AI Didn't Replace My Dev Team — It Became My Ops Team

There is a version of the AI conversation that happens in almost every founder circle right now. Someone mentions AI, and the discussion immediately goes to code generation — GitHub Copilot, Claude Code, Cursor, vibe coding. It's a real story and it's worth having. I wrote about it last week.

But there's a different AI story that I don't hear enough, and it's the one that has actually changed how Threefold operates day to day. It's not about code. It's about the ten hours a week of overhead that used to disappear into proposals, follow-up emails, internal documentation, meeting recaps, and the general bureaucracy of running a small professional services firm.

That's where AI has become genuinely indispensable for me. Not as a developer. As an operator.

The overhead nobody budgets for

When you're running a services business — especially a small one — a large percentage of your working hours go toward work that isn't the actual work. The deliverable your client pays for might take twelve hours. But the proposal that won the engagement, the onboarding documentation, the status update emails, the post-project recap, the process template you built so the next migration goes faster — that's another eight or ten hours on top of it.

For the first couple of years at Threefold, I just accepted this as the cost of running the business. Some weeks it felt like I spent as much time managing the business as doing the business. That ratio is unsustainable, and it's invisible until you try to scale.

AI changed that ratio. Not by writing better code — by handling the layer of work that sits between the code and the client.

What this actually looks like

Here's where I use AI at Threefold that has nothing to do with software development:

Client proposals. Every new engagement starts with a proposal. Before, I'd write each one from scratch or from a rough template I'd saved somewhere. Now I start by describing the client, their situation, the scope of work, and what they've told me they care about most. I get a structured first draft in minutes. I still revise it — heavily sometimes — but I'm reacting and refining instead of generating from nothing. The thinking is mine. The blank-page problem is gone.

SOPs and process documentation. Threefold does a lot of church data migrations. We've done enough of them now that we have a clear process, but for a long time that process lived in our heads. Getting it out of our heads and into documentation that a new person could follow — that used to take forever. AI makes that dramatically faster. I describe what we do and why, and I get a structured document I can edit into something real.

Client-facing communication. Not every email, but the ones that require care. When a migration hits a snag and I need to communicate a delay to a client without creating panic, I'll draft what I want to say, then ask AI to help me pressure-test the tone. Is this clear? Does it sound defensive? Am I being specific enough about next steps? It's like having an editor in the room.

Meeting notes to action items. After a client call, I'll paste in rough notes and ask for a clean summary with action items separated by owner. This sounds minor until you're running six concurrent client relationships and trying to remember who promised what by when.

function generateClientProposal(client, scope, priorities) {
  const context = {
    clientName: client.name,
    currentPlatform: client.currentSystem,
    painPoints: client.expressedChallenges,
    decisionDrivers: priorities,
  }

  const structure = {
    executiveSummary: summarizeSituation(context),
    proposedApproach: defineScope(scope, context.painPoints),
    timeline: buildTimeline(scope.complexity),
    investment: calculateInvestment(scope),
    whyThreefold: tailorValueProp(context.decisionDrivers),
    nextSteps: defineNextSteps(),
  }

  return structure
  // The thinking is still yours. The blank page problem is gone.
}

What AI doesn't replace

It doesn't replace the relationships. The reason a client hires Threefold is not that our proposals are well-structured. It's that we understand their world, we've earned their trust, and they believe we'll take care of their data and their people. AI can help me communicate that. It can't create it.

It also doesn't replace judgment. When a client is in a difficult situation and needs real advice about whether to migrate now or wait, that's not a prompt. That's experience and context that I've built over years in this industry. AI helps me prepare for that conversation. It doesn't have it.

And there's a category of work where AI consistently underperforms: anything that requires deep knowledge of our specific clients and relationships. The more I can brief the AI on context, the better it does. But there's a floor below which it's just going to produce something generic that I'd have to rewrite anyway.

The honest math

If AI is saving me eight hours a week on business overhead, and I bill at anything north of $100/hour, that's over $800 in recovered capacity per week. The tools cost me $20-40/month. The ROI math is not complicated.

The harder math is behavioral. You have to actually change how you work to capture those savings. It took me a while to stop writing proposals from scratch by habit. It took intentional effort to build the habit of briefing AI before starting a document instead of after I was already stuck.

The tools are there. The question is whether you build the muscle to reach for them at the right moment.

Start with the thing that costs you the most time and has the least emotional complexity. For most founders, that's something like proposals or documentation. Build one good habit. Then add the next.

The code stuff is real. But don't sleep on what AI can do for the ten hours a week you weren't billing for anyway.