Perspective · Strategy
Build, buy, or configure: how to get your AI.
Once you've picked a workflow to improve, the next question is how to get the AI to do it: build something custom, buy a finished product, or configure tools you already have. Vendors will push you toward whichever one they sell. The honest answer depends on the workflow — and for most first projects, the option people overlook is the right one.
The three paths
Strip away the jargon and there are three ways to get AI into a workflow:
- Buy: a finished, off-the-shelf product built for a specific job (an AI scheduling tool, a support copilot, a niche SaaS).
- Build: a custom solution made for your exact process, from scratch or on a development platform.
- Configure: take flexible tools you already have or can rent — your CRM's built-in AI, an automation platform, a general assistant — and shape them around your workflow.
Most "AI strategy" debates frame it as build vs. buy. The third option, configure, is usually the fastest, cheapest way to prove value — and it's the one I steer most first projects toward.
When to buy
Buy when your need is common and a mature product already does it well. If thousands of businesses have the same problem — scheduling, e-signatures, basic support deflection — someone has built a polished tool for it. Don't rebuild what you can rent. The risks to check: does it fit your real workflow (not just the demo), what does it cost at your volume, and what does it do with your data?
When to build
Build when the workflow is genuinely specific to your business and is a real source of advantage — and only after you've proven the value a cheaper way. Custom build gives you an exact fit and full control, but it costs more, takes longer, and you own the maintenance forever. Building should be a deliberate investment in something that differentiates you, not the default reflex for project one.
Why "configure" usually wins first
For a first project, configuring tools you already have (or low-cost flexible ones) is almost always the smart move. It's fast, it's cheap, it reuses systems your team knows, and it lets you learn how AI behaves in your environment before you commit real money. You can often prove or disprove a workflow in weeks by configuring, then make a clear-eyed build-or-buy decision after you've seen real results. Configure to learn; build or buy to scale.
The decision in one question
When a team is stuck on this, I ask one thing: "If this works, is it worth owning custom software for — forever?" If the honest answer is no, configure or buy. If it's a genuine yes, still prove it by configuring first, then build the version you'll commit to. That sequence quietly avoids the two most expensive AI mistakes: building something nobody ends up using, and buying a platform that doesn't fit.
A caution on lock-in
Whichever path you choose, protect your ability to change your mind. Keep your data portable, avoid tools that hold your information hostage, and prefer options your team can actually maintain. The goal isn't the cleverest stack — it's the one you can govern, afford, and walk away from if it stops earning its keep.
Not sure which path fits your workflow?
An AI Opportunity Assessment looks at your specific workflow and recommends the build/buy/configure path with the best payback — plus a 30/60/90-day plan to prove it. About two weeks.