Systems | Development | Analytics | API | Testing

You opened the AI assistant for the first time with a fair amount of hope. You typed "write test cases for the login page." You got eight test cases back in about three seconds. Valid login. Invalid password. Empty username. Empty password. The kind of list you could have written in your sleep, missing every scenario that actually matters for your product. So you closed the tool and thought: this is fine for the basics, but it does not really get testing. AI is overhated.

Until recently, we extensively used HashiCorp's CDK for Terraform. Then they announced the end of its development. We didn't want to spend months migrating to a new ecosystem, but we needed a replacement. In the Infrastructure Team at Ably, we like Go. I wondered if there was an opportunity for something simpler.

Most AI content assumes your data is in the cloud. But for a meaningful segment of enterprises, cloud-only AI tools block them at the pass. For regulated industries like manufacturing and healthcare, data residency requirements, compliance mandates, security policies, and simple operational reality mean sensitive data must remain on-premises.

Most QA teams already track the basic metrics: how many tests ran, how many passed, how much coverage exists, and how many defects turned up. Those numbers still matter, and engineering leaders will keep asking for them. The real challenge is turning those numbers into decisions, and that gets harder as AI-assisted development speeds up the volume, frequency, and complexity of software change.

In the first part of this series, I argued that discernment, the ability to recognise when an AI-generated answer is wrong, is becoming one of the most valuable capabilities inside an organisation. The question this piece addresses is simpler and harder: who is actually being given the opportunity to develop it? The AI opportunity gap is real. It is not primarily a gap in access to tools. It is a gap in permission. And I believe that gap starts earlier than most leaders realise, often in school.

When a PropTech product struggles, the issue usually isn’t the code. A skilled team can build the wrong thing well: a listing platform on a data model that buckles when the second MLS feed arrives, a CRM that ships to spec while agents quietly stop using it, an MVP that investors, agents, and end users each expected to do something different. That gap between what the business assumes, what users need, and what the architecture can support is what we mean by PropTech product misalignment.