What Top AI Labs Are Really Doing With Observability

Pomel says developers and operations teams often blamed each other, while traditional monitoring tools served operations and left developers without visibility into production. Datadog began in 2010 with a cloud-first platform that brought development, security, and operations teams together. Calling it infrastructure monitoring helped customers understand why they needed it.

Pomel says Datadog did not write down company values, choosing instead to make culture visible through how leaders act and whom they hire, promote, or let go. He stays close to product details and samples customer conversations, support requests, and employee feedback. He says this helps leaders understand what is happening on the ground rather than relying only on polished summaries.

Datadog often builds from customer workflows, but Pomel says AI requires companies to act before demand is clear and accept that some bets will be wrong. Its current priorities include making products usable by agents and adding more capabilities directly into the observability data plane. He says the company sees leading AI labs use its products in unusual ways, offering clues about possible future needs.

Pomel says Datadog is moving toward a way of working where teams automate first and write code when needed. After developers independently rebuilt systems in days that might once have taken six people six months, the company decided to change how it operates. Smaller teams can now explore more ideas, though the right mix of engineers, product managers, designers, and security staff remains uncertain.

Looking back, Pomel says he would advise himself to move faster, especially on hiring and firing. He recalls spending too long trying to find the perfect hire and keeping a poor fit too long. The host adds that quicker hiring and course correction can give a company more chances to find the right people.