Every software company has a legacy story. Here’s how I’m keeping Pivot from writing ours.

Getting old is inevitable, even for procurement software companies. We used to think legacy meant Coupa, Zip, and friends. I think AI has made that definition obsolete. If a procurement platform behaves legacy, why wait twenty years to call it legacy?

A self-proclaimed modern procurement company can pioneer the “front door,” have a sleek interface, and an army of agents and superagents, and still behave like the software it set out to replace.

AI slop thrown on top of messy architecture? Legacy. AI built around workflows it should be making unnecessary? Legacy. Agents that require extra headcount to check and correct their work? Legacy.

That’s the legacy I’m interested in. Not how old the software looks, but what sits underneath it and how quickly it adapts when expectations change. And by that definition, Pivot isn’t exempt.

Modern is getting harder to buy

Everything looks modern now. The interfaces are polished. The roadmaps say AI. Everyone has agents. Great. I’ve already written about how yesterday’s disruptors start to resemble the legacy software they replaced. 

Here’s what I’d look at when buying it:

1. Legacy makes the license the cheap part: Implementation, integrations, governance, admins, change management and now a menu of AI capabilities on top. If AI means another module, another implementation, and another thing to govern, put those on the bill too.
2. Superagents are telling us something: If agents need superagents to coordinate them, another layer to govern them and people to check their work, what actually disappeared? I want fewer workflows, approvals and manual interventions, not a more autonomous agent org chart.
3. “Agentic” is becoming a very generous definition: Every procurement vendor has AI now. Congratulations to all of us. But AI won’t clean supplier data, create adoption or recover context the system never captured. A good demo can make old complexity look intelligent. Production tends to find it again.

I’d ask a simple question: what has AI actually made unnecessary? Workflows? Administration? Manual work? Software to manage the software? If most of it is still there, I’d be careful about calling the new layer a new foundation.

Pivot can become legacy too

There is a consequence to everything I’ve just written: Pivot can become legacy too. Probably faster than we think if we buy into the AI hype and start building just to build.

We were lucky with timing. We started Pivot when AI was already changing what software could do, so we didn’t have twenty years of technical debt and have to work out where to force-fit AI. We could build with AI from the beginning and invest heavily in the system of record and system of engagement underneath it.

AI is only as reliable as the foundation it has to work with. Clean, structured data. Context. Adoption. Without them, better models can produce more convincing artificial confidence. Add another agent. Make one a superagent if you’re feeling ambitious. It still inherits whatever sits underneath it.

AI-native has a shelf life

“Born AI-native” can become its own piece of legacy marketing surprisingly quickly if we spend the next ten years protecting what we built in the first three.

That’s the part I think about as a founder. As Pivot grows, we’ll have more customers depending on existing functionality, more revenue attached to old decisions and more perfectly reasonable reasons not to change things.

Eventually someone suggests adding another layer because touching the foundation is harder. Then another. The product gets broader. The packaging gets more creative. Everyone can explain why each decision made sense.

Nobody puts “become the incumbent we used to make fun of” on the roadmap. Legacy arrives one perfectly reasonable decision at a time.

How I’m keeping Pivot from becoming legacy

Being born AI-native gave us a natural head start. But I don’t want us confusing that with a permanent status. As Pivot grows, there are three things I want us to keep doing:

• Obsolete ourselves first: If new technology gives us a better way to solve a problem, we should be the first to rip out what we built, ideally before a competitor or a customer has to point it out.
• Measure progress by what procurement teams no longer have to do: More agents and features make excellent release notes. But I’d rather eliminate a workflow, a manual step, or another piece of administration than add another agent to manage it.
• Move fast, but only if we don’t have to redo it: Speed that creates rework isn't speed. It's technical debt with a launch date. Legacy vendors made "we'll fix it in phase two" a business model. We want speed and excellence, not speed or excellence.

I’m not going to claim that any of this makes Pivot immune to becoming legacy. No software company can make that promise with a straight face. What we can do is keep questioning our own good decisions, especially once they become comfortable, profitable and difficult to undo.

A perfectly good decision can become a legacy decision if you leave it alone for long enough.

Legacy doesn’t announce itself

Nobody sets out to build legacy software. You get there gradually, surrounded by perfectly good reasons for keeping what you already have. That’s what makes it difficult to spot from the inside. The architecture still works. Customers still depend on it. Revenue is attached to it. There’s always a reasonable argument for adding one more layer instead of changing what sits underneath. AI moved the line for procurement software. It will move again. I’d like us to notice when it does.