Why AI Agents Won't Replace Enterprise Software:
Atlassian's Chief Product and AI Officer

Tamar Yehoshua took her current job while the industry was predicting the death of companies like hers. "You know, I joined Atlassian right around the time of the SaaSpocalypse, and people are like, wait, what do you mean you're going to a SaaS company? And I just never believed the narrative," she told CXOTalk. Yehoshua is Atlassian's Chief Product and AI Officer, introduced by host Michael Krigsman as taking the role "after leading product at Google Search and Slack."

Yehoshua's case hinges on how AI changes who uses software. "You have to build great products that people love, that agents can work with, and that are valuable," she said. The user opening your product may not be a person at all: "They might be an agent coming in working with it."

Agents are a second kind of user

Yehoshua argues that enterprise software now has two kinds of users, people and agents, and vendors must build for both. Atlassian's mission is to "unleash the potential of every team," and she extends it: "now that team includes agents, humans and agents."

A Jira issue has to be readable and writable from a developer's coding agent, and in Confluence, "we've gotten to the point where 95% of everything you can do in the UI, you can do through MCP or a CLI." Atlassian's "MCP and CLIs have over 500 tools from Atlassian that we expose to the agents."

Context, not intelligence, is what agents lack

Yehoshua says the missing ingredient for an agent within a company is context, since the models already provide intelligence. "An agent needs the context of an organization," she said, comparing one to "somebody coming into your organization with no background whatsoever." Her formula: "we think of acceleration as intelligence plus context."

Atlassian's answer is the teamwork graph, "the context graph that we build for each customer for their organization." It maps people, knowledge, data, code, and communication, and integrates with Google Docs, SharePoint, email, Slack, and Teams. "Our teamwork graph has over 200 billion entities in it," she said.

In A/B tests of coding agents such as Claude Code with the graph versus MCP alone, "the quality increased by 44%. And very importantly, the token usage went down by 48%." Access follows the person: an agent works with "whatever permissions the human has."

Accountability starts with whoever builds the agent

Yehoshua holds that responsibility for a chain of agents belongs to the human who built the first one, no matter how many steps run unreviewed. Asked by a LinkedIn viewer who is accountable when a chain breaks: "I don't think anybody has figured out the answer to this." The principle she commits to: "Whoever built the agent that starts is still accountable. There always has to be a human accountable."

Yehoshua expects enforcement to become agentic, predicting that "agents are going to be putting in guardrails for other agents." Atlassian already uses the pattern in testing: "We use LLM as a judge." For sensitive data, "We have a product called Guard in our enterprise offering that does a whole data sensitivity layer."

Aim for tenfold gains and measure them

At Atlassian, Yehoshua's answer is to point AI at processes that can be improved tenfold, then verify the gains. "You have to start with what problem you're trying to solve and what your goals are for that problem," she said. The bar: "we ask people to think about what are the processes where you think you can get a 10x improvement. So this isn't where you want just an incremental 10 to 20% improvement."

Her grounding example is Mercedes-Benz, an Atlassian customer of more than a decade whose test engineers file defect tickets in Jira. A Rovo agent now auto-triages those tickets: "that triaging agent reduced the manual work for engineers to triage the bugs by 85%." The caution: "if you're not measuring what the value was, then you might be going too fast and not being able to see the value."

Vibe coding stops at enterprise scale

Yehoshua rejects the idea that large companies will vibe code their way out of buying software. "I think that works for small companies. I don't think it works for large companies," she said. The trap is everything after the prototype: "if you vibe code something, you got to support it," and once internal users depend on it, "your job becomes a software vendor."

"I could just vibe code Slack. I could just vibe code Jira. These just don't last," she said of the industry chatter, adding that "your vibe-coded stuff, it's just not secure. It's not enterprise-ready." Atlassian does see "a ton of vibe coding internally to help people do their jobs, but they're not instead of large SaaS vendors that we buy their software."

That is why the SaaSpocalypse never worried her. "We have 370,000 customers. We are in 85% of the Fortune 500. People run their tier 0 workflows on Atlassian. That's just — you can't substitute that." Her one condition: "Now, if we did nothing, and didn't introduce AI into our products, I would be worried."

Leadership without the hierarchy

Yehoshua believes information no longer has to travel through the org chart, and leaders who do not adapt will see only marginal returns. In her framing, "if you think pre-AI, information flowed through hierarchies." Anyone can now ask Rovo, Atlassian's AI product, for the status of an OKR or the cause of an outage. Her own routine changed: "I, for example, no longer have monthly OKR meetings." An agent reads the Confluence docs, Jira tickets, and Slack messages instead.

Yehoshua thinks leaders face a choice: "all leaders need to change how they're thinking about how they run organizations and where their points of leverage are," and "if they don't, then their organizations will not see efficiencies and speedups. Then it'll be just micro improvements."

Yehoshua's advice to CIOs compresses the argument: "be ambitious, but be educated." Her largest claim is about Atlassian's position with its customers: "We say that we're like the bridge to helping them become AI-native because we're already embedded there."

Watch the full conversation with Tamar Yehoshua and read the complete transcript on the episode page.

CXOTalk prepared this article with AI assistance, based on the verbatim transcript of episode 930. Quotations are unedited from the broadcast.