Talkdesk CXA Operations Center: One Workforce, Humans and AI
Talkdesk announced CXA Operations Center on March 10 at Enterprise Connect. The aim is simple: operate human and AI agents as one team, with shared metrics, shared visibility, and shared accountability.
If you manage operations or HR, this isn't a future problem. AI agents are digital labor. They need schedules, guardrails, metrics, and performance reviews-just like people.
What the new Operations Center actually does
CXA Operations Center brings two core capabilities to the table:
- AI Agent Observability (AAO): Monitor live activity, inspect session-level execution traces, and pinpoint failures with detailed error data.
- AI Agent Evaluation (AAE): Measure behavior across scenarios, compare versions, and track progress against agent-level metrics.
Talkdesk also upgraded Interaction and Quality Analytics with a dynamic dashboard that connects what happens in interactions to business outcomes. Less guessing, more signal.
Why this matters for Ops and HR
Most workforce tools still treat AI as invisible. That breaks your planning. As Pedro Andrade, VP of AI at Talkdesk, put it: "Currently, workforce management solutions only look at humans, because they don't consider AI as part of the workforce. Well, that is changing."
Here's the kicker: when AI handles low-to-medium complexity tasks, the remaining calls routed to humans are tougher. Average handle time (AHT) for human agents will likely go up-not down. If your system can't see the AI workload, it looks like your team is getting slower. They're not; the work got harder.
Andrade again: "Traditional workforce management systems will only see that AHT is skyrocketing because they can't see that part of that work is being taken by AI."
Key metrics to add for a hybrid workforce
- AI containment rate: % of interactions fully resolved by AI.
- Escalation (handoff) rate and quality: How often AI passes to humans and how clean that pass is.
- Deflection to self-service: Movement from voice to digital or automated channels.
- Resolution split: First contact resolution for AI vs. human.
- AI reliability: Failures per 100 sessions, timeouts, policy violations.
- Impact metrics: CSAT by path (AI-only, human-only, blended), revenue saved or generated, and productivity gained.
Practical steps to capture value now
- Map your work by complexity. Route low/medium complexity to AI with clear rules. Keep clean escalation paths for high-risk and emotional scenarios.
- Update WFM models. Treat AI as part of the roster with its own capacity, SLAs, and "shrinkage" (downtime, maintenance windows, model retraining).
- Redefine KPIs and targets. Expect human AHT to rise while overall handle time per issue (AI + human) should fall. Track both.
- Stand up observability. Use session traces and error inspection to resolve failures fast and prevent repeats.
- Version and evaluate. A/B test AI agent versions. Compare against agentic metrics before promoting to production.
- Redeploy capacity. As AI absorbs repetitive tasks, move skilled agents into complex support, QA loops, or product feedback.
- Coach differently. Train supervisors to manage mixed teams, and teach agents how to collaborate with AI during live escalations.
Quotes worth noting
"The framework for human management has existed and is something we've been building on for years. We're taking the AI pieces and putting them in [as well] so you'll see one workforce," said Pedro Andrade. "If you're looking at things as two separate silos, you're actually missing a lot of the opportunity."
On planning impact: "If you're seeing that you don't need human effort in one area, those people can be moved into something else - improving products, for example. You basically redirect that budget into something else."
Risks to avoid
- Siloed reporting. If AI work is invisible, you'll misread AHT and make the wrong staffing calls.
- Over-automation. Poorly handled edge cases will erode trust. Require clean handoffs and clear escalation logic.
- Compliance gaps. Log decisions, protect data, and audit model behavior-especially on denials, refunds, or sensitive interactions.
What this means for your next quarter
- Add AI agents to your workforce plan and capacity model.
- Adopt a shared dashboard for AI + human metrics.
- Revise incentives so teams aren't penalized for higher AHT on harder work.
- Set a monthly AI evaluation cycle with version comparisons and rollback plans.
Level up your team
- AI Learning Path for Call Center Supervisors - practical training on routing, AI containment, staffing, and KPI resets when AI takes part of the load.
- AI Learning Path for HR Managers - workforce planning with digital labor, policy and governance, and skills development for hybrid teams.
Bottom line
You can't manage what you can't see. Treat AI agents as part of your workforce, measure them with the same rigor, and make staffing and budget moves based on the full picture. That's how you speed up resolution, protect CX, and free up people for higher-value work.
Your membership also unlocks:
