June 24, 2026
BPO Viewpoint – The Agentic AI Jobs Truth Vendors Won’t Tell You
At CCW Las Vegas, the vendors sold one story: humans and machines, side by side, nobody worse off. The outsourcers who actually run the agents told a sharper, more divided, and more honest one. CXM reports the BPO view from the floor.
Walk the floor at Customer Contact Week, and the same reassuring line follows you from conversation to conversation: humans and AI, together, everybody wins. It is a tidy story, and a popular one. It is also not quite the story being told by the people who run contact centres for a living.
The outsourcers are worth listening to precisely because they have the least room to posture. A platform vendor needs you to believe agentic AI is transformative; a sceptic needs you to believe it is hype. The business process outsourcer (BPO) sits at the centre of that argument, carrying both the headcount that AI might displace and the technology clients now expect them to deliver. So when four of them broke from the conference script this week — on jobs, on cost, on what customers actually want — the disagreements between them turned out to be more revealing than any keynote.
The short version
- Agentic AI does not rescue a struggling contact centre. It magnifies whatever is already broken underneath.
- The jobs question splits the industry, and the most candid answer is the one most vendors won’t say aloud.
- Nobody yet knows what agentic AI costs at full production volume.
- Cutting costs is the boardroom’s real goal, whatever the conference language pretends.
- Customers want to be told when they’re talking to a machine, and the youngest resist AI hardest.
Does agentic AI fix a struggling contact centre?
No — and according to Molly Moore, President and COO of Liveops, it often does the opposite. Her line was the bluntest of the day: AI exposes a company’s operational weaknesses and then makes them worse. Bolt an agent onto fragmented data, tangled processes and unclear ownership, and you don’t get transformation. You get the same mess, accelerated, with a bigger invoice attached. Some businesses, she added, have spent millions this way for no measurable return.
Her prescribed fix is deliberately unglamorous, and it begins before anyone signs for a tool. Governance comes first: decide what must stay human, what can safely be automated, fix the data, and only then automate. It is a discipline she has set out publicly — that operations, not IT, should lead these rollouts, with technology in a supporting role, and that the quickest way to destroy trust in a deployment is to scale something that only works because staff are quietly compensating for it. That same conclusion, that the returns gap is an organisational failure rather than a technical one, has echoed across this month’s CX events on both sides of the Atlantic.
Will agentic AI cut contact centre jobs?
This is where the BPO consensus splinters, and where the conversation stops being comfortable.
Liveops and TTEC hold the industry line, more or less. Moore argues that the expected wave of mass layoffs simply hasn’t materialised; the value, she says, lies in effectiveness and trust, not headcount reduction. Leslie Kornaker, Chief Sales Officer at TTEC and a 35-year veteran who started on the phones, puts it more firmly still: AI augments, it was never meant to replace. When a client asks her to take 10 per cent out of next year’s budget, she will find efficiencies — but she refuses to pretend that an agent removes half the workforce. “The reality is we’re not there,” she told me. “And I don’t know if we ever will be.” The rollback pattern the industry has started to recognise — the companies that cut deep, found the automation couldn’t carry the load, and quietly rehired — is exactly the outcome her “augment, never replace” stance is built to avoid.
Then there is LeGrand Bonnet, SVP of Global BPO Operations at CBE Companies, who refused the script altogether. His complaint is about candour. At CCW, he pointed out, everyone repeats the “humans and AI” mantra; at the technology vendors’ own conferences, the message flips to “we’re getting rid of your agents”. Someone, he argued, ought to say the quiet part out loud. Agents will be replaced. Some of the workforce will shrink, and it is honest, not heartless, to admit it. And here is the line that should give the redeployment evangelists pause: it isn’t always the simple work that gets automated. Sometimes the complicated work goes first.
Four outsourcers, two irreconcilable positions, one uncomfortable truth sitting between them. The honest answer to “Will agentic AI take the jobs?” is neither yes nor no. It is that some will go, gradually, including skilled ones, and that the industry, for all its reassurance, isn’t being straight about it.
What does agentic AI actually cost?
If the jobs question is contested, the cost question is barely addressed, and the BPOs are the first to admit it. Bonnet was candid about the unknown. The moment you move from experimenting with a chatbot to running enterprise-scale agentic AI, you hit token costs, and the real question — whether those costs land above the offshore or nearshore labour they are meant to replace — has no settled answer yet. Complex and back-office use cases are already startling companies with the bill, he noted, and the contact centre is next in line. His sharpest observation cuts through the marketing entirely: nobody runs a fully autonomous contact centre at scale yet, so until real volume flows through one, every cost projection is a guess.
This is the thread the vendors smooth over with talk of outcome-based pricing. The direction of travel is real — Liveops itself describes a market shifting from capacity to outcomes, and Salesforce made the same claims recently at their Agentforce event last week. But an outcome-based model only helps if you know what the outcome costs to produce, and right now, the people who would have to underwrite that economics don’t.
Eric Guarro of IBEX is the one willing to say what the boardroom is really thinking. The industry, he argues, has tied itself in knots, avoiding two words: “deflection” and “cost”. Everyone wants to sound enlightened — I’m not really trying to deflect calls, I’m improving the experience. “Sure you are,” is Guarro’s response. Once you get into the boardroom, he says, those things simply have to happen: taking cost out, deflecting routine contacts, and standing up support that runs around the clock is the actual brief, and pretending otherwise helps nobody.
What separates a good deployment from a bad one is never the ambition to cut costs; it is the approach. Done well, automated deflection is a better experience for the customer than waiting on hold. When done badly, it is the kind of broken interaction that gives cost-cutting a bad name. The intent was never the problem. The execution is.
How companies choose AI tools — and why so many choose badly
Kornaker offered the clearest explanation of why so much CX spend misfires: the buying itself has changed shape. She used to sell seats; now she sells seats and technology, which means selling to IT, compliance and marketing as well as the contact-centre head — more stakeholders, competing priorities, and a longer, more tangled sales cycle. Guarro described the same shift without prompting, watching his own buyers multiply from the contact-centre owner into a whole committee. Two outsourcers are naming the same change. Firms that try to navigate it alone tend to get patchy results, Kornaker says, where the slickest pitch beats the best fit, which is, conveniently, the gap a BPO that has already run the tool and its rivals is selling itself to fill.
Guarro adds the part most buyers underestimate. “AI just doesn’t work without the right foundations.” Clean data and real access to the back-end systems have to come first, and that plumbing, unglamorous as it is, is where ibex claims its edge: decades spent wiring into client CRMs, data warehouses and contact platforms for its human operation, now repurposed for the agentic one. His other warning is about how teams brief the technology in the first place. Too many fixate on getting an agent to perform discrete tasks, when the better question is which parts of the customer’s whole journey should be automated, which should be streamlined, and where a human should pick the interaction back up.
Bonnet frames the same discipline more bluntly: find your why. Are you chasing efficiency, a headcount cut, or a better experience? Define what winning looks like, prove the return, and if you can’t do either, don’t start. Moore’s version is to pilot on real, live data before scaling anything. For all the daylight between these operators on the jobs question, they converge completely here — the technology is the last decision a company should make, not the first.
Should you tell customers they’re talking to AI?
On disclosure, the operators are unanimous, and they puncture a fashionable assumption while they are at it. Customers want to know when they are dealing with a machine, and the group most insistent on that is the youngest, not the boomers. Bonnet sees a real backlash among people leaving school and university, possibly sharpened by fear for their own prospects. Kornaker rejects the cliché that younger customers only ever want to message, pointing to how readily younger staff embraced the return to the office once it came; people, she argues, still want human contact. Moore’s 2026 Resolution Gap research lands in the same place: around half of the youngest customers still want the option of a person.
The operational lesson they all draw is identical, and worth pinning above the desk: keep a fast, obvious route to a human. Not a buried menu option, but a quick one. Bonnet, fresh from a maddening exchange with a retailer’s AI agent earlier that day, was unequivocal. If the AI is failing the customer, let them out of the interaction, and let them out immediately.
The thing that actually scares them
For all the talk of cost and jobs, the moment an outsourcer’s composure visibly cracked came on a different subject altogether. I asked Bonnet about machines talking to machines — agent-to-agent interactions — and he didn’t reach for reassurance. It is already happening, he said, and it terrifies him. Anyone can now stand up their own autonomous agent, and those agents have begun negotiating with one another, unsupervised and unaudited. From a man who runs government contracts, the instinct wasn’t excitement but caution: put the brakes on, he urged, until the industry can get a better handle on what these things are doing.
It was a fitting end to a day of BPO conversations. These are not AI sceptics — they sell agentic AI, deploy it, and have staked their businesses on it. They are simply the least romantic people in the building, and the middle of the market has a habit of telling the truth first. Agentic AI won’t rescue a broken operation. The cost model is still a black box. The boardroom wants to cost out, whatever the conference vocabulary pretends. And the customer, especially the young one, still wants the option of a human. The vendors will tell you the future is settled. The people who run the floor know it isn’t.
Frequently asked questions
Does agentic AI reduce contact centre headcount? It depends on who you ask. At CCW 2026, BPO leaders split on the question. Liveops and TTEC argue that AI augments staff rather than replaces them, and that fears of mass layoffs have been overstated. CBE Companies’ LeGrand Bonnet argues the industry should be honest that some roles, including complex ones, will be automated over time. The realistic answer is that some jobs will go gradually, while the nature of the roles that remain changes.
Why do so many agentic AI projects fail to deliver ROI? Liveops President and COO Molly Moore argues that AI is too often bolted onto operations that are already broken — fragmented data, weak processes, no clear governance — which it then amplifies rather than fixes. The recommended remedy is to sort out governance, data and process first, and to treat AI as an operational programme led by the business rather than a technology project led by IT.
What does agentic AI cost to run at scale? There is no settled answer yet. CBE Companies’ LeGrand Bonnet notes that enterprise token costs are unpredictable and may even rival the offshore or nearshore labour they are meant to replace. Because no organisation has yet run a fully autonomous contact centre at production volume, the true unit cost remains unknown.
Should companies tell customers when they are talking to AI? Yes. BPO leaders agree that customers want to be told, and that, contrary to popular assumption, the youngest customers are often the most resistant to AI. A fast, obvious route to a human agent is considered essential.
What is a BPO, and why does its view on agentic AI carry weight? A BPO, or business process outsourcer, runs customer service operations on behalf of other companies. Because it is commercially exposed to both AI adoption and the human workforce AI affects, its perspective tends to be more calibrated and less promotional than that of a technology vendor.
