AI is in the Apple II Era

Picture the scene.

It’s 1979. One person seated, two others looking over the shoulder. They’re all crowded around an Apple II, their faces lit by glowing green characters on the screen.

One of them is pointing excitedly at a spreadsheet.

“See this? It’s called VisiCalc. I can change the interest rate right here—and it updates all the rest of the numbers on the page. Can you believe that?”

A pause, and then the statement: “I BUILT THAT.”

Of course, that person didn’t build VisiCalc. They built a spreadsheet inside it. But for the first time, the average intelligent human being could make a computer do something genuinely useful.

That is exactly where we are with AI today.

Two reactions to an unpredictable future

I see business leaders splitting into two substantial camps.

The first camp watches someone build an application with AI over a weekend and reaches a breathtaking conclusion: We can build all our own software now. Why keep paying dozens of vendors? We’ll just describe what we need and build it ourselves.

The second camp admits it has no idea what the future holds—and by “the future,” it means six months from now. So it buys Claude for everyone, encourages experimentation, and waits for the fog to clear.

Both reactions are understandable. Neither one is a strategy.

The truth is that AI is going to destroy some categories of software. It is also going to make the best software dramatically more valuable.

You probably should not build your own software

AI has made it astonishingly easy to build the first 80% of an application.

Unfortunately, the first 80% is the fun part.

The remaining 20% is permissions, security, data governance, backups, edge cases, telemetry, monitoring, integrations, and documentation. It’s dealing with package dependencies that break everything, or APIs get deprecated and the process fails.

The demo took a weekend. The product takes forever.

If you decide to replace a meaningful piece of commercial software with something your team built, you haven’t eliminated a software vendor. You’ve turned yourself into a software vendor. You get to amortize that cost across exactly one customer, and you call yourself when something breaks.

For some companies, that trade can make sense. If the software is central to your competitive advantage, you may absolutely need to own it.

But most leaders who decide to build everything are not escaping a software bill. They are buying themselves a brand-new evenings-and-weekends job maintaining a private software company they never intended to start.

Some software is about to get crushed

That does not mean the current software stack is safe.

Years ago at Riskalyze, an enterprise client required us to maintain a public status page. Someone on the team said “I found a solution.” We bought it, hooked it up to our logs and telemetry, and spent around $3,000 a year. It was perfectly rational: buying was easier than building and maintaining.

As we started to roll out our Partner API at Contio, my teammate Tom told me we needed a status page as well. I confidently said “ah, I’ve got a pretty good vendor for that.”

“No need,” he said. “It’s just a pretty display for our logs. I whipped one up.”

Think about it — there’s no compounding data, no network effect, and no proprietary algorithm that adds value to a status page. For our needs, something that was static and simple worked great.

To put it simply, software that is mostly a pretty display for your own data is in trouble.

AI is collapsing the cost of narrow applications without moats. Dashboards, web sites, calculators, and basic reporting tools will increasingly be built on demand. Products that justified a few thousand dollars per year because building was annoying will struggle when building becomes a Tuesday-afternoon project.

But the opposite is also true.

The best software will become more valuable

Great software is not just a collection of screens and features. Its value compounds beneath the interface.

• Compounding data. The product improves as it accumulates customer history, decisions, relationships, and outcomes.

• Network effects. Every additional customer, participant, or connection makes the product more valuable to everyone else.

• Proprietary algorithms. Hard-earned logic produces an outcome that a clever prompt and a weekend cannot reproduce.

AI does not erase these advantages. It amplifies them.

Think about my former company, Nitrogen (which started life as Riskalyze). They synthesize data feeds from thousands of sources — markets, custodians, asset managers — and they crunch all of that through the lens of the Risk Number® — and that Risk Number becomes more of an industry standard every day. Compounding data, network effects, proprietary algorithms. Real value.

A generic AI model can write code for a new interface. It cannot instantly recreate ten years of clean data, a trusted network of participants, or an algorithm refined across millions of real-world outcomes.

So no, the future is not that every company builds every piece of its own software. Most software will remain hard, expensive to maintain, and extremely valuable.

But the boundary between the products in your stack is about to change completely.

Build your own glue

This is the part almost every company should build.

Your glue is the custom layer that makes the systems you buy work together the way your business works. It connects your data to your unique processes and bridges gaps vendors cannot design for because those gaps are particular to you.

Your glue might:

• Turn a client conversation into updated relationship history, assigned work, and a drafted follow-up without anyone moving data by hand.

• Combine information from your billing system, project system, and CRM into one view that answers the question your team asks every Monday morning.

• Integrate two different products who have an integration, but that integration doesn’t quite do exactly what you want it to do.

This used to require an internal engineering team, a consulting project, or a fragile mess of no-code automations. AI has made it feasible for far more companies to durably build this. The project is contained, the business logic is yours, and if something changes, you are repairing a connection—not maintaining a substitute for Salesforce.

Buy the systems with compounding value. Build the glue that makes them yours.

When you get this right, the result feels like magic to your employees and clients. Information appears where it is needed. Rote work evaporates from people’s to-do lists. Your people stop serving the software, and the software finally starts serving your people.

Fire any software that traps your data

There is one requirement for this future: every important product in your stack must give you full-fidelity access to your own data.

Not a monthly export that loses half the context. Not an “open API” that exposes twelve safe fields while trapping the valuable information inside. You should be able to retrieve everything your software knows on your behalf and use it elsewhere.

I’ve come to believe this simple truth: if a piece of your tech stack won’t give you full-fidelity access to your own data, you must aggressively fire them.

A product that traps your data cannot participate in your glue. It forces employees to keep visiting its little island, copying and pasting information into the rest of the business. No matter how beautiful that island is, it will increasingly hold you back.

This is one reason we built a Partner API at Contio early on, and why we’ve now delivered full user and workspace APIs with complete MCP access. With these four capabilities, you can build an app that spans across multiple Contio workspaces; natively pulls data across your entire workspace; connects your personal Contio account into any other system; or allows you to use your AI tool of choice to ask questions of Contio data.

We didn’t build those things because APIs are fashionable. We built them because I think in less than 12 months, any software company not delivering them will be hopelessly behind and incapable of catching up.

Bottom Line

We eventually got to the other end of the Apple II era. We moved on from thinking we all had time to be hobbyist builders, we leveraged the work of others, and we ended up far more productive than ever before.

AI will do the same. It will make building feel effortless, kill products whose only advantage was that building used to be annoying, and make software with real compounding advantages even more valuable.

Your job is not to predict every category winner six months from now. And it is not to turn your company into a mediocre software vendor for itself.

Your job is to draw the boundary in the right spot.

Buy the software that compounds. Build the glue that makes it yours. And fire anything that gets in the way.