nata.Fieldnotes

TECHNOLOGY DECISIONS

Should You Build a Custom Workflow or Use Existing Software?

Compare the complete decision, including configuration, integration, maintenance, and a way out.

Fran Herrera · Co-founder, Strategy & Operations

Prepared and revised

The right choice depends on the task, its constraints, and the cost of keeping the solution useful. Existing software may fit after configuration. A small integration may solve the gap. A custom build may be justified where the requirement is important, specific, and difficult to meet otherwise. Keeping the current process can also be reasonable when change offers little benefit.

How should you define a workflow software requirement?

Describe the user, input, next action, expected output, and important exceptions. Express the requirement as work someone can complete, such as finding the governing quote before a project handoff. Include failure and recovery so a successful demonstration with clean inputs does not conceal an unsuitable operating model.

What alternatives should you compare with a custom build?

Compare keeping the process, redesigning it, configuring an existing tool, integrating tools, buying another product, and building custom functionality. The NATA Build Decision Record below supports that comparison. Combinations may work: a custom interface can sit above an established record system without replacing it.

Some options belong together. The point is to understand the minimum intervention that meets the requirement. A custom interface may be useful while an established system remains the underlying record.

DimensionEvidence to seek
Task fitA normal case completed from start to finish
Exception fitAn incomplete, conflicting, or unusual case handled appropriately
OwnershipA named operator and maintainer after launch
DataDefined sources, permissions, and export paths
ChangeThe effort required when a business rule changes
FailureA visible error and a workable recovery route
ExitA realistic way to migrate or continue without the solution

What should you test first in a build-versus-buy decision?

Test the requirement that makes your workflow distinctive or difficult, including an exception. In an illustrative design-studio proposal process, maintaining unique exclusions and open decisions through revisions matters more than demonstrating reusable biographies. The result should help choose the implementation, rather than justify a choice already made.

Test that part before investing in a polished editor. Can the solution show what changed? Can someone distinguish approved information from assumptions? Does it preserve the final version the client accepted? The answer may favor a configured product, a custom component, or a simpler workflow.

How do you compare the total cost of custom and existing software?

Compare setup and recurring fees alongside administration, data preparation, maintenance, training, changes, recovery, and exit. State which figures are known or estimated and identify who will maintain the solution. A low build price or familiar subscription is not enough to establish long-term fit.

A low initial build cost can hide dependence on one developer. A familiar subscription can hide manual reconciliation. Neither option deserves an automatic preference. Ask who will change the solution when the business changes and how the company will know whether it still works.

When should AI be part of a custom workflow?

Use AI for a defined task whose output can be assessed in the workflow. State which decisions remain with people, how uncertainty is exposed, and how errors will be reviewed. Drafting a summary is different from confirming a technical specification or committing the business to commercial terms.

NIST’s AI Risk Management Framework is a voluntary framework for incorporating trustworthiness into AI design and use. It does not certify a particular workflow or remove the need to test your application.

How do you avoid dependence on a custom software supplier?

Define record export, access, critical business rules, maintenance ownership, and a workable way to continue during interruption. Test a bounded implementation before committing more of the business to it. Record the evidence and review trigger so the decision can be revisited when requirements or supplier arrangements change.

Run a bounded trial with a stated decision date and success conditions. Include the people who will operate the system, not only those selecting it. A technically successful build can still create work the team cannot sustain.

Record the chosen option, alternatives rejected, evidence, uncertainties, and review trigger. The business then has something more useful than a preference for custom software or a belief that buying is always safer: a decision it can explain and revisit.

Frequently asked questions

When is a custom workflow justified?

When an important requirement is specific enough that available options cannot meet it acceptably, and the business can support the build’s full operating life. Demonstrate that need with real tasks and exceptions before committing.

Is existing software always cheaper than building?

Not necessarily. Configuration, integration, administration, and manual reconciliation also cost time and money. Compare the complete requirement and operating costs rather than only the subscription and initial build price.

Does including AI make the custom option more appropriate?

No. AI needs a defined role, suitable inputs, review, and a recovery route. Its presence does not demonstrate that a custom build fits the business better than another option.

Sources and case basis

This guide presents a working method, not measured client results. Examples are illustrative unless explicitly attributed. Sources linked in the text support the stated principle or published example; they do not validate the NATA worksheet as an assessed instrument. Prepared and revised November 4, 2026.