The right way to think about AI - Twilio’s Chief Product Officer Inbal Shani on the difference between successful and not successful companies

For the past nine months, Inbal Shani has been the Chief Product Officer and Head of R&D at Twilio, a customer engagement platform (CEP). She's already working hard to build a trusted platform that leverages AI to support a constantly shifting customer engagement market.

Shani previously worked at Github, where she was responsible for what is now called Copilot (along with other platform elements). She has over twenty years of experience in AI/ML and has been through every phase of AI growth, from neural networks to deep learning, predictive, and now generative AI. Shani's experience spans industries including aerospace, automotive, robotics, retail, cloud developer tools, and now communications and data. In all her work, she has looked at how AI can improve the business.

Twilio has created trends and followed trends. But as it looks to the future, it's all about building "one platform that is trusted, simple, and smart.' Shani argue that every company has to leverage AI, but many are taking the wrong approach.

The questions companies do ask focus more on AI and how to monetize it. The questions they should be asking are: what is the value-add that automation or having a conversational agent brings to the customer experience, and how are you using AI as a tool? How will AI make you more productive, remove friction, or provide a better customer experience?

It’s the difference, she says, between successful and less successful companies. If you are grounded by the customer problem you're trying to solve and figuring out how AI can help solve that problem, you will realize ROI and see successful adoption. Companies that sprinkle a little generative AI on something that doesn't materialize into a product or feature that customers want to adopt and buy will struggle.

The other challenge she sees is that the AI strategy doesn't necessarily align with the company strategy.

I think that's the biggest kind of difference in terms of perspective on AI; AI is a very useful tool. It's a very powerful tool. And you need to know how to use these tools to solve problems. But it's not a magic solution for everything, and it's not a magical solution that is going to solve everything. So, knowing how to use it in a smart way to really focus on the customer. If it's better user experience, if it's better productivity, if it's creating shorter time to value because of automation, then you have a successful AI strategy.

Shani says Twilio is successful because it uses AI as a tool to improve customer engagement, security, and customer satisfaction.

How Twilio CEP uses AI

Twilio is not for one department. It provides the tools to support the customer end-to-end. The CDP (customer data platform) gives them customer data, and the CPaaS (communications platform as a service) provides the "last mile connection" - email, SMS, and voice interactions. Bringing them together creates customer engagement across the entire customer lifecycle. AI is used throughout the CEP. Shani cites the example of Fraud Guard, Twilio’s first product to market, which provides SMS protection. They developed predictive AI models for this product, which they continually tune:

In Twilio, for example, in our Fraud Guard product that I mentioned before. It was never about, okay, how do we monetize predictive AI? We were grounded in solving a problem that was real and increasing, especially now in the age of AI, of spam, and impersonation. So when we built Fraud Guard, it was always the intention: how do we protect our customers, how do we create these secure communications, and we see real value? We see real savings for our customers who are using the product. And they never once asked us, are you using LLMs? Are you using predictive AI? It was not even a question. So, it was not about how you're monetizing it. But it's more in terms of what is the value that I'm getting?

But Shani is also looking into leveraging AI in the Segment business, focusing on customer data and engagement. Here, predictive audiences help companies understand what their customers are trying to do, how they will behave, and what the next best action for them is.

In addition, Twilio has launched generative AI products across communications, including their personalized IVR and Flex contact center product with a copilot agent. In this product, an AI assistant helps agents be more helpful to customers by providing advice and recommendations, sentiment analysis, and summarizing conversations.

What’s changing for developers

With a ten-million-strong developer community, it makes sense that Shani would have a strong view of how AI is affecting developers.

We are hearing more and more that in the next five years, everyone will be a developer. Shani doesn’t agree. She reckons there is a big difference between a developer and a builder. Yes, everyone can use an AI tool and build something, she said. You can write better content and better code, but that doesn’t make you a software developer:

It's kind of a new language that developers need to learn how to use and how to use it well.

Shani pitches that AI is a productivity and assistant tool, with rhe caveat that you need to know what questions to ask to get the right answer.

You see the evolution of software development languages all the time. And I think when I'm looking from a software developer perspective, what we see right now, with all these copilot assistants to write code, is really about making a developer much more productive so they can focus on getting the full value out of their time writing code which is something that they didn't have before.

Shani sees three themes in how AI is impacting software development. First, developers need enough software skills to validate that the AI-generated code is good and does what they want it to. 

Second, AI uses data to create code, but there’s a creative aspect to people that Shani doesn’t see AI replacing in the near future. 

Finally, Shani believes there will be a shift in how developers think about code. Before, it was all about generating code to solve a problem. Now, it’s thinking about systems design and overall architecture from the start. And it’s about user experience, something that is becoming very critical in the world of AI, she said.

What does this mean for developers exactly?

She explains: 

If we're using an AI tool to write a document, we don't just copy, paste, and call it done. The same needs to be done in code. You have to validate whatever the AI assistant has been generating. I think that's one element. And the second thing that I advise computer science students asking me similar questions is you need to learn the basics without AI. You need to learn how to be a software developer before you earn the right to use an AI tool because you need to be able to look into the code. You need to be able to look into the product that you have generated and know if it's good or not. Because, if not, and that goes back to then, we don't really need you because that's the validation that is missing if you delegate the entire creation of a code to an AI assistant.

The key to consider here is how students are taught computer science today. Not all universities have embraced AI trends—not for developers, marketers, or writers. Shani said she is seeing some effort to figure it out. But there is a balance to find. You want students to learn to code without AI first, but you also want them to learn to leverage AI tools to help them be more productive and efficient.

At the same time, these schools also need to devote more teaching time to the thinking and designing aspects of systems development. There aren't likely many schools teaching user experience in computer science courses—at least not as a core subject.

My take

I appreciated Shani's perspective on how organizations should think about AI. By focusing on the customer and creating better experiences, it's easy to find ways to leverage AI. Incorporating AI into a product without thinking about what benefits it truly brings may have been okay at first, but it isn't the right approach today.

I also liked how she talked about developers' use of AI. This is very similar to how content marketers and writers can use AI—as an assistant, as another tool. It doesn't replace the need to understand the work; it's simply a way to take more of the basic operations out of the hands of developers so they can spend more time designing products and experiences that customers will love.

Does this mean companies won't replace developers? Not necessarily. As I have said before, with AI and writers, some will get replaced. It's just the way things go. What developers want to do is find the companies that understand the true value of AI in the toolkit. And it sounds like Twilio is one of those companies.