The promise of AI in support is that it takes the dull work and leaves people the interesting problems. A CNBC report published on October 24 talked to agents who have lived with that promise, and the picture is more mixed. Some say they finally have time to look after each customer. Others say they spend their days “constantly fighting fires”. For anyone designing a human-plus-AI support team, the details are worth reading carefully.
What Agents Reported
CNBC’s Lindsay Dodgson spoke with contact-center workers, executives and a researcher. A few points stand out:
- Less admin, more attention. Kaylee Harris, who works in customer service for a credit card company, said her days used to be filled with calls, note-taking and lengthy disclosures. With AI handling some of that, she can focus more on each customer, though system glitches sometimes create extra work.
- Errors that add work. A telecom worker in Cardiff, identified only as Shannon, said AI-generated notes at her current job are often inaccurate, so she writes her own. At a previous job, AI agents booked appointments but handed cases to humans when a requested date wasn’t available, which led to heavy workloads and burnout.
- Workloads rising. MaxContact’s 2025/26 UK Contact Centre KPI Benchmarking report found that more than half of surveyed decision-makers said agent workloads had increased, up from around 40% in 2024. The survey does not establish what is driving the rise.
The Hidden Cost of Automating the Easy Contacts
Jones’s point deserves attention because it is easy to miss on a dashboard. If AI resolves the simple contacts, the average difficulty of what reaches a person goes up, even if total volume stays the same. Each human conversation becomes harder, longer and more emotionally demanding. Jones also notes that workers may lose the chance to build the experience needed for harder work, because the easy contacts were where newcomers learned.
This does not mean automation is a mistake. It means the human role changes and needs to be designed deliberately rather than inherited as whatever is left over.
Four Design Changes That Help
1. Revisit handle-time targets
ArvatoConnect CEO Debra Maxwell told CNBC that average handling times may need to rise, and that agents should be judged on whether they understood the customer’s needs and reached the right resolution: “Not whether they followed a script or hit an arbitrary time target.” If your human contacts are now the complex ones, a handle-time target set in the old world will push agents to rush exactly the conversations that need care.
2. Fix the handoff
Shannon’s story shows the failure mode: automation that collects information, hits a snag and passes the customer to a person without context. When a conversation reaches a human, they should see the customer’s goal, what was already tried and what the AI could not do. That cuts both customer repetition and the agent’s workload. Our escalation matrix guide covers deciding in advance what should hand off and to whom.
3. Check the AI’s output before agents depend on it
If AI-written notes or summaries are often wrong, agents will rewrite them, and the time saving disappears. Sample the outputs agents rely on, share error rates with the team and give agents a quick way to flag bad ones. Treat their corrections as training data.
4. Build in recovery and growth
If the easy contacts are gone, create other ways to learn and recover: structured coaching on complex cases, rotating time on lighter work such as knowledge-base improvement, and deliberate breaks after difficult conversations.
Questions to Ask Your Own Team
| Question | Why it matters |
|---|---|
| Has the share of difficult contacts reaching humans gone up? | Shows whether the job has quietly changed |
| Are handle-time and quality targets still sensible for those contacts? | Prevents rushing the hardest cases |
| How often do agents edit or discard AI-generated notes? | Reveals hidden rework |
| How often do customers repeat themselves after a handoff? | Measures the quality of handoffs |
| How do new agents build experience now? | Protects the pipeline of skilled people |
We covered how roles are shifting in The New Shape of the Support Team. The front-line view is a useful reality check on that picture.
The Takeaway
AI can make support work better for people, and some of the agents CNBC interviewed say it already has. But the benefit is not automatic. The same automation that removes tedious tasks can concentrate the hardest conversations on fewer shoulders. Ask your agents what their days look like now, and design the human role on purpose.