Agentic AI is quickly becoming one of the biggest talking points in customer contact.
Not just because it can answer questions, summarize conversations, or help agents work faster, but because it can now begin to do things.
For years, contact centres have been looking at AI through the lens of extremely simple automation and assistance. But agentic AI changes the conversation from “What can it do?” to “What should it be allowed to do?”
For customer service directors, agentic AI will not simply be another tool you add to your contact centre, because it now raises a more strategic operating question:
How should operations be structured when AI can participate directly in the work?
In a recent conversation at Content Guru’s global headquarters, Xander Freeman checked in with Rob Mansfield, CTO at Content Guru, to discuss the next frontier of AI, where we’re heading next, and what that means for customer contact.
Start With the Process, Not the Technology
“A person has to look at the processes that you’re following – what you need to automate and what you need to make efficient – and only once you’ve cut out all of the waste in your process, then you automate, and then you use AI.”
Before automating anything, leaders need to ask what the contact centre is actually trying to improve – because one of the biggest mistakes organizations can make with AI is starting with the tool rather than the process.
For example, is the goal to use agentic AI to:
- Reduce handling time?
- Improve first contact resolution?
- Speed up after-call work?
- Support quality assurance?
- Help agents navigate complex policies?
- Create more consistent customer outcomes?
- Or something else?
AI in general is most valuable when it is attached to a well-defined task.
The danger is not that AI fails to improve your already pre-defined process, the danger is that it makes a flawed process look more efficient while the underlying problem remains untouched.
So, before contact centres give AI more work to do, leaders need to be clear that the work itself is still worth doing in the first place.
Decide Where AI Can Work in the Background
“It’s really important to see how agentic AI can work for our customers in the contact centre space. How we use it internally within our own company to see how we can improve performance, how we can drive automation, and get better outcomes.”
Agentic AI is not just about giving people better suggestions while they work, it’s about giving AI defined tasks that it can carry out initially before handing the result back to a human.
He gives the example of coding agents that can take an instruction, check work out into “a safe sandbox”, pull context from sources such as design documents and repositories, and complete tasks autonomously. The important detail is what happens next: “at the end, you’ve got to review those tasks.”
That is a useful way for Customer Service Directors to think about agentic AI in the contact centre. The real question is not whether AI can support a workflow, but where it can safely work without constant human prompting, and where the hand-off back to a person needs to happen.
For example, an AI workflow might prepare a case summary before an agent opens the record, check whether a follow-up has been completed, draft a response using the relevant policy, or flag a possible compliance issue for review.
In each case, the value comes from AI doing preparatory or repetitive work in the background, while the human stays responsible for checking the output and making the final judgement.
This is much more specific than treating AI as a general productivity tool. It means contact centre leaders need to design clear task boundaries: what the AI can access, what it can change, what it can only recommend, and what must always be reviewed.
That way your agentic AI will never end up as a “runaway co-worker” but rather as a controlled agent at your contact centre that you can build upon and improve with.
Give AI the Right Context Before You Give It More Responsibility
For contact centres, this may become one of the most valuable concepts in AI adoption. If the AI does not understand the context, it may produce an answer that looks polished but is completely wrong for the situation.
“Context engineering is being able to provide the right information to enable or empower the AI to do the job that you want it to do.”
Customer service decisions are rarely based on one piece of information, and so, if agentic AI is expected to support that work, it cannot be treated like a clever search box. It needs the right context before it can produce a useful output.
An agent may need to know what the customer has already been told, whether the case has been escalated before, which policy applies, whether the customer is vulnerable, what the complaint history looks like, what the next promised action was, and whether there are any regulatory or commercial risks attached to the response.
If the AI is working from incomplete, outdated or poorly structured information, the problem is not just that the answer may be weak – but even worse, the AI may act confidently on the wrong understanding of the customer, the policy or the process.
This means context engineering is a leadership issue, not an afterthought.
Before agentic AI can be trusted with more meaningful work, leaders need to know what information it is using, where that information comes from, how current it is, and whether it reflects how the contact centre actually operates.
Break Big AI Ambitions Into Smaller Tasks
“The technology’s now getting more sophisticated about breaking long running tasks down into sub-tasks – so they can choose the right model and the right tool for the job.”
Instead of asking AI to “handle customer service” as one huge broad task, leaders should always try to break the work into smaller, specific sub-tasks or use different agents.
For example:
- One AI workflow might identify customer intent
- Another might retrieve the right knowledge article
- Another might draft a response
- And another might recommend whether the case should be escalated.
That way, if something fails, leaders can better identify where the issue occurred rather than blaming “the AI” overall. It’s a far better controlled path to scale: start small, prove the value, manage the risk, then expand.
Governance Cannot Be an Afterthought
The more AI is allowed to act, the more important governance becomes. And if an AI tool makes a mistake, the business still owns the outcome.
“We can’t go to any of our customers and say, ‘Do you know what? This was an AI problem.’ We have to work alongside them to educate and advise on how to get the best from AI and have the necessary guardrails in place to mitigate issues.”
As Mansfield stated, this should be a guiding principle for every contact centre using agentic AI, particularly for contact centres handling sensitive, regulated, emergency or high-trust services.
Customer service directors need to work closely with IT, security, legal, compliance and operational teams to define what responsible AI use actually looks like in day-to-day service delivery.
The Future Belongs to AI-Enabled Teams
“AI won’t take your job, but someone using AI may take your job. Embrace the tools and learn to use the tools well.”
A common fear is that AI will replace people.
The more practical view is that people who know how to use AI well will become far more valuable.
That applies to agents, team leaders, quality managers, workforce planners and senior leaders. The best contact centres won’t simply reduce headcount and hope automation fills the gap – they’ll redesign roles around higher-value human work like judgement, empathy, escalation, coaching, problem-solving and customer trust.
AI can support throughput, speed and consistency, but humans will always be in charge of, and will still need to own responsibility, context and care.
Final Thoughts
Agentic AI isn’t just another tool; it actually signals a new stage in customer contact, where the best organizations will not only automate more, but also design smarter, more powerful and human-centred ways of working.
Now that AI can take action inside a workflow, leaders will have to think differently about how work is designed, decisions are made, and where human judgement adds the most value. AI supports the right work, humans remain accountable for the right decisions, and the whole operation gets easier.
The organizations that benefit most will not be the ones that rush towards the biggest or most impressive AI use case, either.
It will be the ones that do the harder, more valuable work first: understanding their processes, improving the quality of their data, defining clear governance, and training their people.
As Mansfield makes clear, AI still needs context, oversight and governance. The future of contact centres will not be AI completely replacing the operating model, but it will be redesigned around what AI and humans can do best together.
Author: Xander Freeman
Reviewed by: Megan Jones
Published On: 25th Jun 2026
Read more about - Expert Insights, Agentic AI, Artificial Intelligence (AI), Content Guru, Customer Experience (CX), Employee Experience (EX), Management Strategies, Service Strategy, Technology Enablement Strategy, Technology Roadmap
