What are ambient agents? Why AI that acts will redefine automation at work – Computerworld

What are ambient agents? Why AI that acts will redefine automation at work

Episode 355 What are ambient agents? Why AI that acts will redefine automation at work

Jan 12, 2026 47 mins

Artificial Intelligence Generative AI

Overview

Within the agentic AI space, there's a new generation of 'ambient agents' that are emerging - AI tools that observe, decide and then act across the enterprise without needing a prompt. In this episode of Today in Tech, host Keith Shaw interviews Bhavin Shah, CEO of Moveworks, about how these proactive agents will transform everything from IT support to HR compliance and software license management.

Learn how ambient agents differ from traditional chatbots and co-pilots, how they deliver real ROI through automation and cost savings, and why enterprises need to rethink privacy, trust and system architecture. Topics discussed include:

Transcript

Keith Shaw: Most people think of AI agents as tools you call on — you ask a question, they respond. But what happens when AI doesn’t wait for a prompt? New tools called ambient agents are emerging, capable of taking action before employees even realize help is needed.

How will that transform the workplace? We're diving into that on Today in Tech. Hi, everybody.

Welcome to Today in Tech. I'm Keith Shaw. Joining me on the show today is Bhavin Shah, CEO of Moveworks. Welcome to the show, Bhavin. Bhavin Shah: Thank you, Keith. Glad to be here. Keith: Let’s start with the basics.

Most people know what an AI agent is, but what makes an ambient agent different from a traditional agent or chatbot-style generative AI? Bhavin: With tools like ChatGPT, Copilot, and other assistants, users interact in a reactive way — you ask a question, get an answer, and move forward.

Ambient agents take that a step further by incorporating deeper automation capabilities into the architecture of large language models. Instead of waiting for a prompt, these agents take action based on conditions and reasoning. They don’t rely on user initiation, which is essential for scalable automation. Keith: Right.

Sometimes when someone talks about AI agents on a show like DEMO, it still feels like a prompt-response interaction. Even though they claim it’s an agent, it feels more like a chatbot. Maybe they’re just trying to keep a human in the loop? Bhavin: There’s a reason for that.

Think of an enterprise like a city, with each tall building representing a different SaaS tool. You might build agents on top of those tools, making the buildings taller. But the roads and bridges between them are often missing. So agents can’t easily move between systems on your behalf.

That’s why a lot of automation today still requires user direction. Ambient agents are designed to synthesize automation across those “buildings” and streamline tasks that cross systems. Keith: I've also seen terms like “autonomous agents,” “always-on agents,” and “active agents.” Are these interchangeable, or are there differences?

Bhavin: They build on each other. “Always-on” means the agent is always listening. But what matters is what it does after listening. We like “ambient agents” because it conveys both listening and acting — AI that behaves. That’s the real frontier: AI that acts with purpose and intelligence.

Keith: Why would a large enterprise need an ambient agent instead of a chatbot or Copilot-style tool? Bhavin: We're entering a new transformational phase in business — like we saw with the internet, SaaS, or the cloud. AI is driving a wholesale reinvention of automation.

Traditional systems are brittle — one small change in logic and they break. Ambient agents bring reasoning and flexibility. For example, a manufacturing system might get too hot and send an alert, but someone still has to respond.

An ambient agent could analyze that event, act autonomously, and fix the issue — no human needed.

Show more Show less

Take one example: A customer built an ambient agent that monitors employee schedules. If someone appears consistently overbooked, the agent recommends taking time off. It considers their calendar, compares them to peers, and checks how long it’s been since their last break. You couldn’t do that with traditional automation.

Keith: That's impressive. You're saying these agents can monitor systems and take contextual action. What makes this now possible? Is it just the rise of generative AI? Bhavin: Yes, it’s the large language models (LLMs) that make this possible.

The connectors and APIs still matter — they provide access to data — but now you can hand off that data to a reasoner. Instead of hardcoding every rule, you let the agent analyze and decide.

We've seen agents that monitor scrum plans and calendars to raise flags if a team is short on resources. Or agents that detect underused software and initiate deprovisioning to save costs.

Show more Show less

One customer created an agent that checks if someone hasn’t used a tool in a while. It evaluates usage patterns — seasonal or sporadic — and if it determines the tool isn’t needed, it suggests removing the license. If needed again, it can quickly restore access.

That’s a new level of cost-saving automation. Keith: That’s huge for ROI. Are these ambient agents delivering better ROI than traditional generative AI tools? Bhavin: ROI is complex. We break it into two categories: shallow ROI and deep ROI. Shallow ROI is productivity — writing memos or generating reports faster.

But that’s a soft gain. Deep ROI happens when you transform processes. Take recruiting: Hiring managers constantly ask about open requisitions. With an agent monitoring applicant tracking systems, it can proactively provide updates. That saves time and reduces friction.

Show more Show less

Ambient agents can also offer negative time to resolution — solving problems before they’re even raised. That changes everything. Keith: What about the fear that AI and agents are just there to replace jobs? Is that concern valid with ambient agents?

Bhavin: It’s a real concern, but today’s agents mostly handle repetitive tasks. We give them guardrails and specific scopes. That frees up humans to work on more strategic issues. Some companies tried to replace entire systems with AI and had to reverse course due to poor outcomes.

Eliminating headcount is complicated. Good automation complements people, not replaces them. Keith: And it gives employees more fulfilling work. Like not having to constantly check OKRs or request progress updates from HR. Bhavin: Exactly. One customer had 800 people per month asking HR partners for goal progress updates.

With ambient agents, those questions are handled automatically. HR professionals can focus on higher-value tasks. Keith: Some companies are training employees for more advanced work — level 3 or 4 support — because AI can now take over levels 0 and 1. Bhavin: That’s right.

Ambient agents improve the quality of work and life. Studies show many AI initiatives don’t impact P&L because companies overpromise on automation and underdeliver. The key is augmenting — not replacing — human effort.

Show more Show less

Keith: Give me a concrete example from IT support. Can you share a real case where an ambient agent caught something before a help desk did? Bhavin: Sure. Let’s say there are 10 VPN disconnects in a region. An ambient agent detects the pattern and alerts IT.

That’s hard to preprogram because the thresholds and triggers vary. Another example: an agent can guide employees to the right DRI (Directly Responsible Individual) to resolve an issue based on patterns of past resolutions.

Or it might notify a new manager about security trainings they need to complete — something that might otherwise fall through the cracks. Keith: That’s the kind of proactive “nudge” people respond to. Like a compliance reminder that doesn’t come from a no-reply email. Bhavin: Exactly.

People respond better when the reminder comes from an agent they already use. One customer saw compliance jump from 30% to 60% just by switching from email reminders to ambient agents. Keith: Wow. That’s a massive improvement. What about time-sensitive tasks like approvals? Bhavin: Great example.

One customer tracked software approvals. Before agents, it took VPs an average of 10 hours to approve. After introducing ambient agents, that time dropped to 9 minutes. It’s not just speed — it’s behavior change driven by context and helpful nudges. Keith: Right.

In my experience, some of the slowest processes are approvals for things like vacation or expenses — often buried in inboxes. Bhavin: That’s another perfect use case. The agent keeps track of requests, reminds managers, and ensures nothing falls through the cracks.

Keith: Are some industries more ready for ambient agents than others? Bhavin: It’s less about the industry and more about enterprise readiness — the systems in place and the mindset of leadership. We’ve seen success in healthcare, defense, retail, pharma, and tech.

It really comes down to whether the company is cloud-forward and already invested in automation. Keith: That makes sense. But I can hear IT folks asking: “Is this just another tool to bolt onto our existing stack?” How easy is integration? Bhavin: Ambient agents are challenging the current model.

Copilots today mostly enhance the platform they live in — like Salesforce or Microsoft. But ambient agents work across systems — Slack, Workday, Jira, ServiceNow. That requires more expansive integration but delivers far more value. Fortunately, platforms are becoming more extensible, which helps.

Keith: But that also raises privacy and security concerns. How do you manage trust and access? Bhavin: Privacy and autonomy are top of mind. Agents must respect the same enterprise rules already in place — no “god mode” access. You can’t have an agent guessing access permissions.

Access must be deterministic, based on roles, ACLs, and existing policies. Autonomy is earned over time. We advise starting with low-risk actions, like password resets. Only later should agents handle more sensitive tasks.

Keith: And if trust is built, users are more willing to give up control — especially if the agent reassures them it can restore access quickly. Bhavin: Exactly. The key is reducing friction.

If people know they can get a tool back instantly, they’re more willing to let go of unused licenses. That shift in mindset is crucial. Keith: So overall, ambient agents increase speed — and speed solves a lot of problems. Bhavin: Speed is the real game-changer.

When you act fast, you fix mistakes fast. That agility makes businesses more efficient and profitable, which fuels growth. It’s why AI is driving such a significant portion of GDP growth. Keith: Do you think ambient agents will become standardized in the next few years?

Bhavin: Not standardized in the strict IT sense. But communication protocols like MCP are becoming more common — helping LLMs access data. Some agent standards, like ATOA, are still developing. What matters is how easy it is to build useful, sophisticated agents. That’s where the competition is heating up.

Keith: What about the day-to-day experience for workers? Will people start managing agents instead of doing the work themselves? Bhavin: Potentially. It’s like conducting an orchestra of agents. But that requires maturity in systems and coordination. Not all roles will shift this way.

Many jobs involve complex decisions, collaboration, and judgment. That said, over time, more of the “work of work” will be automated. Keith: And humans don’t like change overnight. So it’ll happen gradually. Bhavin: Yes. It’s going to take time and intentional rollout.

But we’re at the beginning of a very exciting frontier. Keith: Bhavin, thank you for joining us today. This was a fascinating conversation. Bhavin: Thanks for having me, Keith.

Keith: That’s it for this episode of Today in Tech. Be sure to like the video, subscribe to the channel, and leave a comment if you’ve got thoughts on AI agents or automation. Thanks for watching!