AI agents are slowly becoming more mainstream and a vital part of the enterprise IT toolkit for increasing productivity—both internally and in consumer-facing functions. A new study this week from Databricks zooms in on how more than 20,000 of the company’s agentic platform customers are using the technology.

Right now, many companies are moving to the point where they’re using multi-agent systems to orchestrate and execute more complex workflows. Databricks found that this type of usage grew 327% in just four months—meaning more companies are reaching the point where they’re more comfortable and trusting in the AI and agents they’ve been developing. About 37% of Databricks’ agent usage goes toward these systems. But agents are still being used for less complex tasks. About a third of Databricks’ agent usage is for information retrieval: Converting unstructured text data into tables that generative AI can use.

AI agents are also taking the lead in database creation and management. Databricks found that agents create four out of five new databases, as well as 97% of all branches and variants of databases. This database work, when given to humans, often takes hours if not days. Databricks found that human-optimized databases might not have always been built for optimal query response, while AI agents can easily handle databases to work with more in-depth and intensive query use.

But even with all of this action, the vast majority of AI pilot projects don’t make it into large-scale use. A large part of the hesitation, Databricks says, is likely a lack of confidence in an enterprise’s own governance and security. In the last year, Databricks found a 7X increase in the number of companies using its governance tools. Compared to others, companies using governance tools put 12 times more pilot AI projects into production.

AI agent use will continue to grow as more companies build trust and enhance their governance standards. But another important aspect of AI agents is making sure they understand your workplace context and handle tasks using the same types of logic and processes as humans. I talked to Cognizant CEO Ravi Kumar S about why context engineering of AI agents is so important. An excerpt from our conversation is later in this newsletter.

Nominations are now open for Forbes AI 50. The eighth-annual list, in collaboration with sponsoring partner Mayfield, will recognize the most promising startups deploying artificial intelligence in financing, scientific discovery, construction and more.

It’s another day when tech companies show their larger strategies—and markets react dramatically. Meta and Microsoft both reported quarterly earnings after markets closed on Wednesday. Both showed larger-than-expected growth, but Meta’s stock surged—increasing more than 7% by Thursday morning—and Microsoft’s crashed—down almost 12%.

Meta’s quarterly revenues hit $59.8 billion, smashing estimates and showing year-over-year growth of 5.8%, with 24% growth in advertising revenue for its entire fiscal year. And the company plans huge increases in AI infrastructure spending, with expected capital expenditures of $115 billion to $135 billion in the next fiscal year.

Microsoft’s revenue also beat expectations, hitting $81.3 billion in the last quarter—a 17% increase. But the company is spending more than expected on infrastructure, and its cloud business is growing—but not as quickly as some investors might have liked.

The huge rise and fall of the tech stocks show how volatile current market conditions are. Erik Gordon, an entrepreneurship professor at the University of Michigan’s Ross School of Business, told Business Insider that the AI bubble is almost as big as the planet Jupiter—and when it bursts, debris will be everywhere. The valuation bubble seems reinforced by other major deals. Earlier this week, Nvidia made a $2 billion investment in cloud computing provider CoreWeave—a move that boosted........

© Forbes