Deloitte published new survey results this week, drawn from 3,500 business and IT leaders across 25 countries: 28 per cent of companies now report revenue gains from AI, up from 20 per cent in January, and 81 per cent report faster decision-making and execution.
Then comes the number that explains the rest. Only 25 per cent of companies use AI to create new products or services, redesign core processes or change their business models. The other three quarters are using a generational technology to do the same work slightly faster — and wondering why the returns feel thin.
Speed is not a business model. A faster version of a broken process produces the wrong answer sooner. The companies reporting revenue gains are not the ones with the most AI tools. They are the ones that rebuilt a workflow — how an order moves, how a quote gets priced, how a schedule gets set — and put AI inside it, where it changes a decision instead of decorating one.
AI pays when it is built into the process. Bolted on, it just makes the process louder.
That is custom software work, and it is why the build matters more than the tool. Off-the-shelf AI features are written for the average company; a Toronto manufacturer quoting custom jobs or a logistics firm dispatching its own fleet does not run the average process. Software shaped around the real workflow, integrated with the systems already in place, is what turns faster into revenue. It also forces the ownership question the survey flags: 23 per cent of companies say accountability is unclear when an AI system produces a negative outcome. A proper build names that owner on day one.
Before the next AI purchase, pick one process that touches revenue and redesign it end to end. That single project will teach you more about your AI readiness than another year of pilots.