CIOs: Are you ready to kill your AI darlings?
CIOs: learn a disciplined framework to stop underperforming AI pilots, consolidate initiatives, and govern AI for lasting value....
by José Parra Moyano, Karl Schmedders, Olivier Laplace, Benjamin Torben-Nielsen Published November 19, 2026 in Brain Circuits • 4 min read
Generative AI (GenAI) lets a domain expert build a working software solution by handling decisions about how the solution is structured. The human supplies the intent; the AI determines the architecture. This can speed up the development process and requires less engineering expertise, but also creates blind spots. The team building the tool may not know how it’s actually structured, what the cost of deploying it will be, and what maintenance it will entail. Leaders are also subject to these blind spots, and their expectations about how easy it is to build with GenAI can lead them to ask for more than their teams can deliver.
Engage with software engineers early to anticipate the true complexity of the solution that needs to be built. Keep them sufficiently involved to understand the solution’s architecture, and have them assess the real cost of deploying at scale – before the team commits to putting it into production.
Every link between the new solution and the rest of the company creates an obligation. Each data feed, each integration, each downstream process that comes to depend on the tool is something that breaks when the tool changes. The more connected the solution, the more expensive it becomes to fix, replace, or retire.
Keep the number of connections to other systems to a minimum. There’s no right or wrong number here: evaluate the value that each additional connector provides, and whether it outweighs the complexity of the requirements that connection would add.
The temptation after a successful homegrown software pilot is to push it everywhere – more teams, more regions, more use cases. This almost always requires new functionality that is built by someone who was not involved in the original build, never saw the original blueprint, and cannot fully grasp the design beneath. When failures surface, the new teams don’t know how to diagnose or fix them. The solution is then slowly abandoned by the new groups because it’s not viable without the new functionality and fixes.
Be disciplined about who the core users are. Define which users and needs a build will serve before the pilot succeeds, then hold that boundary when a tool that works for the original set of users tempts everyone to widen it.
This is not a case against internal teams. It is a case for deploying them where they create the most value. Build where you are different, for narrow internal tools, or to prototype what you need, and partner on solutions where specialists live the problem every day. A specialist AI startup works on one problem, for years, across tens or hundreds of clients. Every edge case solved for one client makes the product better for all the others – a learning curve that is hard to copy-paste.
Before you build your own AI solutions, ensure there are people in the organization who understand the intended build well enough to avoid blind spots, limit interdependencies with other systems, and stick to a defined and focused scope.
This article is based on Turn Vibe-Coded Prototypes Into Tools Your Company Can Actually Use by the same authors in Harvard Business Review, 16 October 2026.
Professor of Digital Strategy
José Parra Moyano is Professor of Digital Strategy. He focuses on the management and economics of data and privacy and how firms can create sustainable value in the digital economy. An award-winning teacher, he also founded his own successful startup, was appointed to the World Economic Forum’s Global Shapers Community of young people driving change, and was named on the Forbes ‘30 under 30’ list of outstanding young entrepreneurs in Switzerland. At IMD, he teaches in a variety of programs, such as the MBA and Strategic Finance programs, on the topic of AI, strategy, and Innovation. He is also the Program Director of Leading Teams in the AI Era Sprint.
Professor of Finance and Strategy
Karl Schmedders is a Professor of Finance and Strategy, with research and teaching centered on sustainability and the economics of climate change. He directs the Strategic Finance (SF) program and teaches in the Executive MBA programs. Passionate about sustainable finance, Schmedders believes that more attention needs to be paid to on the social (S) and governance (G) aspects of ESG to ensure a fair transition and tackle inequality.
Tech Venture Capital at Vi Partners
Olivier Laplace is Managing Partner at VI Partners. With 18 years’ diverse experience as a tech entrepreneur, software developer, investment banker, and VC, Laplace led Swiss Post’s venture arm, Swiss Post Ventures, investing in 14 startups. An MBA holder, he initiated corporate VC, led M&A at Kudelski, and founded a mobile tech startup.
AI and Data Strategy Expert
Benjamin Torben-Nielsen, PhD, MBA, is an AI expert with more than 20 years of experience turning artificial intelligence research into practical business solutions. With a background in AI and neuroscience, he has led the development of AI-powered products, contributed to multiple patents, and authored more than 40 peer-reviewed scientific publications. His work focuses on AI strategy, responsible technology adoption, and the use of data and AI to improve business performance and decision-making.
November 10, 2026 • by Achim Plueckebaum, Michael R. Wade, Konstantinos Trantopoulos in AI
CIOs: learn a disciplined framework to stop underperforming AI pilots, consolidate initiatives, and govern AI for lasting value....
November 4, 2026 • by Amit Joshi in AI
CHROs, assemble: six practical moves to master honest AI adoption, win staff buy-in, and keep transparency, upskilling, and safety at heart....
November 3, 2026 • by Misiek Piskorski in AI
Unlock competitive edge via four data-strategy moves: consolidate your knowledge, collaborate data clean rooms, deploy agents for data quality, and use error data....
October 29, 2026 • by Carlos Cordon, Konstantinos Trantopoulos , Michael R. Wade in AI
Master AI-enabled operations: identify persistent bottlenecks, unite ops/data/IT in one hub, and design human-centered AI systems...
Explore first person business intelligence from top minds curated for a global executive audience