Buy when the job is already a category
Use an established product when many businesses share the same need and the product already handles permissions, reliability and ongoing change. Scheduling, e-signature, help-desk ticketing and core CRM are common examples. Rebuilding their basic capability creates maintenance work without creating an advantage.
Evaluate the product against the real workflow rather than a feature list. Check export and API access, user permissions, audit history, failure handling and the effort required for staff to adopt it. A capable tool that cannot exchange data may move the problem rather than solve it.
Connect when the tools are good but the journey is broken
Many operational problems live between products. An enquiry arrives in one place, qualification happens in another and the appointment sits elsewhere. A focused integration can preserve the tools people trust while moving context and decisions through the full journey.
Connections need more than a successful API call. Decide how duplicate records are handled, which system owns each field, what happens when a service is unavailable and how a person can replay or correct a failed step.
Build when the workflow is distinctive
Custom work makes sense when the process reflects genuine operating knowledge, the gap is valuable enough and existing products force damaging compromises. The useful custom layer may be small: a qualification console, a proposal assistant or an interface that combines three systems around one decision.
Start with a thin release. Avoid replacing systems that already work. Build the part that expresses your rules or removes the costly join, then rely on established services for identity, payments, communication and storage where practical.
Use AI where the input is variable
AI is helpful when the work involves messy language, summarisation, classification or drafting. Conventional rules remain better for permissions, money, eligibility, irreversible actions and calculations that must be repeatable. Strong systems combine both: AI prepares or proposes, deterministic logic constrains, and a person handles high-impact uncertainty.
The decision should follow the workflow, risk and ownership. Buy the commodity capability, connect the journey and build only the part that deserves to be yours.