Phone calls are back-powered by AI agents handling thousands at once
Businesses spent years hiding phone numbers and pushing customers toward chatbots to reduce contact centre costs. Now they're reversing course, deploying voice AI agents that can manage thousands of simultaneous calls while upselling products and services.
Voice AI revenues at cloud communications company Twilio grew 49% year-on-year in the fourth quarter of 2025, signalling momentum in the market. The shift reflects a fundamental change: organisations are no longer trying to avoid phone volume. Instead, they're putting phone numbers front and centre on their websites-but with AI handling much of the load.
From black boxes to modular systems
Early voice AI implementations relied on monolithic "black box" solutions. That approach is giving way to modular architectures that integrate AI with existing back-end systems, allowing organisations to choose and train their own AI models.
Twilio implements voice AI in layers of increasing complexity. Organisations start with high-volume, predictable tasks like flight cancellation triage. Once refined, they move to more complex conversations involving multiple back-end systems-returns, shipping, account changes.
The company remains agnostic about which language model customers use. A customer can deploy small models, large models, locally-hosted models for privacy, or cloud-based options. Twilio handles the communications layer-speech recognition, voice generation, interruption handling-while customers control how AI interacts with their CRM and databases.
Voice AI agents can also switch channels mid-conversation, moving from phone to SMS, email, or WhatsApp to complete a transaction. The agent carries context forward: customer loyalty status, purchase history, past issues.
Results from early adopters
Philippine Airlines deployed Twilio's contact centre platform after its 2021 restructuring. The airline now handles routine queries like flight status checks with AI, reducing average wait times to under one minute and cutting monthly customer service costs by around 30%.
PAL aims to reach what it calls a "super AI agent state" by April 2026, where AI fully automates 80% of tasks currently handled by live agents.
The economics favour voice AI even in lower-cost markets. An airline facing mass flight cancellations can instantly deploy AI to handle 50,000 concurrent callers. This eliminates the need to overstaff for seasonal spikes or crises.
The revenue opportunity
Historically, contact centre automation meant headcount reduction. Now organisations are discovering a second benefit: revenue generation.
PAL identifies passengers who haven't purchased checked baggage and sends outbound messages reminding them it's cheaper to buy an allowance now rather than at the airport. The contact centre shifts from cost centre to retention engine.
Obstacles remain
Deploying voice AI at scale is not straightforward. Data silos plague many enterprises. Regional fragmentation complicates matters-a model effective in Thailand may fail in Japan.
As voice AI becomes indistinguishable from human speech, fraud and deepfake concerns grow. Twilio implements infrastructure-level protections including know-your-customer vetting, deepfake detection, and traffic monitoring. The company requires customers to be vetted before sending outbound communications.
Twilio also works with telecommunications carriers, device manufacturers, and messaging services to ensure voice AI calls are branded and identifiable as legitimate.
Learn more: Explore AI for Customer Support and Speech-To-Text technologies shaping contact centre operations.
Your membership also unlocks:
