AI Agents at Work

AI Agents at Work: How They Work, Where They Fit, How to Run Them

A practical workshop for business leaders on how agents work, where they fit, and how to manage their results.

Participants leave with a first agent operations map for one of their own workflows. You will follow a business task through its instructions, context, tools, decisions, and output, then map that workflow into inspectable increments of work with clear context, permissions, handoffs, and human responsibility.

Date
Thursday, December 17, 2026
Time
10:00 AM to 3:00 PM ET
Location
In person, Cambridge

This in-person session is a planning workshop. No coding background is required.

A blue AI Agent character holding a task checklist and pointing forward

AI Agent: works toward a defined goal through permitted steps.

Agents need more than access to AI.

AI agents are opening new opportunities to improve productivity, streamline workflows, and carry out complex tasks. Building one that performs reliably takes more than powerful technology. An agent has to understand the work it is asked to do, including objectives, roles, processes, rules, and knowledge.

Without that context, even capable models produce results that vary from one run to the next. The gap between a promising experiment and a dependable agent is usually missing context.

Just getting started
  • Agent ideas are coming from every team.
  • Few people can name the first workflow worth building.
  • A problem-first approach turns those ideas into a short, ranked list.
Experiments are running
  • Agents perform well in demos but vary in daily operations.
  • Outputs drift because the agent lacks roles, rules, and process knowledge.
  • Organizational context gives agents the same grounding your people rely on.

Agents perform when the work, the rules, and the people behind the work are defined. This workshop shows you how to define them for one workflow.

The participant promise

Understand the agent. Map the workflow. Manage the results.

Understand what happens inside an AI agent, then map one business workflow into inspectable increments of work. Each increment comes with clear context, permissions, handoffs, and human responsibility.

You leave with a method you can reuse on the next workflow, along with a planning map for the first one.

Where agents fit

Know which type of AI fits each job.

AI assistants

Respond to a person in the flow of their work. The person directs each step and reviews every output.

Best for: drafting, research, and analysis

AI automations

Run a defined workflow when a trigger occurs. Each step follows fixed rules, so the result is predictable.

Best for: routing, data handling, and repeatable approvals

Autonomous AI agents

Work toward a goal by choosing steps, using tools, and checking progress. Guardrails and approval points set the limits.

Best for: multi-step work that changes from case to case

Three questions organize the workshop

Three questions. One working plan.

Question one

How does an AI agent work?

  • Follow a business task through its instructions, context, tools, decisions, and output.
  • Make the process visible with a worked example.
Question two

What work should we give an agent?

  • Choose a useful increment and define its inputs and completion criteria.
  • Decide which steps need AI judgment.
Question three

How do we run agentic operations?

  • Establish ownership, action permissions, and checks.
  • Set failure handling and handoffs between people and agents.

Understand the work. Scope the increment. Run it with clear ownership.

The demonstration

Follow one client-meeting brief from source to check.

Source information

The notes, records, and account details the brief draws on, shown before the agent starts.

Shows what the agent is given

The assignment

The objective, the context the agent receives, and the format the finished brief must take.

Shows what the agent is asked to do

The brief and its checks

The resulting brief, the checks needed before anyone relies on it, and the next step in the workflow.

Shows where a person takes over

The same business process runs through the whole session. Each completed increment connects to the next step in the workflow.

The tangible takeaway

Leave with an agent operations map for one workflow.

Each participant completes a first map that runs from business outcome to next handoff.

01Business outcome
02Inspectable increment
03Required context
04Permitted actions
05Verification
06Next handoff

This is a planning artifact. A working deployed agent would require a different workshop scope.

Shared agent infrastructure

Set up the foundation once. Reuse it across every agent.

Agents built one at a time create one-off solutions. Shared infrastructure lets each new agent draw on the same organizational context, tools, and controls, so every team builds on the work of the last.

  • Shared context blocks for roles, processes, and rules
  • Common tool connections and access permissions
  • One set of guardrails and approval points
  • Measures that show how each agent performs
An orange person character in a talking pose, presenting an idea

Bring the problem. The agent comes after.

Who should attend

Built for the people who will work with agents.

  • Business leaders deciding where agents fit
  • Operations and functional managers who own a workflow
  • AI transformation and innovation teams
  • IT and technology leaders who will support agent operations
  • Professionals who will review and own agent results in their roles

No coding or technical background is required. Bring one business workflow you want to examine. You will map it during the session.

Reserve your seat

Apply for AI Agents at Work

Complete the form to reserve your place at the in-person workshop in Cambridge on Thursday, December 17, 2026.

Get started

Bring one workflow. Leave with an agent operations map.

Join The Kendall Project in Cambridge on Thursday, December 17 for AI Agents at Work, a practical workshop on how agents work, where they fit, and how to run them.