AI agent, bot or copilot? Choosing the right AI for each contact centre journey

The labels vary by vendor, so we use the terms Genesys publishes. A scripted bot follows a path designed in Architect. A Genesys Virtual Agent is a conversational bot powered by Genesys AI, available in voice and digital bot flows, which understands natural language and can answer from knowledge; AI Guides, written as natural language instructions in AI Studio, are charged as Virtual Agent interactions. An agentic virtual agent pursues a goal with tools, knowledge and guardrails. Copilots do not talk to customers at all: Agent Copilot supports the agent during and after the conversation, and Virtual Supervisor brings AI to supervisors and quality teams.

The four also differ in what they need from you. A scripted flow needs a clear design and little else. A virtual agent needs well-maintained knowledge and good examples of how customers phrase things. An agentic virtual agent needs tools that are safe to expose, written guardrails and a testing regime that covers the awkward cases as well as the common ones. Copilots need agreed summary formats, current knowledge and supervisors who use what they produce. The further along the scale, the more of the work sits in integration, governance and testing rather than in the AI itself, and the more important it is to plan that work from the start.

We start every AI decision from the journey, not the product. For each contact reason we ask the same handful of questions, and the answers usually point clearly to one kind of AI, or to a person. Where they point to more than one, the simpler option generally wins: it is cheaper to run, easier to test and easier to explain to a regulator or a customer.

Copilots are the least visible kind of contact centre AI and often the quickest to show value, because they improve conversations customers already have with people. Genesys describes Agent Copilot as determining customer intent and providing relevant next best actions during the interaction, then producing a summary with the reason for contact and its resolution and suggesting wrap-up codes. Virtual Supervisor brings AI Scoring, AI Translate and AI Summary and Insights to supervisors and quality teams, so more interactions can be reviewed consistently.

Copilots do not remove the need for good knowledge, coaching and quality management; they depend on them. Next best actions are only as good as the knowledge and actions behind them, and summaries need agreed formats and a check that agents read and correct them. Because Genesys charges Agent Copilot per user each month, it also shares the organisation's free token allowance with any bots, which is worth modelling before launch.

The kinds are not alternatives so much as stages. A typical voice journey might begin with a scripted flow that identifies and verifies the caller securely, pass to an agentic virtual agent to resolve a delivery problem across order and courier systems, and transfer to an agent when the customer asks for a refund outside policy, with Agent Copilot presenting what has already been done and suggesting the next best action. Afterwards, AI Summary and Insights records the outcome and AI Scoring helps the quality team review it.

The joins are where journeys succeed or fail. Context must travel with the customer at every hand-off, so no one is asked to repeat themselves; outcome codes and reporting must be consistent, so analysts can see where AI resolved a contact and where it passed it on; and every stage needs a defined route to a person. We design and test these hand-offs as carefully as the AI itself, with real calls placed from outside the platform, the way customers call.

1. Scripted flow: greet, identify and verify securely
2. Virtual agent or agentic agent: understand and resolve what it can
3. Hand-off: transfer with context when a person is needed or asked for
4. Agent Copilot: next best actions, knowledge and a summary for the agent
5. Virtual Supervisor: scoring, summaries and insight for quality and improvement

Each kind is metered differently on Genesys Cloud CX: scripted voice bots by the minute, virtual agents and agentic virtual agents per interaction, and copilots mostly per user each month, all in AI Experience tokens from one organisation-wide allowance. That means the cheapest option depends on call length, volume and how many agents use copilots, not on the headline rate. Our AI Experience tokens guide explains the rates, and our voice AI cost calculator compares native and third-party voice AI per resolved call at published list prices.

QVCCS is a Genesys referral partner, not a reseller, and we take no commission from any AI vendor, so our recommendation is simply the kind of AI that resolves each journey most reliably for the least complexity. We map contact reasons to the four kinds, record the reasons in the design decision log, and prove each choice with real calls before it reaches customers.