How it works
An AI agent is software that can take actions through tools, not only generate text. In the workplace, the tools are the systems that hold space and people data: booking, floor plans, the directory, calendars and occupancy records. The agent reads from them, proposes an action and, within its permissions, carries it out.
Agents work at two levels. For employees, an assistant books a desk near the team, finds a room or invites a visitor from a chat. For workplace teams, a planning agent drafts a stack plan or seating layout from headcount and occupancy data, compares options and hands the result to a person to approve.
Open standards such as the Model Context Protocol let the AI assistants people already use connect to workplace systems. Governance does the rest: agents act with each user's existing permissions, keep a record of what they did and leave consequential decisions, such as moving a team, to people.
Agents are only as good as the data and decision history they can see. Clean occupancy data, current floor plans and a record of past decisions matter more than the model.
How to measure it
- List repetitive tasks by volume (bookings, visitor invites, seat assignments, plan revisions) and start where the volume is.
- Check that agents act with each user's existing permissions and that every action is logged.
- Measure time saved and errors avoided against the manual process.
- Keep a human approval step for moves, closures and anything that affects many people.
Examples
- An employee asks their AI assistant for a desk near the design team on Tuesday; the agent books one within the booking rules.
- A workplace lead asks a planning agent to fit a growing team onto two floors; the agent drafts a stack plan, the lead adjusts it and publishes it.
Benchmarks
In JLL's Global Real Estate Technology Survey, only 5% of corporate real estate occupiers piloting AI had achieved all their program goals.
JLL, Global Real Estate Technology Survey news release
Common mistakes
- Letting agents act with broader access than the person using them.
- Automating decisions that need human judgment, such as team moves, without an approval step.
- Pointing agents at stale floor plans or booking data and trusting the output.
- Calling a chatbot that only answers questions an agent.
How Kadence helps
Kadence AI books desks and rooms by chat in Microsoft Teams and Slack. The Kadence MCP server (early access, enabled by an admin) lets Claude, ChatGPT or Copilot book desks, rooms, parking and visitors and check people in, with per-user sign-in and each person's existing permissions. In SpaceOps, the AI planning agent builds or changes stack and space plans from plain-language requests, and a person approves a plan before it is published to WorkOps.
Kadence MCP · SpaceOps · Kadence AI
Frequently asked questions
What is agentic AI in the workplace?
Agentic AI is AI that can take actions through connected tools, not only answer questions. In a workplace it books space, invites visitors, drafts seating or stack plans and prepares changes, within the permissions and rules the organization sets.
How is an AI agent different from a chatbot?
A chatbot answers questions. An agent can also act: it calls tools, such as a booking system or a floor plan editor, to complete a task, and it can plan several steps toward a goal. That makes permissions and approval steps essential.
Is it safe to let AI agents act on workplace systems?
It can be, with controls: agents that act only with the permissions of the person using them, per-user sign-in, a log of every action and a human approval step for consequential changes such as moves and closures.
What does an agentic workplace need to work?
Accurate, connected data (occupancy, floor plans, directory and bookings), a standard way for agents to reach it, such as the Model Context Protocol, clear permissions, and a record of past decisions the agents can reason over.
Related terms
- MCP (Model Context Protocol) An open standard that lets AI assistants use tools and data in other systems, such as booking a desk, without custom integrations.
- AI space planning Using AI to draft and test stack plans, seating plans and scenarios from headcount, occupancy data and floor plans.
- AI workplace assistant A chat-based assistant that books desks and rooms and handles workplace tasks from a plain-language request.
- Office move automation with AI Software and AI that recommend seats, assign people in bulk, notify them and update records for workplace moves.
- Workplace operating system One platform that connects daily office use (booking, check-in, visitors) with space planning, on one set of data.
Sources
- Real estate’s AI reality check: 90% of companies piloting, only 5% achieved all AI goals, JLL
- Introducing the Model Context Protocol, Anthropic
Kadence editorial team · Last reviewed