Agents & Tokens: What Are They, and Why Should You Care?
AI agents are becoming the next big thing in legal tech. But what actually makes something an agent, what are the tokens everyone keeps mentioning, and why can one AI task cost almost nothing while another gets expensive fast? Pieter Buteneers, Co-Founder and CTO of Emma Legal, breaks down agents, tokens and agentic workflows in plain language.
What you'll walk away with
"Agent" is doing a lot of work in legal tech marketing right now. It gets used for a chatbot with a search box, and for a system that reads a full data room, decides what to look at next, and works through it step by step.
Those are very different things. They also cost very different amounts of money.
Pieter takes the lid off. How an agent actually reasons, what tokens are, and why a workflow that looks like one question can quietly become a hundred. By the end you'll be able to read a vendor's pricing and know what you are looking at.
What we'll cover
- What actually makes something an agent.
A model that answers a question is not an agent. A model that plans, picks a tool, checks the result and tries again is. We'll draw that line clearly, and show where most legal tools sit today. - Tokens, explained once and properly.
What a token is, how a model counts them, and why the length of your document matters far more than the length of your question. - Under the hood of an agentic workflow.
Reasoning, tool use, document retrieval, multi-step execution. Pieter walks through a real legal task from prompt to output, and shows what happens at every step in between. - Why the cost varies so wildly.
The same question can cost a fraction of a cent or several euros, depending on how it is built. We'll show exactly where the multiplication happens, and why agentic workflows are where budgets get surprised. - How to read the numbers vendors quote you.
Per seat, per document, per token, per deal. What each pricing model rewards, what it quietly hides, and which questions get you a straight answer.
Who this is for
M&A lawyers, legal innovation and knowledge leads, and anyone evaluating AI tools or defending their cost internally. No AI hype and no computer science degree required. Pieter explains this the way he would explain it to a partner, not to a developer.
About the speaker
Pieter Buteneers is Co-Founder and CTO of Emma Legal, with a PhD in AI and 18+ years of applied machine learning experience. He founded Chatlayer.ai before building Emma. He is a regular speaker at MLcon, one of the leading European AI conferences, and runs workshops for DPOs on setting up an AI security framework. He speaks plainly, thinks in systems, and has no patience for AI hype that doesn't translate into better legal work.