Short answer
Test existing software against one real workflow before commissioning a new tool. If the gap is between systems, evaluate an integration. Consider a custom build when an important need remains unmet and you can support the tool after launch. Compare all three options with the same usage, cost period and acceptance criteria. The method below is a decision framework, not a claim about measured client savings.
The UK government's Technology Code of Practice provides criteria for designing, building and buying technology, including user needs, integration and purchasing strategy. It is government guidance; we use these topics here as comparison prompts for a business.
A prototype is not the full project
A convincing demonstration answers one question: can this approach produce a useful result on the examples tested? It does not establish how the tool behaves when an integration fails, a user has the wrong permissions or a supplier changes its service.
Whether code is written with AI assistance or by hand, include specification, integration, security review, testing and ongoing support in the decision. Agree which tasks the prototype validates and what remains before production. Use representative test data and keep actions that affect customers behind appropriate approval steps.
For security acceptance criteria, OWASP ASVS provides a framework of requirements for developing and testing web applications. Select the requirements relevant to the project and agree how they will be checked.
Three options, not two
| Question | Buy existing software | Connect existing tools | Build an application |
|---|---|---|---|
| Scope | Configure a product for your workflow. | Transfer information and apply rules between systems. | Implement the agreed workflow and interfaces. |
| Costs to include | Licences, setup, migration, training and integrations. | Existing subscriptions, integration build, usage fees and maintenance. | Build, hosting, third-party services, testing, maintenance and support. |
| Key checks | Exceptions, access roles, export options and vendor support. | APIs, permissions, limits, duplicate handling and recovery. | Acceptance tests, ownership, documentation and operational responsibility. |
| Possible fit | A product handles the important steps with acceptable compromises. | Your systems work, but information does not move reliably between them. | An important workflow remains unsupported by products or integrations. |
Four reasons to evaluate a build
- A distinctive workflow matters to customers. Identify the specific step and test whether existing products can support it before assuming it requires custom development.
- A product leaves a costly gap. Describe the missed outcome, the workaround and its frequency. Compare that cost with the full cost of building and maintaining the replacement.
- Several tools require repeated manual reconciliation. First evaluate an integration. A focused application is another option if the required workflow cannot be handled reliably that way.
- Hosting or access requirements rule out some products. Confirm the actual requirement and assess suitable hosted or self-hosted products too. A data-location constraint alone does not require a custom build.
When to evaluate existing software first
Start with established products when your need is a standard business function or requires ongoing specialist maintenance. Buying does not remove your responsibility for configuration, permissions or validating that the product meets your requirements.
- Accounting and payroll: check local requirements, supported workflows and the provider's update and support arrangements with the people responsible for these functions.
- Authentication, payments and email delivery: evaluate maintained services before taking on the engineering and operational responsibilities yourself.
- A subscription you want to avoid: compare the fee with migration, hosting and maintenance costs. Removing a monthly licence does not establish an overall saving.
The costs each side hides
For a purchased product, estimate setup, migration, training and any workarounds that remain. Check pricing at your expected user and usage levels, what happens when you exceed limits, and whether you can export usable data when you leave.
For a custom tool, include operation after launch: hosting, external services, monitoring, backups, maintenance and support. Ask which items the proposal includes rather than assuming they are covered. Our guide to commissioning custom software explains the scope and contract questions to discuss.
Compare the options over the same period, for example three years, using the same expected headcount and transaction volume. Test a low-usage and a high-usage scenario. Record uncertain costs and the assumptions behind them. There is no predetermined year when one option becomes cheaper.
The integration option
Consider a business where the CRM holds the customer, the accounting tool holds the invoice and a scheduling tool holds the job. If people copy information between them, a focused integration may address the gap without replacing all three systems. This is an illustrative scenario, not a reported client result.
Check API access, permissions, record identifiers, usage limits and what happens when a step fails. Define how duplicates are prevented, who reviews exceptions and how to resume work safely. A small integration still needs documentation, monitoring and someone responsible for changes.
As one concrete example, n8n's error-handling documentation explains how an error workflow can respond to an execution failure. Check the equivalent mechanism in whichever integration tool you evaluate, and test it before launch.
The question that settles it
Name the person responsible for accepting the workflow and the person or service responsible for operating it. They need time, access and clear escalation arrangements. If ownership is unresolved, settle it before launching a custom tool or integration. Purchased software also needs an internal administrator and a defined support route.
Run a focused comparison
- Document one workflow, including the exceptions your team handles manually and the outcome each user needs.
- Try two plausible products against the same examples. Record what works, what needs configuration and what remains manual.
- Estimate buying, connecting and building with the same usage assumptions. Include ongoing costs and the cost of leaving.
- Agree acceptance criteria, a project owner and the support route. If uncertainty remains, test a limited prototype before expanding the scope.
Bring the workflow and the gaps you found to our AI automation and internal tools service. A first conversation helps assess whether a project together makes sense; it does not commit you to a build.
Before accepting a custom tool or integration, define the workflow tests to run before launch and the evidence your team will review.
Frequently asked questions
When is a custom internal tool worth considering?
Consider it when an important workflow cannot be handled adequately by existing products, or when connecting those products would still leave a costly gap. Test that gap against real examples. Compare the build, integrations, maintenance, access controls and exit arrangements before deciding. A custom tool is an option to evaluate, not an automatic saving.
Does a quick AI-assisted prototype settle the build-or-buy decision?
No. A prototype can help you test a workflow, but it does not establish the cost or readiness of a production system. Check integration failures, permissions, backups, support and changes to connected tools. Estimate those requirements separately from the first demonstration.
What costs should we check before buying?
Check licence terms at your expected user count, setup, migration, integrations, training and export options. Test the exceptions your team handles manually. Include the cost of any workarounds that would remain and confirm which support services are included.
What costs should we check before building?
Include specification, implementation, testing, hosting, third-party services, monitoring, backups, maintenance and support. Agree who owns the accounts, documentation and code rights. Ask which costs are included in the proposal and which continue after delivery.
Is there a middle option?
Yes. You can keep existing systems and connect them with a focused integration. This can suit a business whose main difficulty is transferring information between tools. Validate API availability, permissions, usage limits, error recovery and maintenance before selecting that approach.
How do we start a focused evaluation?
Document one workflow and its exceptions. Test two plausible existing products against the same examples, then compare buying, connecting and building over the same period and usage assumptions. Name the person who will validate the result and maintain it. This is a proposed decision method, not a benchmark from client results.
Bring one workflow, your current tools and the gap you want to resolve to a 30-minute first conversation. Book a call.

