About 7 years ago, we were running Cro Metrics on Airtable and spreadsheets. With my product background, I spun up a proof-of-concept to replace our disjointed systems and showed it to our CEO. He also has a product background, immediately recognized the value, and we were off to the races to fully productionize it.

It served us well.

But by 2025, the cracks were obvious: technical debt, patchwork refactors, and an aging presentation layer duct-taped to modern stateless UIs. But the biggest problem wasn’t the code; it was trust. Data anomalies started creeping in. The product team patched quickly, but rogue spreadsheets began appearing.

When people go rogue, that’s the beginning of the end.

Not because spreadsheets are inherently bad (well, they kind of are), but because trust, efficiency, and scale all require a shared system of record.

A traditional refactor would’ve taken six months and meaningful opportunity cost. We’re a small but mighty team. Those resources should create client differentiation, not maintain aging infrastructure.

So we reframed the problem: Refactor the data layer first. Then rebuild everything else with AI-assisted velocity.

Our lead developer built a proof-of-concept using Claude Code. It wasn’t an incremental improvement, it was step-change feasibility. Game on.

Within a month, we had a rebuilt system.

Three things emerged:

  1. Our lead developer evolved into a hybrid Product Manager/Developer working directly with teams, shaping requirements, and iterating in real time.
  2. Development velocity increased dramatically.
  3. We decoupled core integrations. Harvest, our time tracking tool, for example, was embedded in our schema. We generalized it. Now we can roll our own time tracking or swap vendors with minimal friction.

That last point matters.

The moat in vertical SaaS is eroding. When you own your schema and normalize your backend, switching costs collapse. Control shifts back to the operator.

With our current architecture, replacing Harvest is likely a week-long project.

The interface isn’t the moat anymore.

The data model is, and we own it.

What This Means for Your Business

  1. The cost of reinvention has collapsed. Legacy codebases are no longer immovable objects.
  2. The backend is the leverage point. Refactor the data layer well, and you unlock integration flexibility, internal tool velocity, and data democratization through APIs and normalized tables.
  3. The hybrid PM/Developer role is becoming essential. It’s not just hands-on keyboard execution; it’s problem framing, architecture decisions, and rapid iteration with stakeholders.
  4. You should be actively evaluating which SaaS tools are core infrastructure and which are convenience layers. We won’t rebuild Slack. But tools like Harvest? That’s a strategic decision now, not a foregone conclusion.

This isn’t solely about cost-cutting.

It’s about control.

It’s about velocity.

It’s about designing systems that scale with your ambition.

This article was written by me, not AI. I used AI as my editor to tighten up language and enhance concepts.

Ready to turn ambitious growth goals into deeper customer connections and measurable business impact?