The programmable business: why AI needs more than a chat window

Key takeaways

+

• A programmable business keeps its important work in one system that people and AI agents can both understand, operate, and improve.
• AI has made software cheap to build. The hard part now is running it: deployed, connected to real company data, shared with the team, and governed.
• Adding an AI assistant to every app leaves you with agents stuck in silos. Agents need the same shared context and permissions as the people they work with.
• Growing companies, complex enough to need it and small enough to change quickly, will become programmable first.
• Opser is one programmable system where teams and AI agents build, run, and keep improving how a company works.

Every company already runs on a system. It just isn't one system. It's a handful of inboxes, a few spreadsheets, a project board, a CRM, a dozen chat threads, a shared drive, and the memory of the two or three people who know how things really get done.

That arrangement holds up for longer than it should. The founder remembers the follow-ups that matter. The operations lead knows which spreadsheet to update after every customer call. Someone in finance knows which invoices need a second look before they go out. The real pipeline lives in the head of whoever runs sales.

Then the company grows. New people join, customers multiply, and the exceptions pile up. Coordination gets harder, and not because the work got more complicated. It gets harder because the knowledge of how the work gets done is spread across too many tools and too many people.

This is the problem AI is about to solve, and we think it will change how every growing company operates.

AI made software cheap to build. Now it has to run.#

For the first time, a small team can describe a workflow and get software built for it. An operator can put together a lead tracker, an invoicing flow, a customer dashboard, or an internal tool in an afternoon. Not long ago, each of those meant a development team, a budget, and a few months.

That is a huge shift, and it is only the first half of the story.

A tool generated in a chat window isn't part of the business yet. To be useful, it has to run somewhere real. The rest of the team needs to be able to sign in to it. It has to read and write the company's actual data, not a copy pasted into a prompt. Someone has to be able to change it next month, long after the original prompt is forgotten. And the work it handles has to be visible and governed, because there are real customers and real money on the other end.

The gap between building software and running a business on software is where most of the value sits today. The companies that close it get what used to take an engineering department: a twenty-person team operating with the reach of a much larger one.

What is a programmable business?#

A programmable business keeps its important work in one system that people and AI agents can both understand and operate.

Customers, deals, projects, invoices, conversations, and approvals stop being isolated records in separate tools. They become connected parts of one living model of how the company works. In practice, that means:

• Every process has an owner, a current state, and a next step anyone can see.
• Agents handle the repetitive work with the full picture: the customer's history, the open invoice, the promise made on the last call.
• People step in where judgment matters, and approve anything that shouldn't go out unreviewed.
• Permissions and approvals are part of the system from day one, so nothing an agent does is invisible.
• When the business changes, you describe the change and the software follows.

This is a different bet from adding an AI assistant to every app you already pay for. That path leaves you with a dozen assistants, each stuck inside its own silo, none of them aware of what the others know.

It's also a different bet from handing the company to a model and hoping for the best. Businesses run on judgment and accountability, and they should. What changes is that people and agents finally work from the same context, each doing the part they're best at, with a person answerable for every decision that needs one.

Why growing companies will get there first#

The companies best placed for this shift aren't the giants with the biggest IT budgets. Their processes have hardened over decades, and changing them takes years and a team of consultants.

The first programmable businesses will be growing companies: big enough to have real operational complexity, small enough that one strong operator can still see how every piece fits. Companies where a critical process still depends on one person, one spreadsheet, or one chat thread.

These teams have every reason to move now. Their workflows are worth encoding, and nothing is stopping them. They don't have to bend to software designed for somebody else's company. They can run on software that fits how they work today and keeps up as that changes.

Within a few years, we expect this to be the normal way a growing company runs. The ones that start now will compound the advantage every month.

What we're building at Opser#

Opser is one programmable system where teams and AI agents build, run, and keep improving the way a company works.

It starts with the apps a company runs on: sales, customers, email, tasks, finance, and more, all on one shared record. Each app comes with an agent that already knows your customers, your work, and your numbers. When an app is missing, you describe it and it's built into the same workspace, on the same data. Every agent works inside the permissions, approvals, and audit trail you set.

None of these pieces is new on its own. What's new is having them in one place. The agent that drafts an invoice can see the deal it came from. The person who approves it can see why it was drafted. And when a process should work differently, the fix goes into the system, where every person and every agent picks it up.

Where business software goes from here#

For thirty years, business software meant buying tools built for the average company and bending your operations to fit them. That era is ending.

The next generation of companies will describe how they work, put that knowledge into their systems, and improve those systems every week. AI made building software possible for everyone. Programmable operations will make what gets built useful to everyone, and the companies that get there first will be very hard to catch.

Frequently asked questions

What is a programmable business?

A company whose important work (customers, deals, projects, invoices, approvals) lives in one connected system that both people and AI agents can read and operate, and that the team can change by describing what it needs.

How is that different from adding AI to the tools we already use?

An assistant inside each tool only sees that tool's data. In a programmable business every agent works from the same shared context, so the agent drafting an invoice can see the deal it came from.

Does a programmable business replace people?

No. Agents take the repetitive work and people keep the judgment. Anything that shouldn't go out unreviewed waits for a person's approval, and every action is logged.

Who is Opser for?

Growing companies with real operational complexity that can still change how they work: startups finding their first customers, and teams whose processes live in spreadsheets, inboxes, and chat threads today.

Simone Zamparini

Founder · Senior Software Engineer

Simone Zamparini is a founder and senior software engineer.