Last week Twilio, a provider of voice, video and messaging services, reported its second quarter results and they were certainly standout. Revenues spiked by 86% to $275 million and there was a profit of 2 cents a share.
On the earnings call, CEO Jeff Lawson noted: “We have the opportunity to change communications and customer engagement for decades to come.”
And yes, as should be no surprise, one of the drivers will be AI (Artificial Intelligence). Just look at the company’s Autopilot offering (at today’s Signal conference, Twilio announced that the product is generally available to customers). This is a system that allows for the development, training and deployment of intelligent bots, IVRs and Alexa apps.
Now it’s true that there is plenty of hype with AI. Let’s face it, many companies are just using it as marketing spiel to gin up interest and excitement.
Yet Autopilot is the real deal. “The advantage that’s unique to Twilio’s API platform model is that we build these tools in response to seeing hot spots of demand and real need from our customers,” said Nico Acosta, who is the Director of Product & Engineering for Twilio’s Autopilot & Machine Learning Platform. “We have over 160 thousand customers of every size across a huge breadth of industries and we talk to them about the solutions they need to improve communication with their customers. What do they keep building over and over? What do they actively not want to build because it’s too heavy a lift? Those conversations inform the products we create that ultimately help them differentiate themselves through a better customer experience.”
Consider that Autopilot breaks the conventional wisdom that there is an inherent tradeoff between operational efficiency and customer experience. To do this, Twilio has been focusing on pushing innovation with AI, such as with:
- Classification: This involves grouping utterances and mapping them to the correct task. With AI, the system gets smarter and smarter.
- Entity Extraction: This uses NLP (Natural Language Processing) to locate details like time, place, cities, phone numbers and so on. This means it is easier to automate repetitive tasks like setting up appointments (if the customer says “7 at night,” the NLP will understand this).
There are definitely some interesting use cases for Autopilot. One is with Green Dot, which is a next-generation online bank. A big challenge for the company is that its customers are often new to financial services. But with Autopilot, Green Dot has been able to develop a conversational agent that works on a 24/7 basis, on both the IVR and chatbot system in the mobile app. The company also gets a view of important metrics like usage, time spent and common questions about products.
Here are some other interesting use cases:
- Insurance: Generate quotes automatically, file claims easily, and answer FAQs.
- Hospitality: Offer virtual concierge services, answer FAQs, and manage reservations programmatically.
- Real Estate: Field and generate leads, schedule appointments programmatically, and answer questions about listings.
- Retail & Ecommerce: Allow customers to search products, take advantage of promotional offers, and check delivery status.
Keep in mind that changing a traditional IVR system can be complicated and time-consuming—much less having the ability for AI. But with Autopilot, the development is lightning fast. You can create bots with simple JSON syntax and deploy them on multiple channels with zero code changes. There are also easy-to-use tools for training the AI models.
Takeaways With Autopilot
With the development of Autopilot, there were some important learnings for AI. “No single model can handle the many different use cases,” said Acosta. “Because of this, we created a multi-model architecture that adjusts in real time. For example, if there is large amounts of data, a deep learning algorithm might be best. But if not, a more traditional model could be better.”
Regardless of the technical details, Autopilot does point to the incredible power of AI and how it is poised to upend the software market.
“The potential of AI to transform communications is huge, but there’s a big delta between that potential and how companies are actually using it at scale,” said Acosta. “So at Twilio, we are focused on the building blocks that customers need to put the innovation to work.”