Legacy Workflows Are a Bigger AI Obstacle Than Technology

Harvard Business Review published an article by Graham Kenny and Ganna Pogrebna on how AI can free organizations from legacy workflows. The authors open with the example of logistics company Virtal Systems, whose manager Elena says: “We do not lack capabilities.” The more important continuation is that the company is burdened by its own past. This sentence captures the problem of many enterprise AI projects. The technology is deployed, but processes, decision rules and organizational habits remain the same. If AI merely serves an old process, the result will be limited. Automating a poor workflow creates a faster poor workflow. Companies often ask which model to buy, how quickly to train employees or where to find the first use cases. But if they do not first conduct an inventory of work, they may add another cost layer. Employees keep optimizing for old indicators, managers keep demanding old outputs and the system only adds more administration. The business impact is clearest in companies with long histories. They have experienced people, customers, data and capital, but also old approval chains, complex exceptions and metrics that measure performance according to a past operating model. AI then runs into internal resistance even when nobody explicitly says no. The governance and regulatory impact lies in responsibility for process change. If a company deploys AI in decisions about customers, employees or suppliers, it must know who decides, which data are used and how exceptions are handled. Many old workflows were created at a time when decisions were made by people and the audit trail was informal. The practical conclusion is strict: AI transformation should not start with buying a tool, but with an inventory of work. Management should first describe key processes, remove steps that exist only by habit and determine where AI should decide, recommend or merely prepare supporting materials.

Legacy Workflows Are a Bigger AI Obstacle Than Technology

The full article is available to users registered at Hard Skills