In her analysis "Unbundling the BPO: How AI Will Disrupt Outsourced Work," Andreessen Horowitz partner Kimberly Tan argues that AI is unbundling the roughly 300 billion dollar BPO industry away from a monolithic seat-based model toward cross-system automation and coding agents, with voice AI and browser-native agents making previously un-automatable workflows automatable. She cites illustrative examples such as Salient in auto-lending voice calls and Avoca for overflow home-services calls. Buyers should treat this as a16z's thesis and opinion, not as settled fact, and combine it with their own workflow evidence and market data before restructuring a live BPO relationship.

The global BPO market is valued at approximately 353.64 billion dollars in 2026 and projected near 741.60 billion dollars by 2034, about a 9.7 percent CAGR, per widely cited market-sizing reports. Gartner projects that 75 percent of customer interactions will be AI-powered by 2026, powered by AI and not fully handled by AI, and the distinction matters. Reported unit economics put a per-interaction cost floor of about 6 to 12 dollars for human handling versus under 1.50 dollars for AI, with generative AI resolving roughly 60 to 80 percent of routine queries. The Philippine IT-BPM sector is on track for approximately 42 billion dollars in export revenue in 2026 employing roughly 1.97 million workers. Together those numbers describe a restructuring industry, not a collapsing one.

It changes four things at once. Scope stops being defined by processes and becomes a task inventory, each task with its own definition, success criterion, cost profile and optimal human-AI split. Pricing stops being uniform per seat and becomes task-differentiated, with high-volume routine tasks priced at the AI-plus-supervision cost floor and low-volume high-judgement tasks priced at a full skilled-human rate. Governance stops being 'is the team hitting AHT and CSAT?' and becomes task-level quality management on containment rate, escalation rate, resolution accuracy and cost per resolved instance. Risk is repriced, with tooling and automation risk moving partly to the partner that operates the AI while judgement, brand and exception risk stay with the human tier and closer to the buyer.

One, map your workflows down to the task level as a buyer-owned artefact, with volume, variance, judgement content, failure cost and current handling cost per task. Two, adopt a measured 80/20 human-AI hybrid as a target reached by moving tasks one at a time. Three, shift from seat-based to task or outcome-based pricing progressively, running the two schedules in parallel until the economics validate. Four, keep a smaller, higher-skill human judgement layer for exceptions, empathy, fraud, regulatory-sensitive handling and high-value relationships. Five, choose a partner that operates the AI stack — models, prompts, workflows, evaluations, guardrails, monitoring, escalation, retraining — rather than one that resells a tool. Six, treat data-readiness and governance as the first workstream of the engagement, not a phase two.

Because when the human tier shrinks in headcount but rises in skill floor, its location, language mix, time-zone alignment and governance access become materially more important than in the volume-BPO era. The human tier is where exceptions, empathetic conversations, fraud calls and audit-defensible handling actually happen, and where the buyer's brand and risk sit. A multilingual nearshore team with delivery depth across English, French, Spanish, Arabic and German on EU time zones, with CNDP Law 09-08 and GDPR alignment and a 3-hour flight from European capitals for governance access, is a strong operational fit for that judgement layer while the volume tier is operated in AI or in lower-cost geographies.

Call IT Dev operates as a BPO, customer support and AI automation partner for mid-market buyers restructuring their outsourcing relationships toward task-based delivery, from Morocco, with nearshore EU-time-zone delivery in English, French, Spanish, Arabic and German, aligned with CNDP Law 09-08 and GDPR. The engagement pattern covers the six-step playbook: workflow mapping to the task level, a measured 80/20 human-AI hybrid moved one task at a time, task and outcome-based pricing alongside the seat schedule, a human judgement layer for exceptions and empathy, operation of the AI stack rather than resale, and data-readiness plus governance as the first workstream with monthly evidence on containment, quality and cost per resolved instance.