AI agents are taking over routine customer service tasks, pushing human contact centre workers towards more complex problems that require judgement, specialised knowledge, and human intervention.
In an exclusive interview with AIM, Thomas Wyatt, COO of communications-platform-as-a-service provider Twilio, says enterprises should not look at AI as a wholesale replacement for their contact centres. Instead, companies can route routine level-one and level-two queries to AI while reserving human agents for more complicated cases.
A large share of customer service queries do not require a highly skilled human agent. AI agents can be trained on a company’s knowledge base, FAQs, policies, and other information to resolve routine requests.
Wyatt says this can reduce the amount of human capital required for cases that are relatively straightforward. But AI will not necessarily be able to resolve every customer problem.
“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,” Wyatt says.
But handing the conversation to a human comes with its own requirement: context cannot be lost. Wyatt says customers should not have to repeat their entire problem after an AI agent hands the conversation to a human.
Wyatt explains that Twilio Conversation Memory helps preserve the context of a conversation, allowing a human specialist to understand what has happened and continue resolving the issue without asking the customer to repeat themselves.
AI or Human Support
Wyatt believes customers will increasingly choose between AI and human support based on the complexity of their needs.
He points to Philippine Airlines, which gave customers the choice of interacting with an AI or waiting for a human.
A simple request, such as increasing baggage allowance, could be handled by AI. A more complicated situation involving an elderly passenger, wheelchair assistance, or special food requirements could be directed to a human.
“People are actually pretty smart to self-serve,” Wyatt says. He expects this choice to become a bigger part of customer service as AI agents become more capable.
Voice AI a Major Growth Driver
Messaging remains Twilio’s largest business, followed by voice, while WhatsApp is the fastest-growing channel, albeit from a smaller base.
Voice AI has become particularly important. Wyatt says Twilio’s voice business is growing faster than it has in years, driven by building applications around voice agents.
Per Wyatt, customers want Twilio to provide capabilities such as voice recording, transcription, sentiment analysis, real-time connections to large language models (LLMs), and customer memory.
“They want to focus their energy on building the workflows and the verticalisation of those use cases to meet the customer need in a very specialised way,” Wyatt says.
Twilio is also extending these software capabilities across channels. A customer using the platform for voice can add messaging without having to build a separate software stack.
Twilio’s Moat
Wyatt says that while AI has made it relatively easier to generate software, Twilio’s underlying communications infrastructure is much harder to replicate.
Twilio has around 4,800 carrier relationships across 180 countries, built over more than a decade.
Whether a company sends an SMS through a carrier, uses RCS through Google or communicates through WhatsApp, the interaction ultimately depends on physical communications infrastructure and relationships with carriers and platforms.
“You could vibe-code a wrapper around our communication infrastructure, but you can’t vibe-code the communication infrastructure,” Wyatt says.
The company wants to move beyond connectivity and provide the software layer that sits on top of that infrastructure.
Wyatt sees value in vibe coding for simple applications, prototypes, and proof-of-concept work. But he cautions against using it as the primary way to build production-grade enterprise software.
The problem is maintainability. If developers do not understand how an application was built, debugging and modifying it can become difficult.
He expects enterprises to increasingly use tools such as Claude and Codex as coding assistants, rather than handing over entire software projects to AI.
However, without human judgement, companies risk creating technical debt and what he describes as “slop” that will have to be fixed later.
From APIs to AI-Native Communications
Twilio historically provided APIs for individual communication use cases such as SMS, voice calls, and email. Wyatt says the company is now bringing these interactions together into a single, continuous conversation.
He gave the example of a customer applying for a mortgage. The process could begin with a phone call, continue through a text message confirming information, and later involve an email updating the customer on the progress of the application.
“It really is a unified conversation that happens over the course of a relationship between a brand and a consumer,” Wyatt says.
AI is now being added on top of this infrastructure. Twilio wants AI agents to be able to access its communication capabilities directly, allowing them to communicate with customers across different channels.
The company sees an opportunity in the expected growth of enterprise AI agents. According to Gartner, more than 80% of enterprises will deploy industry-specific AI agents by 2030. Many of those agents will eventually need to communicate with humans through voice, messaging, or email.
India Market
India is an important market for Twilio within the Asia-Pacific and Japan region, which Wyatt describes as one of the company’s fastest-growing regions globally.
According to Deloitte, India's enterprise AI market is expanding rapidly, valued between $7.6 billion and $9.5 billion as of 2024–2025 and projected to surpass $130 billion by 2032 at a 39–42% CAGR.
The company does not disclose specific revenue figures for India, but Wyatt identifies Reliance as one of Twilio’s flagship customers in the country.
“It’s a market where we have a lot of R&D as well and customer care capabilities in terms of Twilio's investment in the region,” he says.
Wyatt also points to the growing number of AI-native companies in India and the wider APJ region using Twilio, although he said the Indian market remains in its early stages for the company.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.