From my perspective, this is the wrong question, based on a flawed view of how software is developed and deployed in large organizations.
What is the right question? Actually, there are many. Here are some examples:
- Can AI help reduce technical debt ? In other words, can it help organizations migrate away from mission critical legacy software that is increasingly difficult to enhance and maintain. This article from the Wall Street Journal describes how Morgan Stanley is using AI to reverse engineer decades old COBOL code with the intent of replacing it. How Morgan Stanley Tackled One of Coding’s Toughest Problems - WSJ
- Can AI help reduce the backlog of user requests for maintenance and enhancements of existing software? Every IT organization faces this, and every user organization is frustrated by it. Use AI to accelerate the analysis, development and testing, and begin working down the list. Your users will thank you.
- Can AI help integrate the various pieces of software that have been built, bought and subscribed to over the years? Large organization have heterogeneous IT ecosystems, with subsystems that need to share data. This often happens today in spreadsheets and other "sneakernet" artifacts. AI Agents may be a solution, but using AI to streamline the collection, matching and reconciliation of disparate data sources may be a simpler solution that generates value in the short term.
AI can help you reduce costs but framing it exclusively as a cost cutting mechanism is myopic and unnecessarily threatening to your organization. Treat it as a quality improvement opportunity, and you might find adoption improves too.
