Twilio bets on AI agents for customer support
Twilio aims to delegate simple requests to AI agents while keeping humans for complex cases. The company also advocates a model where context follows the customer across channels.
Twilio encourages companies to split customer support between AI agents and human advisors, rather than replacing call centers entirely. Thomas Wyatt, Chief Operating Officer, explains that level 1 and 2 requests can be automated, while complex cases must remain in human hands. The challenge, according to him, lies as much in the quality of response as in maintaining context continuity when the conversation switches from one system to another.
Automatiser le simple, garder l’humain pour le reste
Twilio's approach is pragmatic: a significant portion of customer service requests does not require a highly specialized agent. An AI agent can learn from the knowledge base, FAQs, and a company's internal policies, then handle repetitive requests. Thomas Wyatt sums up the limit right away: "AI agents can only take it so far. There’s usually a reason why a certain percentage of those cases need to be escalated to a human," he says. In other words, automation helps, but it does not replace human judgment when the case becomes complicated.
The sensitive point lies elsewhere: when AI transfers the case to an advisor, the customer should not have to start over. Twilio highlights Conversation Memory, a mechanism that retains useful history so the specialist can pick up the thread without making the caller repeat themselves. This kind of detail makes all the difference in customer support, especially when the customer's patience is not unlimited.
Le support devient un choix, pas un couloir obligé
Wyatt anticipates a more deliberate use of free choice between AI and human. He cites Philippine Airlines, where customers can opt for the AI agent or wait for an advisor, depending on their needs. A baggage allowance change can be handled by automation; a request involving an elderly passenger, wheelchair assistance, or a special meal is more likely to require a human. The distinction seems simple, but it outlines a new sorting in the customer experience: what is standardized on one side, what requires attention on the other.
Twilio does not only sell this role distribution. The company also promotes voice as a growth driver. According to Wyatt, messaging remains the group's largest segment, voice comes next, and WhatsApp is growing the fastest, even from a smaller base. The voice segment is accelerating significantly thanks to applications built around voice agents, with features like recording, transcription, sentiment analysis, real-time connection to LLMs, and customer memory retention.
Pourquoi la voix redevient stratégique
This direction is not just a trend. Twilio aims to provide the software layer that allows building business workflows over communication channels, without forcing companies to rebuild the entire infrastructure. Wyatt emphasizes continuity between channels: an interaction can start with a call, continue via SMS, then end with an email. The result, according to him, forms a single conversation throughout the relationship between a brand and a customer.
The group also believes its advantage remains hard to imitate. Wyatt recalls that Twilio relies on about 4,800 operator relationships in 180 countries, built over more than a decade. Whether a message passes through a telecom operator, via RCS through Google, or WhatsApp, it always depends on physical infrastructure and distribution agreements. "You could vibe-code a wrapper around our communication infrastructure, but you can’t vibe-code the communication infrastructure," he says. The phrase is well chosen, and not just to make you smile.
L’IA aide à coder, mais ne doit pas tout écrire
Regarding software development, Twilio adopts the same caution. Wyatt sees real interest in vibe coding—tools that generate code from natural language instructions—for prototypes, mockups, or proofs of concept. However, he advises against making it the main method for building production enterprise software. The problem is not generation speed but maintainability: if no one really understands how the application was assembled, debugging and updates become difficult.
In this context, Twilio envisions tools like Claude or Codex as development assistants, not autopilots. Without human judgment, the company risks accumulating technical debt and what Wyatt calls "slop," sloppy code that will need fixing later. The message is clear: AI speeds things up, but it does not absolve architectural responsibility.
L’Inde, terrain de croissance mais encore au début
Twilio also sees an opportunity in Asia-Pacific, particularly India. Wyatt describes the APJ region as one of the group's most dynamic, while India remains, at this stage, a still young market for the company. He cites Reliance as a flagship client in the country, without disclosing local revenue. Investment in R&D and support capabilities is nevertheless described as significant.
The market context fuels this strategy. According to Deloitte, the Indian enterprise AI market was worth between 7.6 billion and 9.5 billion dollars in 2024-2025 and could exceed 130 billion dollars by 2032, with a compound annual growth rate of 39% to 42%. Gartner adds that by 2030, more than 80% of companies are expected to deploy AI agents specific to their sector. For Twilio, the challenge is simple: if these agents talk to customers, they will need voice, messaging, and email to do so at scale.
Key points
- Twilio wants to automate simple requests, not all support.
- Philippine Airlines serves as an example of choice between AI and human.
- 4,800 operator relationships in 180 countries support Twilio's infrastructure.
- Gartner forecasts more than 80% of companies equipped with AI agents by 2030.
- Reliance is cited as a flagship Twilio client in India.
By the numbers
- 4,800 operator relationships — Twilio infrastructure, 180 countries, according to Thomas Wyatt.
- 180 countries — presence of Twilio operator relationships, according to the company.
- 7.6 to 9.5 billion dollars — Indian enterprise AI market in 2024-2025, according to Deloitte.
- 130 billion dollars — projected size of the Indian AI market by 2032, according to Deloitte.
- More than 80% — share of companies that will deploy sector-specific AI agents by 2030, according to Gartner.