shiphrahx/ai-for-engineering-leaders/build-vs-buy-analysis

Build vs Buy Analysis

Weigh building a capability in-house against buying or adopting a vendor solution, on total cost of ownership, engineering opportunity cost, operational burden, and strategic fit — laid out honestly, without bias toward either side.

Inputs to gather

Gather these before analysing. If any are missing, ask for them in a single batched question — never invent vendor pricing, effort estimates, or team capacity. Mark anything genuinely unavailable as Unknown.

• Capability needed — the problem being solved, with must-haves vs nice-to-haves
• Business context — why now (the triggering pain or opportunity)
• Vendors evaluated — the candidate solutions, if any are named
• Team context — skills, capacity, and whether they could realistically build it
• Constraints — timeline, budget, compliance, data residency

Steps

1. State the decision summary upfront: what capability, why now, and your recommendation — do not bury the lede.
2. Lay out the build option: estimated effort (MVP and full), ongoing maintenance burden, team expertise required, advantages, and the opportunity cost of what the team would not do instead.
3. Lay out the buy option(s): each vendor with pricing, integration effort, limitations, and advantages.
4. Build a comparison matrix scoring options side-by-side on time to value, 2-year total cost, maintenance burden, customisation, vendor risk, and opportunity cost. Include the open-source/self-hosted option's ops overhead in its cost, not just its license.
5. Surface hidden costs both options underestimate — build: edge cases and the management UI usually skimped on; buy: per-seat pricing scaling with headcount and SDK/vendor lock-in migration cost.
6. State the recommendation with reasoning, and name the explicit conditions that would reverse it (e.g. "revisit if seat cost exceeds $X/yr or we need custom logic the vendor can't support").
7. Adapt to context: cost-sensitive startup — weight toward cash cost and surface the open-source option prominently. Enterprise/regulated — add a Compliance & Security row (data residency, SOC 2, audit); these often eliminate options fast. Strong internal platform team with capacity and charter fit — build becomes more attractive; factor team morale. Non-technical audience — drop SDK/implementation detail, lead with cost, time-to-value, and risk.
8. Assemble the output in the format below.

Output format

**[Capability]: Build vs Buy Analysis**
 
**Recommendation:** [Build / Buy (vendor) / Hybrid] — [1–2 sentence why, stated upfront]
 
**Build Option**
- *Effort:* [MVP and full estimate]
- *Ongoing maintenance:* [burden]
- *Advantages:* [...]
- *Risks / opportunity cost:* [what the team won't do instead]
 
**Buy Options**
| | [Vendor A] | [Vendor B] | [Vendor C] |
|---|---|---|---|
| Cost (annual) | [...] | [...] | [...] |
| Integration effort | [...] | [...] | [...] |
| [key capability] | [...] | [...] | [...] |
 
**Comparison Matrix**
| Dimension | Build | [Vendor A] | [Vendor B] |
|-----------|-------|-----------|-----------|
| Time to first value | [...] | [...] | [...] |
| 2-year total cost | [...] | [...] | [...] |
| Maintenance burden | [...] | [...] | [...] |
| Customisation | [...] | [...] | [...] |
| Vendor risk | [...] | [...] | [...] |
| Opportunity cost | [...] | [...] | [...] |
 
**Hidden Costs**
- *Build:* [underestimated complexity]
- *Buy:* [scaling cost, lock-in]
 
**Recommendation: [Choice]**
[Reasoning. Revisit if: (a) [condition], (b) [condition].]

Boundaries

• Never invent vendor pricing, build effort, or reliability claims — mark them Unknown; vendor quotes and reference checks are the user's to obtain.
• Never write a one-sided analysis — be genuinely generous to the option you don't recommend, especially for a controversial or contested decision.
• Never present opportunity-cost or TCO figures as precise — use ranges and state the assumptions.
• Never reduce TCO to sticker price — include maintenance, integration, ops overhead, and opportunity cost.

Chaining

• After the decision is made, offer architecture-decision-record to capture the choice, alternatives, and reversal conditions durably for future engineers.