Skip to main content

Build vs. Buy vs. AI — The New Decision Framework

5 min read
Tech LeadCto

Tech Lead

Use this in design reviews. 'Did we consider the AI option?'

Cto

This framework scales. Teach it to your directors.


Build vs. Buy vs. AI — The New Decision Framework

TL;DR

  • Old model: build or buy. New model: build, buy, or AI-assist. Most teams forget the third option.
  • AI-assist is often cheapest and fastest for internal workflows. Save build/buy for customer-facing or high-stakes.
  • The framework: Is it differentiating? Regulated? Reusable? Answer those, then pick.

You're used to "build or buy." Add "AI-assist" as a third path. It's not always the right one. But it's often overlooked — and for certain problems, it's the obvious winner.

The Three Paths, Defined

Build

  • Your team builds it. Full control. Highest cost, longest timeline. Best when: differentiation, complex custom logic, regulatory needs.

Buy

  • You license a product. Fast to start, ongoing cost. Best when: standard problem, you're okay with their roadmap and limits.

AI-Assist

  • You use AI (public or internal) to augment humans. No product to install, no vendor to manage. Best when: internal productivity, content/code generation, low-risk, high-variance tasks.

The Decision Flow

Step 1: Is it customer-facing or internal?

  • Customer-facing: Build or Buy. AI-assist is rarely a full product. You might use AI in a product, but that's a build decision.
  • Internal: AI-assist is often viable. Drafting, summarization, code gen, triage — try AI first.

Step 2: Is it regulated or sensitive?

  • Regulated (healthcare, finance, etc.): Build or Buy with strict controls. Public AI is usually a no.
  • Sensitive (IP, PII): Same. Keep data out of public models.
  • Internal, low-sensitivity: AI-assist is fine.

Step 3: Is it reusable across the org?

  • One team, one use case: AI-assist. Tweak a prompt, done.
  • Many teams, same need: Maybe buy. Or build a thin wrapper around AI and internalize.
  • Core to product: Build.

Step 4: What's the failure mode?

  • AI hallucinates: Okay for drafts, bad for final output. Use AI for drafts, humans for approval.
  • Vendor goes away: Build or pick a vendor with a strong ecosystem.
  • Build is late: Buy or AI-assist as a bridge.

Examples in the Wild

NeedOld AnswerNew Answer
Code review commentsBuild a botAI-assist (Copilot, Cursor suggest; human approves)
Internal docsBuy Confluence, build wikisAI-assist (draft with GPT/Claude; human edits)
Customer support triageBuy ZendeskAI-assist for routing; human for complex cases
Auth, billingBuyStill buy. Don't AI-assist security-critical infra.
Custom ML model for productBuildStill build. AI-assist for iteration, not replacement.

When AI-Assist Beats Build and Buy

  • Speed. You can have something working in hours, not months.
  • Cost. No license, no team. Just API calls and human time.
  • Flexibility. Prompts change fast. No release cycle.
  • Experiment. Unsure if it's worth building? AI-assist first. If it sticks, graduate to build or buy.

Manual process. Repetitive tasks. Limited scale.

Click "With AI" to see the difference →

Quick Check

What remains human when AI automates more of this role?

Do This Next

  1. List 5 things your team does manually — Which could be AI-assisted? Run one experiment this week.
  2. Apply the framework to one active project — Did you consider AI-assist? If not, why? Document the decision.
  3. Share the framework — Put it in your team wiki. Use it in design reviews. "Build, buy, or AI — which did we pick and why?"