass even, and bet only then.