STKD.DIGITAL · BUSINESS IN MOTION
STKD.digital
Menu
All software and integrations insights

Software and integrations · 4 MIN READ

Buy, Integrate or Build? A Practical Software Decision for Business Owners

Custom software is not the default answer to a disconnected operation. Compare the work you need to own, not just the first invoice.

THE SHORT ANSWER

Buy or configure when a supported product fits the important requirements. Integrate when the products work but their handoff does not. Build when a valuable recurring need cannot reasonably be met either way. If teams disagree about the rules or ownership, repair that process first rather than automating the disagreement.

There are more than two options

For an established company, the question is not whether someone can produce an impressive demo. It is whether the proposed system solves an important recurring problem at an acceptable total cost, with an owner who can keep it working. AI-assisted development changes how code is produced; it does not answer that business question for you.

Describe the work before choosing the tool

Choose one workflow and write down its trigger, inputs, decisions, actions and exceptions. Identify which information is authoritative and who can approve a consequential change. Watch the people doing the work rather than relying only on a manager's idealized process map.

Then distinguish a product limitation from an organizational disagreement. If two departments use different definitions of a qualified customer, connecting their systems will not resolve the definition. If a purchased product supports the workflow but nobody configured it, a custom rebuild may be an expensive way to avoid administration.

The strongest brief names the friction without presupposing the answer. For example: approved orders are retyped into the fulfillment system and errors are discovered only after dispatch. That describes a problem to evaluate. We need a custom platform describes a solution before the investigation.

When to buy or configure

Buying is a strong candidate when the process is common, your requirements fit a supported product and the vendor's operating model is acceptable. Configuration is worth evaluating when the core platform already contains the necessary data and controls. Check the actual plan limits and integration capabilities rather than trusting a broad feature label.

Leon Merkel's public LinkedIn account describes reconsidering a custom CRM once maintenance consumed capacity needed elsewhere. We do not treat that anecdote as proof that every CRM should be bought. It illustrates the opportunity-cost question an owner should ask. Read the discussion.

Ask for a demonstration using representative records and awkward cases. Include a reassigned employee, a canceled order, an incomplete customer record and an attempted unauthorized action. A sales demo that only shows the ideal path tells you little about daily operation.

When to integrate

Integration fits when the core products perform their individual jobs but the handoffs are weak. A small connection can move an approved record, create a task, attach context or reconcile status without requiring the team to relearn every application.

However, integration is not just copying fields. Decide which system owns each field, how duplicate records are recognized and what happens when one system is unavailable. One-way transfer is often simpler than two-way synchronization, but it must fit the actual workflow. Do not promise real-time synchronization when a scheduled reconciliation would be more appropriate and easier to operate.

Illustrative example: a service company keeps its established CRM and billing system while adding a controlled approval-to-project handoff. The custom work lives in the gap, not in replacing both platforms. This is a design example, not a STKD customer story.

When to build

Building becomes more defensible when a recurring, valuable workflow cannot be supported reasonably by existing tools, or when a specific capability genuinely differentiates how you serve customers. Keep the first release narrow enough to test with real operators and retain a manual fallback.

AI-generated code still needs verification. GitHub's own guidance notes that agent-generated code can be incorrect or insecure and should be reviewed and tested. A vendor should explain its review process rather than using AI as a reason to skip it. GitHub responsible-use guidance.

Your scope should include permissions, auditability, recovery, documentation and maintenance, not merely screens and features. If the company cannot name an owner for the system after launch, pause before authorizing a critical custom build.

The decision checklist

Compare all options against the same requirements. Do not give the custom option a best-case estimate while pricing the purchased option with every possible expense.

  • Is the problem a process gap, a configuration gap, an integration gap or a genuine product gap?
  • What business outcome justifies changing the workflow?
  • What existing data and system ownership should be preserved?
  • Can each option handle representative exceptions and permission boundaries?
  • What are implementation, operating, change and exit costs?
  • Who maintains the solution, and how do we recover or leave if it fails?

Our starting recommendation

Preserve working systems unless there is a specific reason to replace them. Repair the narrowest valuable gap, verify adoption and outcomes, and expand only when the evidence supports it. Custom software should earn its place through a clear operational advantage, not because rebuilding has become fashionable.

Sources and further reading

Find my biggest bottleneck