powering productive workplaces
Front page
NewsProductivity AI16m · 15:01 GMT · 5 min read

This Week: Enterprise AI Gets Serious About Action

Enterprise AI is moving beyond chat-based assistance and into the systems where employees, content and customer conversations are managed. Zendesk, Box and Five9 show vendors competing to make agents useful in real workflows - without asking businesses to rebuild everything first.

Workplace team reviewing automated service and content workflows on screens in a modern office
Workplace team reviewing automated service and content workflows on screens in a modern office

Enterprise AI is moving beyond chat-based assistance and into the systems where employees, content and customer conversations are managed. Zendesk, Box and Five9 show vendors competing to make agents useful in real workflows - without asking businesses to rebuild everything first.

For a while, workplace AI has been the colleague who can draft, summarise and occasionally produce an alarming number of bullet points. This week’s news suggests it is being offered a more consequential job: resolving employee requests, moving content through business processes and analysing customer conversations.

The catch is that few enterprises are ready to rip out the platforms where that work already happens. The vendors getting attention are not merely promising smarter agents. They are promising to bring them into familiar systems, with enough control to make the idea plausible outside a demo.

TL;DR

  • Zendesk is turning employee support into an AI-agent use case across Teams, Slack and internal knowledge sources.

  • Box is putting agentic automation closer to the content and approvals that underpin everyday business work.

  • Five9’s Level AI acquisition shows the race to add AI capability without forcing every customer into a platform migration.

Zendesk Wants AI Agents to Resolve the Internal Admin Drain

Zendesk’s employee service AI agents started rolling out on 3 November, bringing AI-powered support to the help centre, Slack and Microsoft Teams. The agents can connect to more than 20 knowledge sources, including SharePoint, Google Drive, Confluence, Notion and Box, while respecting the permissions attached to that content.

That makes this a more substantial move than a chatbot refresh. Internal service teams deal with the low-glamour friction that quietly eats into the working week: access requests, policy questions, onboarding tasks, IT troubleshooting and finance queries. Zendesk’s agents can answer questions, escalate cases to a person when needed, and trigger advanced action flows designed to complete multi-system requests.

For employees, the value is straightforward: less time chasing answers across portals and inboxes. For IT and HR teams, the test is whether the automation reduces repetitive handling without making the exception cases harder to untangle.

Zendesk is also putting a visible price on automated outcomes. Enterprise and Enterprise Plus plans include 15 automated resolutions per employee-service agent each month, after which additional resolutions cost $0.99 each. That will push customers to be clear about which requests are worth automating, rather than treating every interaction as a suitable job for an agent.

The Lesson: An Agentic Workflow Needs More Than a Clever Prompt

Box made a similar move on 5 November with the launch of Box Architect, a tool designed to build, manage and optimise agentic workflows across enterprise content and business processes. The beta is due for Enterprise Advanced customers in Q4.

The interesting part is not simply that Box wants agents to work with documents and data. It is that the company is building the control layer into the pitch. Its announcement includes MCP guardrails for what agent tools can do, classification-based access policies, threshold alerts and human review at key points in an automated workflow.

Aaron Levie, CEO of Box says:

“The real opportunity with AI is not just helping individuals work faster, but rethinking how entire business processes get done.”

That matters because enterprise content is rarely just content. It is the contract awaiting approval, the customer record containing sensitive information, the proposal that reveals a sales strategy, or the policy document that determines what a frontline employee can tell a customer. An agent that can move, create or share material needs clear boundaries before it becomes useful at scale.

As UC Today found in its look at which AI productivity workflows actually deliver ROI, the best candidates tend to be repeatable, high-volume processes with an obvious bottleneck. Box’s approach gives leaders a practical next step: map the workflow first, then decide where automation can take action and where human judgement still belongs.

The Key Question: Can Enterprises Add AI Without Replacing Their Stack?

Five9’s planned acquisition of Level AI, announced on 5 November, points to a pragmatic answer: they may not need to.

The deal values Level AI at roughly $325 million and will add its AI agents, agent assist, quality management, coaching and conversation-intelligence capabilities to Five9’s Genius CX platform. But Five9 also plans to offer the technology as a standalone product for organisations using an on-premises contact centre or another cloud provider.

That is a telling strategy. Most enterprises do not have the appetite, budget or internal patience for a wholesale technology migration just to access a new AI capability. They want useful automation where they are, especially in customer operations where service quality, compliance and workforce management are tied tightly to the existing platform.

Level AI’s technology is built around analysing customer interactions and using that information to improve future conversations. The opportunity is not simply to automate more contacts. It is to give human agents better context, surface patterns in customer demand and direct automation towards routine work while keeping complex or sensitive interactions with people.

Ashish Nagar, Level AI Founder and CEO said:

“We built Level AI so every customer conversation makes the next one better, for customers and for the people who serve them.”

The risk, as ever, is that a standalone AI layer becomes yet another disconnected tool. Five9’s challenge will be proving that Level AI can create value before a customer commits to the wider platform - while still making the eventual integrated proposition worth considering.

Zendesk, Box and Five9 are tackling different parts of the enterprise, but their strategy is remarkably consistent. AI agents are being positioned less as a separate destination and more as a working layer inside the systems companies already rely on. The vendors that succeed will not be those with the loudest autonomy claims. They will be the ones that make automation fit the way work actually gets done.

rate this story
helps rank stories across uc today
Join the conversation

Does this reflect your experience?

opening the room…