News

By Arnab Bose, Chief Product Officer, Asana.

getty​The rapid rise of AI agents has prompted predictions about the “death of the user interface.” As agents become more capable, it’s easy to imagine a future where work happens entirely through prompts and conversations. ​

However, that misses the bigger transformation underway. Enterprise software isn’t losing its interface. Rather, it’s gaining a second participant: the people doing the work and the AI agents helping to execute it.​

The user interface remains critical for collaboration, evaluation and human judgment, while the machine-readable layer allows agents to execute. That’s very different from saying the user interface is dead. It’s more accurate to say its job, purpose and function are changing.

Rethinking What This Means For Product DesignFor decades, enterprise applications were designed almost exclusively for one audience: humans. The focus has been on screens, workflows and dashboards to help navigate information and make decisions.

That’s no longer sufficient. Those same products must now also support AI agents that need to understand work, participate in workflows and execute tasks alongside people.

Therein lies the challenge: Agents don’t navigate interfaces the way people do. They need structured context on who owns what, how work connects and what decisions have already been made, as well as durable ownership, clear permissions, approval checkpoints and machine-readable representations of work. These are the minimum requirements for an agent to operate safely and usefully inside an enterprise system.

What we previously considered great human user experiences now need to be great machine experiences, too. This doesn’t mean abandoning the fundamentals of good product design but raising the bar for what it looks like moving forward. Agentic AI makes the core principles more important while expanding the number of participants the product must design for.

The Interface Isn’t Dying—Its Job Is ChangingThe reality is that good interfaces become even more valuable alongside even more capable AI agents.

That’s because humans still need to direct work, inspect it, correct it, make judgment calls and align with each other. Taste, prioritization and accountability don’t disappear just because an agent can execute specific steps.

Knowledge workers recognize this, too: Our research found that 85% of surveyed workers “believe all AI agent output should be reviewed by a human.”

As agents take on more execution, the opportunity is there for humans to elevate the role they play—from producing every output themselves toward evaluating alternatives, applying taste and judgment before deciding what exactly moves forward. Importantly, those decisions rarely happen in isolation. Enterprise interfaces need to give people space to evaluate agent outputs, comparing recommendations and applying collective judgment.

The role of the interface evolves from helping people understand, supervise and improve work already in motion to helping them evaluate and guide work that agents are carrying out.

Chat Is The Entry Point, Not The ProductOne of the biggest misconceptions people have about AI agents is that they’ll change the way people interact with software by collapsing the interface for work into chat.

That’s not how serious work actually happens. People don’t run companies through prompts alone. They work through plans, approvals, timelines and shared context. That’s why even the leading AI platforms are moving beyond pure chat; they support artifacts, embedded apps, structured tools and hosted experiences.

Chat will matter, but it will matter as an entry point rather than the whole product. In enterprise software, the winning pattern will be natural language to initiate, structured software to coordinate and governed systems of record to track what happened.

Agents will make software more dynamic but not less structured.

Every Product Needs A Shared Understanding Of WorkLooking ahead, the biggest product challenge isn’t building more capable agents but ensuring every participant—human or AI—works from the same foundational understanding of the organization.

Our previous research has found that 73% of IT decision makers say AI initiatives fail or stall at least sometimes because AI lacks complete organizational context—including decisions, dependencies and prior work. Agents can’t execute effectively if the underlying system can’t tell them what matters, who owns what, what has already been decided and where work is blocked.

The reverse is just as important; people need visibility into the context informing an agent’s actions. If someone is expected to evaluate an output or make the final judgment call, they need confidence that the agent has considered the correct goals, dependencies and prior decisions in their output.

The solution is something similar to a “company brain”—a shared understanding of work that both humans and agents continuously contribute to and learn from. With this in place, every interaction begins from a stronger foundation than the last. Local intelligence becomes coordinated execution, and that shared layer will matter more than any individual chatbot or model.

Enterprise software, agents and people will all need to work off the same common ledger of work, ownership, status and intent. Without that, you get activity but not necessarily outcomes.

Enterprise Software’s Next Design ChallengeThe future of enterprise software isn’t about chat replacing interfaces or agents replacing people. Instead, it’s about designing products that give agents the context required to execute while giving humans the visibility to evaluate that work with other humans and steer it toward meaningful outcomes.

Agentic product design is evolving from building a clever assistant to building the right underlying system for human-agent teams. The companies that succeed in the future will build products where humans and agents operate from the same shared understanding of work, with agents able to execute effectively and people equipped to apply the taste, prioritization and judgment that remain distinctly human.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?