In 2026 the world's largest outsourced customer-service company has become one of Europe's most-shorted stocks. According to reporting by the **Financial Times** and **Bloomberg**, short interest in Teleperformance climbed from roughly 3.8% of free float to about 6.4%, with later data points in May 2026 reported as higher still. The same outlets have named hedge funds including **Marshall Wace, Point72, Citadel Advisors and Squarepoint** among those building or holding short positions against the stock.
A widely cited catalyst, as reported by the **Financial Times**, **Reuters** and **Bloomberg**, was a disclosure by Klarna that its in-house AI assistant was handling work equivalent to roughly **700 full-time customer-service agents**. That single data point preceded a sharp single-day drop in Teleperformance shares, and the stock was subsequently removed from the **CAC 40** index. Concentrix, TTEC and several Indian operators saw correlated declines, per the same reporting.
We are not here to commentate the trade. We run a multilingual nearshore BPO out of Morocco and we read the tape carefully, because our clients ask us the same question every week: *is outsourcing dying?* The short answer, based on what the market is actually pricing, is no. **What is being shorted is one specific operating model — the pure headcount-arbitrage call center.** A different model, AI-augmented BPO, is the one that survives the next cycle. This article explains the distinction, and gives buyers a concrete checklist for picking a partner that will still be here in five years.
Read the analyst notes that followed the Klarna disclosure and a consistent thesis emerges. The bear case against legacy contact-center operators rests on three assumptions:
This thesis is internally consistent. It is also, as written, an attack on a *specific operating model*, not on outsourcing as a category. The model under attack is the one that defines value as **bodies per shift per language**. The model that survives is the one that defines value as **resolved interactions per euro, regardless of whether a human or an AI did the resolving.**
The augmentation framing is now the consensus industry view. Per analyst commentary widely cited by **Reuters** and reported in coverage of Gartner research, AI could **save the contact center industry tens of billions of dollars** over the second half of this decade through deflection, agent assist, and automated quality management — while a growing share of customer interactions are automated by 2026 (estimates vary by source and are best treated as directional, not exact).
The operational reading of those numbers is straightforward. The contact center of 2026 looks like this:
The legacy model staffs the second and third tiers as if the first did not exist. The AI-augmented model designs the second and third tiers *around* the deflection profile of the first. The cost difference between those two designs, at scale, is the entire short thesis.
The phrase is over-used and under-defined. Concretely, a partner that deserves the label runs all four of the following in production today:
If a prospective partner cannot demonstrate all four with live dashboards on a kickoff call, the partner is selling 2019 BPO with 2026 wallpaper.
Use this when shortlisting partners. Score honestly; the answers are usually visible in the first 90 minutes of due diligence.
A partner that scores well on six of eight is shortlist-grade. Five or fewer, walk.
The buyer mistake we see most often in 2026 is the false binary: keep the legacy BPO contract because it is cheap and renegotiate later, or rip it out and rebuild in-house with AI. Both end badly. The legacy contract gets re-priced upward as the partner amortises declining volume; the in-house rebuild discovers that AI deflection in a regulated, multilingual environment is an 18-month engineering project, not a quarterly deliverable.
The pragmatic path is a structured transition with a partner that already operates the augmented model. Move tier-1 volume to AI deflection inside 90 days. Re-shape the human staffing to specialise on complex work. Re-price the contract against resolved interactions inside 12 months. The math works at every reasonable assumption set we have run with clients.
This is the model **Call IT Dev** operates from Casablanca and Madrid. AI deflection on the addressable intent set; human escalation in 12 production-grade languages with CET alignment; transparent QA on every interaction; outcome-based pricing available from day one. We are deliberately built for the world the short sellers are pricing in — not the one the short sellers are pricing out.
If your current BPO partner cannot answer the checklist above, the conversation is worth having now, while the market is still rewarding decisive moves.
No. As reported by the **Financial Times** and **Bloomberg**, the short thesis targets the headcount-arbitrage operating model — the assumption that contact-center value scales linearly with agent count. The same reporting frames AI-augmented models, where humans and AI work together, as the model the market expects to survive. Outsourcing as a sourcing strategy is not in question; the specific operating model is.
Per disclosures cited by **Reuters**, the **Financial Times** and **Bloomberg**, Klarna stated that its AI assistant was handling work equivalent to roughly 700 full-time customer-service agents. That is the framing of the equivalence, not a one-to-one statement that 700 named individuals lost their jobs in a single event. The figure is widely treated as directional evidence that AI deflection is operationally real at scale.
The consensus analyst view we see — including commentary attributed to Gartner research — is no. AI handles structured, deterministic volume well. Complex, regulated, and high-empathy interactions still resolve better with humans, and the share of those interactions is not trending toward zero. The operative number for buyers is the deflection ceiling on the addressable intent set, typically reported in the 50–70% range.
Use the eight-point checklist in this article. The fastest single signal is whether the partner will price a meaningful share of the contract on outcomes (resolved interactions, CSAT, FCR) rather than exclusively per agent-hour. Partners that have re-tooled the operating model will accept that frame; partners that have only rebranded will not.
Because AI deflection raises, not lowers, the importance of the human escalation tier. When a human picks up after AI handoff, the SLA expectation tightens — the customer has already been waiting through an AI turn. Operating in a time zone aligned with the served market keeps that escalation SLA achievable 24/5 without expensive overnight staffing. For European traffic, **CET-aligned locations like Casablanca and Madrid** are operationally indistinguishable from onshore from the customer's point of view.
For a mid-market customer with one or two languages and a clean CRM stack, the operational transition is typically 90 days for tier-1 deflection, with the contract structure repriced inside 12 months. Larger, multilingual, regulated environments run 6–9 months end-to-end. The risk is not the technology timeline; it is the change management with the existing vendor.
It depends on the deflection ceiling for your intent mix and on the residual mix between AI-assisted tier 1, specialist tier 2, and the rare tier 3. In production, blended cost per contact typically lands at 35–55% of an all-human baseline once the transition is complete, with CSAT flat or improving. The savings are real; the prudent buyer plans for the low end of the range.
We are the AI-augmented, multilingual, nearshore BPO designed for European and US buyers who do not want to rebuild contact-center engineering in-house. The model is the one this article describes. The fastest way to a concrete answer for your environment is a 15-minute scoping call.
If the analyst notes are landing on your desk and you want a second opinion before the next renewal cycle, we will spend 30 minutes on your operating model, your deflection ceiling, and your realistic cost-out — no slides, no pitch.
The market is repricing the contact center. Decide whether that is a threat or a tailwind on your terms.
No. As reported by the Financial Times and Bloomberg, the short thesis targets the headcount-arbitrage operating model, not outsourcing as a sourcing strategy. AI-augmented models, where humans and AI work together, are framed by the same reporting as the model the market expects to survive.
Per disclosures cited by Reuters, the Financial Times and Bloomberg, Klarna stated that its AI assistant was handling work equivalent to roughly 700 full-time customer-service agents. That is an equivalence statement, treated by the market as directional evidence that AI deflection is operationally real at scale.
The consensus analyst view, including commentary attributed to Gartner research, is no. AI handles structured deterministic volume well; complex, regulated and high-empathy interactions still resolve better with humans. The operative number for buyers is the deflection ceiling on the addressable intent set, typically reported in the 50–70% range.
Use the eight-point checklist in this article. The fastest single signal is whether the partner will price a meaningful share of the contract on outcomes (resolved interactions, CSAT, FCR) rather than exclusively per agent-hour. Re-tooled partners accept that frame; rebranded ones do not.
Because AI deflection raises the importance of the human escalation tier — once a customer has waited through an AI turn, the SLA on the human handoff tightens. CET-aligned operations like Casablanca or Madrid keep that SLA achievable 24/5 without expensive overnight staffing.
For a mid-market customer with one or two languages and a clean CRM stack, tier-1 deflection typically rolls out in 90 days, with the contract structure repriced inside 12 months. Larger, multilingual, regulated environments run 6–9 months end-to-end.
In production, blended cost per contact typically lands at 35–55% of an all-human baseline once the transition is complete, with CSAT flat or improving. The exact figure depends on the deflection ceiling for your intent mix.
Call IT Dev is an AI-augmented, multilingual, Morocco-nearshore BPO designed for European and US buyers. The fastest way to a concrete answer for your environment is a 15-minute scoping call.
CALL IT DEV — Software, AI and dedicated tech teams — Casablanca | Madrid | Dubai — contact@callitdev.com — +212-537-373777