ConstructConnect CTO Gaurav Singal says a rise in AI use, or even a jump in features shipped, cannot show whether AI made engineering better. His practical test follows the work from software quality and delivery through to customer outcomes, giving teams a more useful way to judge what their AI investment is buying.
Watch Desk analysis
What happened
In a Forbes Technology Council essay, Singal says ConstructConnect’s production feature count rose five to ten times in an equivalent quarter after it built an AI adoption dashboard. But he cautions that features vary in size, and the figures alone could not show how much of the increase came from AI or what customers gained.
He recommends tracking quality with measures such as defects found after release, test effectiveness and unresolved security findings. For delivery, he points to deployment frequency, time from commit to production, failed changes and recovery time. For business impact, teams should choose outcomes relevant to the feature, such as customer satisfaction, task completion time, revenue or cost to serve, then agree a baseline and review date. His Forbes essay gives the example of measuring how long construction estimators take to review documents, how accurate the result is and how much manual correction remains.
Why it matters
AI usage dashboards can show that people have adopted a tool. They cannot, by themselves, show that customers are better served or that faster development has not brought more defects and rework along for the ride.
Singal’s framework gives technology leaders a way to connect the investment to the work it is meant to improve, rather than treating activity as proof of value.
Our read
This is a sensible test: measure the job, not the number of times someone opened the tool. Singal’s reported feature increase is an example from his company, not proof that AI caused the change. The stronger takeaway is to establish a baseline, track quality and delivery, and check whether the customer outcome actually improved.
What to watch
- Whether teams publish comparable before-and-after measures, including quality and rework.
- Whether claimed time savings translate into better customer outcomes or simply more output.
- Whether AI investment reviews account for setup, training and human review effort.
Discussion spark: When judging AI-assisted engineering, should customer outcomes outweigh a rise in features shipped, even if those outcomes take longer to measure?
Sources and evidence
- Your AI Adoption Is High. Is Your Engineering Delivery Better? – Forbes (10 November 2026, 13:45 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.