MIT finds digital investments capture 1/3 value

Digital transformation is having a weird second act. Companies spent years buying cloud tools, automation, analytics, and now AI. But the payoff still looks thin. The new MIT Technology Review pieces published on May 11 land on the same uncomfortable point from two angles: most firms still start with the tool, not the problem, and employees are already using AI before leadership has figured out the rules. ### What is the actual finding? The headline number is not new, but it keeps surviving every new wave of enterprise tech. McKinsey’s digital strategy survey said most organizations captured less than one-third of the impact they expected from recent digital investments. In a later banking analysis, McKinsey put harder numbers on that gap: large companies captured 31% of expected revenue lift and 25% of expected cost savings from digital and AI transformations. (technologyreview.com) ### Why do companies miss by that much? Basically, they buy capability first and then go hunting for a use case. The MIT piece frames that as the core mistake — companies begin with what the technology can do and bolt applications onto the business later. That creates fragmented tools, awkward handoffs, and customer experiences that feel stitched together instead of redesigned end to end. (mckinsey.com) ### What’s the alternative? The phrase MIT uses is “customer-back engineering.” Start with the customer problem, the friction point, or the broken journey. Then work backward to the workflow, the data, and finally the technology. Capital One describes this as getting engineers closer to customers through empathy sessions, support embeds, ride-alongs, and hackathons built around real user pain. The point is simple — if engineers only meet the software, they optimize software; if they meet the customer, they can redesign the service. (technologyreview.com) ### Why is finance showing the problem so clearly? Because finance is supposed to be the controlled part of the enterprise. But the MIT finance piece says AI arrived there more like a quiet insurgency. Employees are already using it for things like variance commentary, fraud work, contract review, and close narratives, while executives scramble to add governance after the fact. That flips the normal sequence. Instead of plan, approve, deploy, companies are now deploying and then trying to clean up the controls. (technologyreview.com) ### Is governance the whole issue? Not really. Governance matters, but it is standing in for a bigger operational gap. MIT’s March report on enterprise AI said projects stall without integrated data, stable workflows, and clear operating models. Gartner’s estimate there is blunt — more than 40% of agentic AI projects could be canceled by 2027 because of cost, inaccuracy, and governance problems. So the bottleneck is not just model quality. It is the plumbing around the model. (technologyreview.com) ### Are other surveys seeing the same thing? Yes — and the pattern is striking. PwC’s 2026 operations survey found 89% of leaders said their tech investments had not fully delivered expected results, even while 85% felt ahead of competitors in digital transformation. That is classic enterprise optimism: everyone thinks they are advanced, but almost nobody thinks the returns are there. Poor data quality, siloed operating models, and weak enterprise-wide embedding keep showing up as the drag. (technologyreview.com) ### So where does the value actually come from? Turns out the winners do the boring hard part. They connect strategy, process design, data, talent, and governance at the same time. McKinsey’s banking work showed digital leaders materially outperformed laggards on shareholder returns and profitability, but only after building the organizational machinery that lets digital work compound. The technology helps — but the operating model is what lets the gains stick. (pwc.com) ### What’s the bottom line? The story is not that digital failed. It is that buying software is the easy part, and capturing value is the real transformation. In 2026, the scarce skill is not “can we deploy AI?” It is “can we redesign the business around it before the tool outruns the organization?” (technologyreview.com) (mckinsey.com)