The time between an agent’s first day and their first productive shift has long been treated as a fixed cost. In reality, it’s a readout on how well an operation has captured what its best agents do and made it available to everyone else. AI has improved that transfer of knowledge, and TELUS Digital measured the effect directly. In one deployment, role-play productivity improved 75-85%, with agents completing two to three scenarios in 45 minutes rather than the hours individual role plays once required, and customer satisfaction rose 18%.
Ramp time that still takes half a year usually points to the training and assist design behind it. Erin Walker, Global Vice President, CX AI, Business and Delivery at TELUS Digital, says, “The mistake most organizations make is buying good tools separately and expecting performance to follow. It does not. Agents improve when training, assist, quality and coaching operate as one connected system, because that is the only way the operation learns from its own best work.”
A Contact Center Before AI vs. After AI
The shift is visible when you put the two versions of the job side by side. Each moment gained an AI layer, and none of them lost the agent.
| Category | Before AI | AI’s role | Human’s role |
| Quality monitoring | Supervisors reviewed a small sample of calls and chats | Analyzes every interaction, flagging skill gaps and compliance risks automatically | Decides what to coach and how |
| Agent training | Agents learned complex and judgment-heavy scenarios, largely through live trial and error | Runs realistic simulations against AI personas before a live customer is involved | Still owns every real, emotionally complex conversation |
| Live support | Agents searched multiple systems mid-call for account history and product details | Surfaces knowledge and customer context in real time, suggests next-best actions | Remains the only voice the customer actually hears |
A second TELUS Digital deployment points in the same direction. When a payments processor used Fuel iX™ Agent Trainer to onboard 90 new agents, customer satisfaction scores rose 16% across all channels and 29% in chat specifically. Additionally, 80% of agents who started as underperformers finished the program as top performers.
Why Simulated Practice Changes the Ramp Curve
The conventional new-agent sequence puts classroom instruction first, shadowing second, and a live customer third. The first genuinely hard conversation, usually an escalation or a billing dispute that requires judgment, happens with a real customer on the line.
AI-supported training changes that order. Agents rehearse against AI personas that behave like difficult customers, working through the refund that does not qualify or the caller who has already been transferred twice, before any of it is real. The mistakes happen in rehearsal, allowing a new agent to enter a difficult live call with the experiences required to handle it.
Can AI Handle Customer Service Queries All on Its Own?
Researchers at Bar-Ilan University tested a hybrid system in which every incoming question goes to a virtual agent first and anything the agent can’t resolve routes to a human operator. Trained only on transcripts of human operators answering questions, the agent learned to handle 70% of incoming questions.
The design detail that made it work was the confidence threshold. The agent gets one attempt at each question, and the researchers set the bar deliberately high, sending questions to a human whenever the agent wasn’t confident it understood the customer’s intent. Passing a question along only costs a few seconds, while answering it wrong could mean losing the customer’s trust.
The results favored the hybrid setup on both sides. A single operator assisted by virtual agents served 18 clients against 11 without, with comparable wait times for the human operator. Customers in the hybrid condition received answers substantially faster and rated their satisfaction higher.
How Do AI Copilots Help Agents During Live Calls?
Fuel iX™ Agent Assist gives TELUS Digital’s agents real-time guidance during live interactions so they no longer have to search multiple systems mid-call or ask customers to repeat information already on file.
“We are not trying to take the human out of the conversation. We are trying to make sure they have everything they need, exactly when they need it, and that a person, not a system, stays accountable for the moments that matter most, the ones that require judgment. That is what moves performance and improves your business outcomes,” Walker said.
Ramp Data Is the Audit an Operator Can Run Before Buying Anything
Ramp time is a useful diagnostic because it already exists in every workforce management system and is hard to game. Four readings make it usable.
- Measure time to first unassisted resolution rather than time to graduation from training. Graduation dates reflect a curriculum calendar, while first unassisted resolution reflects capability.
- Don’t just look at the average. If most agents ramp in two months but some take five, the average hides the slow group, revealing a transfer gap.
- Check whether simulation scenarios were built from real transcripts or written from imagination. Scenarios that do not match the interaction mix produce agents rehearsed for the wrong job.
- Confirm whether the assist system records which recommendations agents accept, modify or ignore because a system that doesn’t capture that can’t improve.
The Job Changed, the Agent Did Not
Agents using AI now train against simulations before they train against customers, get account history pushed to them during calls instead of hunting for it, and have every interaction reviewed rather than a sampled few. None of that removed anyone from the floor.
An agent still decides what to do with a frustrated customer and owns how the call ends. AI made the work around that decision easier to do.
Frequently Asked Questions
Question: How does AI help new contact center agents ramp faster?
Answer: It supplies the patterns a new agent hasn’t built yet. Real-time guidance surfaces what experienced agents do at the moment it’s needed, and simulation lets new hires rehearse difficult conversations before facing a live customer. This helps the least experienced agents catch up to the more experienced ones.
Question: How are companies using AI to make contact center agents better?
Answer: Three ways: agents rehearse against AI personas before their first live escalation and during calls, assist tools surface account context and next-best actions. In quality review, every interaction can be scored instead of a small sample, changing what supervisors coach. The agent still handles the conversation.
Question: Should we buy point solutions or work with one partner across the agent lifecycle?
Answer: What matters is whether the tools share data, not how many vendors supply them. When simulation and live assist run on separate records, agents rehearse one version of the job and work another. A single partner guarantees shared data, and integrated tools built on a common record achieve the same thing.
Question: Why do AI quality monitoring insights fail to change agent behavior?
Answer: Most programs stop at the score. Analyzing every interaction solved coverage, but a number in a dashboard never reaches the coaching conversation that would act on it. Behavior changes when the insight routes to a specific agent and feeds the same simulation and live guidance that agent already receives.
TL;DR
Ramp time, or how long a new agent takes to reach full productivity, is the clearest readout on whether a contact center’s AI is working. TELUS Digital’s deployments show the curve can move: training that once took hours runs in minutes, new hires rehearse against AI personas before facing live customers, and quality review now covers every interaction. Through all of it, agents keep the conversations that need judgment. When a ramp stays long, the training and assist design is usually where to look.
