Dynatrace’s $915M Arize deal bets AI agents are just another app to monitor

Dynatrace closed its $915 million acquisition of Arize on October 1, joining Arize’s tracing and evaluation of AI agents to its own monitoring of the applications and infrastructure underneath them.

It also takes an independent AI observability vendor off the market, one Datadog had invested in.

I sat down with Sean O’Dell, a principal product marketing manager at Dynatrace, on September 25 at the WeAreDevelopers conference in San Jose, five days before the deal closed. Dynatrace was in a quiet period, so he kept his Arize comments general.

When I asked where Dynatrace is headed with Arize, his answer started from the premise that an AI app is still an app. When it breaks, the business needs to know what the outage cost, and the engineers on call need to know why it happened.

“We can ask a question, we can do natural language, we can have fantastic RAGs, we can do evaluations, but at the end of the day, it is an application,” O’Dell said.

Teams need to know whether a response is appropriate, whether it’s hallucinating, and what it cost, he said, and both Dynatrace and Arize already answer those questions. Those AI apps “are still a part of a broader ecosystem of legacy applications, mainframes in many cases,” he said.

Until now, that meant two tools. Arize co-founder Aparna Dhinakaran told The New Stack’s Paul Sawers that teams could debug the AI side in Arize, but a software problem sent them digging through separate traces to find the root cause.

Arize also brings Phoenix, its source-available tool for tracing and testing AI apps, and the developers who use it. O’Dell called that community “hard to find because it’s so early and so new.”

Both companies already have agents that act on the data. Dynatrace’s Bluebox, unveiled in June, reads a team’s code in GitHub, GitLab or Bitbucket, matches it against production data and proposes a fix, according to O’Dell. “So here is your PR,” O’Dell said. Arize’s Signal does the same for Alyx, Arize’s own AI assistant. Arize accepts roughly 65% to 70% of the pull requests Signal writes, Dhinakaran told The New Stack.

That leaves about a third for a person to catch. O’Dell said a team can step back as “a human out of the loop” on Bluebox’s fixes, but he still tells customers to validate a fix before approving it.

Ops has rarely let automation change production on its own. Back in the ITIL days, he said, an automated change came with “1,500 checks and balances.”

“I don’t trust everything that comes out of my [chatbot] or my agent,” O’Dell said. “So why would you?”

He’s right.

The post Dynatrace’s $915M Arize deal bets AI agents are just another app to monitor appeared first on The New Stack.