Google Brain cofounder and former Baidu AI chief scientist Andrew Ng today announced the launch of the “AI Transformation Playbook,” a prescriptive guide for business leaders looking to create AI-first companies.

The playbook is the latest initiative from Ng, who has been advising startups, developers, and legacy businesses on how to build successful ventures in the AI era.

In recent years, Ng, who is also a Stanford University professor, has been a mainstay at AI gatherings where business executives meet to discuss what it takes to become a good AI company.

“The ‘AI Transformation Playbook’ synthesizes a lot of lessons I’ve learned leading and building these teams at Google and Baidu, as well as lessons I’ve learned discussing and interacting with many CEOs, including many outside the software industry,” Ng told VentureBeat in a phone interview.

The playbook offers five main rules for how to transform a business into an AI-first company:

1. Execute pilot projects to gain momentum
2. Build an in-house AI team
3. Provide broad AI training
4. Develop an AI strategy
5. Internal and external communications

Business executives interested in making AI-first companies face challenges and can make some common mistakes, Ng said.

He warned that just throwing large amounts of data at engineers and expecting success, or misunderstanding what AI can and cannot do can lead to failure.

Having a broad understanding of AI’s potential and limitations can aid more than CEOs and data scientists.

“I think it’s important for many people in a company to know what AI can and cannot do. Maybe not every single person in a company needs a deep level of knowledge, but it definitely goes well beyond [the engineering team],” he said.

Another common mistake is starting with a small pilot project and failing to build momentum and scale initial efforts to make the impact of AI pervasive throughout a company.

Media reports may skew what people understand to be possible because business news tends to focus on AI success stories. Ng compared CEOs who hear about nothing but breakthrough research and efficiency gains to a computer vision system trained to see nothing but cats.

“If you train a learning algorithm and every single picture you show it is a picture of a cat, then it thinks every single thing in the world is a cat, because it’s never seen anything that’s not a cat before. And I think in a similar way, if you show executives a sequence of success stories but no failure stories, then it creates an impression that AI can do anything, and that’s just not true,” he said.

In addition to the playbook released today, Ng is also cofounder of online learning portal Coursera, which was made popular in part due to Ng’s machine learning course.

There’s also deeplearning.ai, and “Machine Learning Yearning,” an instructional book assembled this year.

Earlier this year, Ng also launched AI Fund, a $175 million investment fund to back artificial intelligence startups.

Initial investments include Landing.ai, a company made to help manufacturers incorporate artificial intelligence into their businesses, and Woebot, a conversational AI company making a therapist bot.

Other investments have been made but are not yet public, he said.