Microsoft: Rob Collie and the Personalized Future of AI

Rob Collie and His Vision for AI in Business

Rob Collie, who held a key engineering position at Microsoft, shares his thoughts on the integration of artificial intelligence in the business world. According to him, the key lies in customizing AI models to meet the specific needs of each company. He emphasizes the importance of designating internal "artisans" to lead this crucial transition. These ideas stem from a discussion with Collie, who is set to publish a book titled "Fair Game" on organizational AI strategy, scheduled for August 11.

Collie spent 13 years at Microsoft, where he contributed to the creation of Power BI, a platform that allows users to visualize data and integrate insights into other Microsoft applications. After this experience, he founded his own consulting firm specializing in data and AI.

Like many leaders, Collie has dedicated the past few years to understanding the potential impact of AI on businesses. His main conclusion is that many companies are approaching AI incorrectly. The future is not just about purchasing AI software or creating in-house language models. The real opportunity lies in learning from existing AI models to understand the internal workings of the company and empowering the right people to drive this change.

The Illusion of Off-the-Shelf AI

In developing his AI strategy, Collie has observed a persistent gap. Although AI is often presented as a revolution, leaders struggle to leverage it effectively. The first wave of AI adoption in business was based on the idea that buying AI software was enough to transform a company. However, this approach has shown its limitations.

Companies need to adapt AI to their actual operations. Satya Nadella, CEO of Microsoft, asserts that every company will eventually develop its own AI, which Collie refers to as the "deep end of the pool": training or modifying specific language models. While this is a future perspective, Collie suggests that most organizations should start with the "shallow end." Current language models are already highly effective if viewed as new employees needing training on the company's specifics.

These models, despite having extensive knowledge, do not understand internal processes or the company's strategy. Rather than creating new models, companies should build systems around existing models to provide them with the right context. Language models can be seen as "Lego bricks": by connecting them to the company's data and workflows, meaningful results can be achieved without the need to develop a proprietary model.

Betting on "Artisans" for the Future of AI

The greatest opportunity in AI does not necessarily lie in hiring specialized researchers. It is more about identifying individuals who are already solving problems within the company. Software developers and data professionals are valuable for connecting language models to enterprise systems, but another group deserves attention.

In his book, Collie refers to them as "artisans." These are the individuals who create essential spreadsheets, automate repetitive tasks, design dashboards, and write scripts. They are deeply integrated into the company and solve problems that the IT department does not have time to address.

According to Collie, about one in sixteen people fits this artisan profile, a number greater than that of professional software developers. AI now enables these artisans to create real software, making them a valuable resource for the company.

When it comes to choosing AI leaders, Collie would not focus on those with the most AI knowledge, but rather on those who are already seen as builders within the company. He recommends looking for "Excel gurus" or those whom everyone turns to for solving complex problems. These individuals, through their ability to create innovative solutions, are best positioned to lead the transition to AI.

Companies should not wait for a top-down AI strategy before starting. They operate through thousands of individual workflows, and no leader can redesign them all. Successful companies will begin with small victories in customization, learn from these successes, and gradually expand.