Anu Bharadwaj on Why Most AI Agents Fail (and What It Takes to Reach Production)

Atlassian
Anu Bharadwaj is President of Atlassian, a leading technology company that helps millions of teams transform the way they collaborate. As an accomplished executive with 20+ years of experience in leading transformations and building billion-dollar businesses for high-growth startups and enterprises, Anu brings deep expertise across Product, Operations, Strategy, CorpDev and Engineering to help teams deliver customer-focused growth. As COO of Atlassian, she was responsible for scaling a distributed workforce in 13 countries while leading Atlassian’s SaaS transformation, building the Cloud enterprise business, and scaling cloud platform teams. Before Atlassian, Anu held various leadership roles at Microsoft, and launched multiple products in Visual Studio. Anu is deeply passionate about growing diverse teams and making the world a better place with technology. She loves helping diverse founders on their mission of building bold and exciting companies. Anu served on the board of Outsystems, advise young and emerging companies and is part of the Operator Collective venture fund that pioneered the collective venture model to empower diverse founders.

Delphina
Hugo Bowne-Anderson is an independent data and AI consultant with extensive experience in the tech industry. He is the host of the industry podcast Vanishing Gradients, a podcast exploring developments in data science and AI. Previously, Hugo served as Head of Developer Relations at Outerbounds and held roles at Coiled and DataCamp, where his work in data science education reached over 3 million learners. He has taught at Yale University, Cold Spring Harbor Laboratory, and conferences like SciPy and PyCon, and is a passionate advocate for democratizing data skills and open-source tools.
Key Quotes
Key Takeaways
Agents are becoming teammates.
Atlassian frames agents as collaborators embedded in workflows, forcing the question of when agents should take action and when humans need to decide.
Most agents end up in the graveyard.
Thousands of agents have been built across product, marketing, engineering, and HR. Many don’t last, but experimentation at scale is what surfaces the ones that reshape work.
Adoption is the bottleneck.
Building agents is easy; making them discoverable, trustworthy, and useful in daily work is the real challenge. Proof-of-concepts only matter if teams actually adopt them.
Grounding determines reliability.
Enterprise agents must be anchored in verified data sources and show their reasoning. Without grounding, trust and adoption stall.
Humans still set the boundaries.
Some workflows can already be automated end-to-end, while others still require human judgment. Over time, more tasks will shift toward autonomy.
The frontier is proactive and ambient agents.
Anu predicts agents will move beyond prompts: surfacing next best actions and triggering background workflows without being asked.
The frontier is self-healing and orchestration.
Agents will increasingly be able to debug themselves and coordinate across workflows, moving closer to true autonomy.
The leap ahead is multiplayer collaboration.
Agents won’t just work one-to-one with humans, but alongside multiple people and other agents in real time.
Governance will decide who scales responsibly.Audit logs, transparency, permissions, and ownership attribution are becoming the foundations for enterprise adoption.
You can read the full transcript here.
Timestamps
00:00 Introduction to AI Agents in Teamwork
00:55 Atlassian's AI Platform and Its Impact
02:15 Interview with Anu: Exploring AI in Daily Work
05:31 The Role of Experimentation in AI Development
10:25 Building and Deploying AI Agents at Atlassian
22:37 Grounding and Reliability in AI Systems
26:26 Exploring Self-Healing Systems
27:15 AI Agents and Their Capabilities
28:44 Evaluating Business Value of AI
32:12 Industry-Specific AI Adoption
33:29 Challenges in AI Adoption
40:39 Proactive and Reactive AI Agents
42:41 Future of AI Agents and Security
49:35 The Next Leap in AI Agent Capabilities
50:36 Conclusion and Wrap-Up
Links From The Show
Transcript
In the spotlight: Our most popular episodes
Listen up: Our latest discussions
Hear the hottest takes on data science and AI.
Get the latest episodes in your inbox
Never miss an episode of High Signal by signing up for the Delphina newsletter.