Enterprise Software's Three Quiet Reckonings

About a month ago, on the first week of a serious run on a personal AI project, my laptop started dying. Wouldn't hold a charge. Battery draining fast. Old machine, reaching the end.

Bad timing. The project was cooking. Momentum like a freight train that had finally gotten up to speed. Losing it would have been rough.

My wife's laptop was available. She rarely uses it, generously offered. So: move everything. My full Claude ecosystem, the Python environments, multiple web servers, the Claude Code setup, dozens of libraries, permissions, browser profiles. All of it, to a new machine, before the old one gave up.

Claude and I sat down and had a serious conversation about it. Took a deep look at my projects, my workflows, my operating environment. Inventoried the configs. Traced back the decisions I'd made along the way and why I'd made them. Discussed the general approach to transfer and reconstitution. Walked through the risks and how to detect and mitigate them. Finalized the plan. Wrote install scripts. Gathered the files. Zipped the whole thing and dropped it on Google Drive.

On the new laptop, I ran the scripts. Got 70% of the way there on the first pass. Spent about 45 minutes working through the rest. Ninety-five percent good enough to keep rolling.

Total time, planning through rebuild: about an hour.

A week became an evening

If my laptop had died five years ago, before AI, I'd have done it myself — I had enough technical chops. But it would have taken an entirely different form. I'd have been racing the clock on the dying machine, copying what I could, dumping what I could. I'd have gotten the files. I probably wouldn't have gotten the inventory — the "what depends on what, which config did I set six months ago at 11pm, which browser profile holds which token" work. That's patient work. You don't do patient work on a machine that's dying.

On the other side, I'd have started from "I have the files. I think." Every evening would open with something broken. Which Python version? Why is this import failing? Oh right, I never copied that config. Spelunking backward through my own work, one breakage at a time. Fifteen to twenty-five hours of it, spread across a couple of weeks. I'd have ended up at 85%, not 95%, because some of the subtle state would never fully come back.

"Faster" is the thing everyone says about AI. But "faster" is one part of what changed here, not the whole of it. It's speed, yes. It's also the span of the time — a week-long recovery became a single evening. It's the posture — scrambling to fix things after they broke became planning for them before they did. It's the feeling — a week of ambient dread became a calm couple of hours on the couch. It's the quality of what came out the other side — 95% intact, subtle stuff preserved, instead of 85% with quiet gaps I'd never fully close. And it's the nature of the work itself — a tool I used became a collaborator I worked with.

Calling it "faster" picks one piece and skips the rest.

What this has to do with enterprise software

Step back from the laptop and look at what that migration implies at scale.

Big enterprise software platforms get talked about as if they're all one thing. They're not. Inside any mature platform is a mix. There's genuinely valuable stuff — crystallized expertise, hard-won patterns, integrations that took years to get right, institutional knowledge about an industry baked into the code itself. Started 10 or 20 years ago, inter-married with other systems and open source projects, evolved across generations, sharing DNA and wisdom and debt along the way. That part is real.

There's also a lot of other stuff in the same package. Platform debt. Features kept alive for the three customers still using them. Marketing bloat that felt essential during the 2017 roadmap review. Lower-value functionality that earns its keep mostly through bundling. And a wide band of "it depends" — where how much value you get from the platform depends heavily on how well it's implemented at your company, and who's operating it.

So the honest picture isn't "these platforms are indispensable." And it isn't "these platforms are replaceable." It's somewhere in the middle, and the middle is fuzzy. The valuable subset is genuinely hard to replicate. The rest of it is more replaceable than the pricing suggests.

You can't rebuild the valuable subset in an afternoon. But it can be replaced over time. Zoom came out of nowhere and took the video-conferencing crown from Skype and WebEx. Salesforce unseated the on-premise CRM giants when SaaS was still considered risky. Shopify built a commerce platform that ate into territory Oracle and IBM had owned for decades. Nothing is untouchable forever. What's usually true is that replacement has been expensive, slow, and risky — which is exactly the assumption AI is starting to pressure-test.

So two things are true at the same time. There's real institutional value in these platforms, and it would be unwise to disregard it. And the packaging, pricing, and delivery around that value was built for a world that's quickly changing.

Three reckonings

One. Right-shaping the product.

The user-facing UI was the right design for twenty years. Knowledge workers clicking through screens, navigating governance, composing in interfaces built for their hands. Fine.

That's not where the work is going. Increasingly, the work is being done by AI collaborators acting on behalf of those knowledge workers. What those collaborators need isn't a better UI. They need structured API access to the logic itself — the reasoning, the patterns, the rules, the context — delivered in a form they can orchestrate against.

Vendors who reshape their offering around that keep their moat and extend their relevance. Vendors who cling to the UI layer and try to block the scrape will lose. Not because anyone's out to beat them. Because the people paying them are going to find the shortest path to the value — and pretending otherwise is an arms race against their own customers. It's a losing battle, on the wrong side of history.

Two. Right-sizing the pricing.

The pricing a customer pays today reflects decades of layered investment — the valuable subset, plus the platform debt, plus the bundled features, plus the premium a customer accepted when alternatives were thin and migrations were painful.

That stack is getting looked at more carefully now. Customers doing AI-assisted work are building a much sharper picture of which parts of the platform actually produce value for them, which parts are carried along for the ride, and what any of it should reasonably cost. The single blended price tag starts to feel less reasonable when the customer can see the pieces underneath it.

None of this means the platform is worth less. It means the customer can see the value itself more clearly, and the pricing needs to hold up under that clearer view. Vendors who can point to specific, durable, hard-to-replicate value will be fine. Vendors whose pricing leans on bundling and incumbency will have a harder conversation.

Three. Revisiting the migration value prop.

For decades, enterprise vendors have priced their platforms partly on the cost of leaving them. The re-platforming. The re-integration. The retraining. The multi-year transition with real risk of breakage. The pricing didn't just reflect ongoing value — it reflected the size of the exit.

That calculation is shifting. The same kind of AI collaboration that let me rebuild my environment in an evening makes larger migrations more tractable than they were. Not easy. Not fast. Not something anyone should do lightly. But less scary than they used to be, and heading in one direction.

Some portion of what justified the price was the difficulty of leaving, not the value of staying. As that difficulty eases, the pricing needs to earn its keep on what the customer gets going forward — not on how hard it would be to go. That's not a crisis. It's a shift worth getting ahead of.

Three reckonings

The assumptions baked into enterprise software pricing and packaging were built for a world where customers couldn't easily reach in or get out. That world is changing.

The response shouldn't be ripping apart your platform ecosystem. It's a good-faith recalibration — from both sides of the contract. Customers get clearer about what these platforms actually deliver that AI can't. Vendors get clearer about what they can still charge a premium for, and where customers are ready to do more on their own.

The companies that handle that conversation well on both sides are going to do just fine. Maybe better than fine. A laptop migration on a Saturday got me thinking about all of this — but the conversation itself has been a long time coming without us realizing it.