We want to automate our inventory reconciliation process, but the external software contractors we interviewed are pitching expensive, custom-built machine learning models that will take six months to deploy. How do we keep these contractors accountable and avoid overpaying for unnecessary custom code?

Question

Tyler Smith · Accepted Answer

External software developers often want to build complex, custom code because it increases their billable hours and keeps you locked into their ecosystem. As a non-technical owner, you must hold them accountable by demanding simplicity. Never let a vendor build a custom machine learning model from scratch when you can achieve the same operational result using simple, off-the-shelf APIs and existing software.

Before approving any custom development budget, challenge the contractor to explain why off-the-shelf AI agents cannot run your documented standard operating procedures. Modern, pre-built models are highly capable and can be connected to your existing databases with minimal custom work. If they insist on custom development, demand that they break the project down into ninety-day milestones that align with your quarterly Rock cycle.

Ensure that every technical proposal is framed around a clear operations-improvement project rather than abstract software development. If the vendor cannot guarantee a working prototype within the first thirty days using standard APIs, they are likely overengineering the solution. Keep your focus on building system-dependent operations using the simplest, most cost-effective tools available.