Converge Methodology artefact
One scenario, four levels of AI: from copilot to agentic system
The words "copilot", "assistant" and "agent" are used interchangeably in most AI conversations, which is why most AI conversations go in circles. We show leadership teams one scenario from their own business at each of four levels, so the shift from AI that assists people to AI that owns outcomes becomes concrete.
Human-led to AI-owned
Each level hands more of the work to the system, within boundaries the business sets.
Copilot — AI suggests, a person decides
A language model with retrieval surfaces context and drafts; a person approves each step. "Help me write this email." Recommends configurations, flags compliance risks, drafts the quote.
Assistant — AI executes, a person instructs
A model with function calling runs defined tasks on demand. "Draft the email, segment the audience, generate variants." Generates the export quote, applies the pricing rules, produces the document.
Agent — AI plans and executes within boundaries
A planning loop with access to tools, iterating until the goal is met. "Increase lead conversion by 15%; I'll decide what to test." Plans the quote workflow, resolves pricing and compliance, delivers an approved quote.
Agentic system — AI delivers governed outcomes
Specialised agents with shared state, orchestrated under policy controls, with escalation paths to people. "Increase qualified lead conversion by 15% this quarter within the existing budget; alert me if the risk to target exceeds 10%." Coordinates specialist agents and delivers a margin-optimised proposal.
The same scenario at every level
Abstract definitions do not change decisions. A scenario the leadership team recognises does. In a manufacturing business we use a customer requesting a complex drive configuration and quote; in a publisher, a subscription campaign; in an agency, a support ticket. The same work is shown at each level, and the room can see exactly what changes: who decides, what the system is allowed to do, what it needs access to, and what happens when it is unsure.
Two things usually follow. The first is relief: most of the value in the next twelve months is at the copilot and assistant levels, which are well understood and low risk. The second is clarity about what the agent and agentic levels actually require, which is governed access to systems and data, an evaluation harness, and confidence-based escalation to people, and which is why they are built, not bought.
From AI that assists people to AI that owns outcomes
The four levels are a progression, but not a ladder every use case must climb. A copilot is the right end state for work where judgement is the point. An agentic system is right where the outcome is measurable, the boundaries are clear and the cost of a wrong decision is bounded by escalation. Deciding which level each use case should reach, and stopping there, is as important as the technology.
This is also where governance enters. Under the EU AI Act's risk-based approach, a customer-facing assistant carries a transparency duty; an agent making decisions that affect people carries human-oversight requirements. Knowing which level you are building tells you which obligations apply.
About the four levels
Which level should we start at?
Almost always copilot and assistant, across the whole business, on governed enterprise tools. Agents come next, for one or two bounded workflows with clear measures. Agentic systems are a programme, not a pilot.
Is an agent the same as a chatbot?
No. A chatbot is usually a copilot or assistant with a conversational interface. An agent plans, uses tools, iterates and decides when to hand over. The Agency51 support agent is an agent: it reads code and documentation, drafts a reply and escalates when its confidence is low.
How do you keep an agentic system safe?
Policy controls on what each agent may do, shared state that is auditable, confidence thresholds that trigger escalation to a person, an evaluation set run on every change, and monitoring for drift. We build those in from the first scope, not after.
See your own scenario at four levels
The exercise takes forty minutes in a leadership session and changes the conversation.