AI
Buy, Configure, or Build: AI Has Changed the Decision
Tony Ruiz
The traditional buy-versus-build discussion assumed that custom software was slow and expensive, while packaged software was faster and safer. AI-assisted development is changing that cost curve—but it has not eliminated lifecycle responsibility.
McKinsey's 2026 global survey found that 32% of respondents said their organizations had decided against buying at least one software product or feature because agentic coding tools made internal development viable.[^2] That is a meaningful shift. It is not a reason to build everything.
The better question now has three options: buy, configure, or build.
Buy commodity capability
Buying is usually strongest when the workflow is common, the vendor has mature security and compliance, integrations are available, and differentiation is low.
Payroll, basic ticketing, commodity collaboration, and standard accounting are rarely where a company should invent its advantage. A strong product can spread maintenance, regulatory, and security investment across many customers.
But price the full commitment: licenses, implementation, integration, data migration, premium modules, contract constraints, and the cost of leaving.
Configure when the product is close
Many decisions are not truly buy or build. A platform may cover 70–90% of the workflow, while APIs, automation, and a focused custom layer close the important gaps.
Configuration works when the vendor's underlying model fits the business and the remaining differences can be isolated. It fails when teams bend a distinctive process until it looks like the software, then rebuild the missing logic through fragile workarounds.
Build where the workflow is the advantage
Custom development deserves serious consideration when:
- the workflow differentiates the customer experience or operating model;
- proprietary data or decision logic creates advantage;
- packaged tools force costly process compromises;
- integration depth is central to the outcome;
- control, latency, privacy, or deployment constraints are unusual; or
- a narrow internal product can replace several overlapping subscriptions.
AI-assisted engineering can reduce the cost of implementation, modernization, testing, and documentation. It does not remove product management, architecture, security, support, or change ownership.
Use a decision sheet, not a preference
Compare each option across seven dimensions:
| Dimension | Leadership question |
|---|---|
| Differentiation | Would owning this workflow strengthen our advantage? |
| Fit | How much process distortion does the option require? |
| Data | Who controls access, retention, portability, and learning? |
| Risk | Which party owns security, compliance, and failure response? |
| Time | When can real users receive a supportable outcome? |
| Economics | What is the three-year total cost, including people and exit? |
| Adaptability | How quickly can the workflow change when the business changes? |
Revisit the decision as development economics and vendor capability move. A product that was sensible to buy two years ago may now be practical to replace. A custom system may also become an unnecessary liability when a mature platform catches up.
Cayru helps leaders evaluate the real tradeoffs, configure platforms where they fit, and build custom AI-enabled products where ownership creates strategic value.
Compare one software decision across fit, economics, control, and adaptability.
