From call center to "empathy centers": How agentic AI is changing the rules

The global call center industry is undergoing a structural shift unlike anything seen in three decades. Agentic AI — systems that not only respond to commands but independently plan, execute, and adapt within multi-step processes — has moved from the concept phase to live operations. According to Gartner estimates published in the latest report, by mid-2026 agentic AI will be part of 56 % of customer interactions, and by 2028 this could reach up to 68 %. Analysts also expect that by 2029 it will be able to autonomously resolve 80 % of routine service issues with a 30 % reduction in operating costs.

These numbers, however, hide an important nuance: only 27 % of enterprise CX teams that launched a pilot project with agentic AI have at least one channel in full production. The remaining 73 % fail for a single reason — they approached the technology as a replacement for human expertise, not as a system that needs it to function. Filipino BPO (Business Process Outsourcing) companies, which manage both front-office and back-office operations for global clients, have become the testing ground where this reality is verified at a measurable scale.

Six-layer architecture: Front office versus back office

The agentic workflow in leading Filipino call centers is not a random collection of chatbots. It is a six-layer operational architecture where autonomous agents handle high-volume, predictable tasks across both the customer-facing and administrative sides, while Filipino specialists manage decisions affecting brand, regulatory compliance, and model accuracy.

This architecture is not theoretical — it is a direct response to the reality that Gartner defined clearly: 40 % of agentic AI initiatives will fail by 2027 if they lack governance, observability, and ROI frameworks. The six-layer stack is precisely that governance framework.

Front office — what the customer sees

In front-office operations, agentic AI handles the first line of communication: ticket classification, basic resolutions, request routing, and instant responses. Data from 2026 shows that purely AI processing achieves a CSAT score of 4.1 out of 5, while hybrid programs with human oversight achieve 4.3 out of 5. The difference may seem small, but in an environment where 70 % of consumers perceive a widening performance gap between AI leaders and lagging companies, every tenth of a point can be decisive.

Back office — what happens behind the scenes

The back office is where agentic AI handles invoicing, refunds, logistical updates, and compliance with internal procedures. Yet here the greatest risk lies without human control — only 31 % of organizations, according to an Adobe and Oxford Economics study from 2026, have a measurement framework for agentic AI. This means that 69 % of businesses have no system to audit AI outputs, record compliance deviations, and feed corrective data back into retention pipelines.

Why the Philippines? The human factor as a competitive advantage

Filipino call centers are not a random choice for global outsourcing. Data from 2026 reveals several structural reasons why this country has become an essential human layer of the agentic workflow:

• AI literacy at scale: 86 % of Filipino white-collar workers already use AI tools (LinkedIn / Microsoft Work Trend Index 2024) — the highest rate in Asia. The Filipino BPO workforce is therefore not joining AI; it is already working in AI-augmented workflows.
• Documented productivity increase: A Stanford and MIT study tracking 5,179 Filipino BPO agents and 3 million customer chats found that AI assistance increased productivity by 13.8 %. The biggest gains were recorded by lower performers, because AI encodes the knowledge of top agents and distributes it to the team.
• Excellent CSAT for complex interactions: While pure AI achieves 4.1/5 and hybrid 4.3/5, escalations handled by Filipino specialists consistently reach 4.6/5. In an environment where customers are increasingly noticing differences between companies, these 0.5 points can be decisive for retaining a client.
• Growing governance role: Filipino BPOs have created 80,000 new jobs in AI oversight and data science since 2023. Prompt Engineering, Knowledge Base Curation, and AI Training are among the fastest-growing categories.

According to John Maczynski, CEO of PITON-Global and a 40-year veteran of the BPO industry: "The deflection rate tells you what AI can do when it works. The CSAT gap tells you what it costs when it doesn't. It is precisely in this space that Filipino providers are creating their competitive advantage — not as a backup for AI, but as an optimization layer that closes this gap."

Four conditions for successful deployment

With 64 % of enterprise CX teams running pilot projects but only 27 % in full production, the production gap is the defining problem of 2026. Advisory work by PITON-Global identifies four essential differentiators of success:

1. Knowledge architecture before software: Every client who came after a failed AI deployment made the same mistake — they purchased software before building a knowledge architecture. A Retrieval-Augmented Generation (RAG) system cannot work on a chaotic document fairy tale and expect 90 % solution accuracy.
2. Governance and observability: Without a measurement framework there is no feedback, and without feedback there is no improvement. Companies need audit layers that check AI outputs in real time.
3. Hybrid workforce: AI handles routine, humans handle exceptions, emotions, and complex decisions. This model is not transitional — it is a permanent structure.
4. ROI framework from the start: Deployment must have clear KPIs not only for cost efficiency but also for CSAT, resolution time, and customer retention.

What does this mean for Czech companies?

For Czech companies considering outsourcing customer support or automating internal processes, the Filipino model offers several lessons. The Czech BPO sector — with centers in Brno, Ostrava, and Prague — may not compete with Filipino costs, but it can compete on quality and proximity to European markets. With the arrival of the EU AI Act, moreover, every company operating in Europe needs auditable AI decisions, which strengthens the argument for human oversight.

Data from 2026 also debunks the replacement narrative: the Filipino BPO sector employs a record 1.97 million people and is growing toward 2.5 million by 2028 — precisely at a time when agentic AI covers 56 % of contacts. Both numbers are true simultaneously because they describe different layers of the same workflow. Czech companies should therefore avoid the idea that AI will reduce the need for people. The reality is exactly the opposite: AI increases demand for specialized human roles that can manage, audit, and optimize it.

For Czech startups and SMEs seeking affordable customer support, outsourcing to the Philippines remains a relevant alternative — especially if the partner can offer a six-layer agentic workflow with guaranteed human oversight. For companies that prefer a local approach, the challenge is to build a similar hybrid model with Czech agents enhanced by AI tools that are now available including Czech language support (for example through translators or native models such as GPT-4 or Claude).