Main content

Twilio CEO Jeff Lawson seeks signal among the AI noise – with Sam Altman’s help

Chris Middleton Profile picture for user cmiddleton August 25, 2023
Twilio’s conference was buzzy and on the money. But it really belonged to OpenAI…(and don't mention CRM!)

Jeff Lawson

For Twilio co-founder and CEO Jeff Lawson, it was all about the numbers. Bounding onstage to deliver his keynote at Signal this week, the comms API provider’s customer and developer conference, he told delegates:

In just the last year, your apps have made over 34 billion voice calls, have sent over 157 billion text messages, and over 1.7 trillion emails, and you've done over four and a half billion digital verifications of your customers. 

And in the contact center, you all have handled 850 million customer interactions. In total, that's more than two trillion interactions that you’ve had with your customers. That is amazing!..And we've ingested over 12 trillion data events on behalf of your applications.

Impressive stuff. But in 2023, the core topic of conversation from a California conference stage was, of course, generative AI. And how, without it, life just sucks. Lawson said:

We've got over 300,000 customers here at Twilio. Companies of every shape and size and every industry: start-ups, Fortune 50 companies, and everything in between. And we've been having a lot of conversations with customers about how they're thinking about the impacts of AI...

Veteran IT conference watchers knew what was coming next: an overarching keynote metaphor, one that subsequent speakers could riff on (and duly did). Lawson said:

Really, it goes back to what it takes to build a great digital business. And the most common thread I see among great digital businesses is that they're all trying to build this flywheel.The flywheel starts with data, all the signals that your customers give you about who they are and what they're interested in. Every website visit, every mobile app, every click, every scroll, everything they buy, and everything they don't buy. Those are all signals that tell you about the customer and what they're all about.

He continued:

And then you use that data to do a better job of engaging with those customers, making yourself more relevant, more timely, and sending them things they're more likely to enjoy. And when you do, they are more likely to click that email or open your mobile app. And suddenly you start getting more data about them and the flywheel turns, right?  This is what great digital businesses know. And you know what's interesting? No matter what your job is, you're actually a part of creating this flywheel. Even if you don't know it.

But there are problems, he added. One, your flywheel gets stuck in a silo (centuries-old foe of flywheels, it seems). And two, all the stuff that Amazon is great at is slow and prohibitively expensive for all but a trillion-dollar corporation that has cracked logistics. Until the advent of generative AI (get the picture?). Lawson said:

Every time a customer hits one of those silos, the flywheel kind of stops spinning, right, because it gets hard and there's friction, and it's not good, you know? [Sad face] And then what do all those teams do? Well, they go buy software to help them do their jobs. And so, the marketers go buy the marketing systems, and the salespeople go buy the sales systems, and the contact centre people buy that system.

And so, what have we done [in that scenario]? We've taken all that software, and we've actually just firmed up all those silos and made it even harder to serve our customers, because now our systems are fighting us too.

Then came the gut-punch of his argument: the tech that is nothing but a malign influence in enterprises’ lives:

You know what is supposed to be solving this problem? Well, I bet three letters come to mind: C… R…. M. […] But it turns out, CRM is where hopes and dreams go to die.

Here's the AI

So, what’s the solution? Its 2023, so it must be generative AI. And this year’s uber-uber-app is here to make sure your customer never, ever, gets a moment’s peace again. Said Lawson:

If you had 10 million customers, what would you need? You’d need 10 million employees. Each one paying attention, laser-focused on one customer. And how would this work and what would those magical employees do? Those magical employees would use their human abilities to actually solve all these problems I just talked about.

And what are humans really good at? Well, first of all, we're really good at perceiving the world around us. […] I see this stage. I see what's going on. And I know that if I take one more step, I will fall off the stage. That's me understanding the world around me.

But hiring 10 million employees to look after 10 million customers “is a ludicrous value proposition, right?” asked Lawson. Wrong. Because AI can be those 10 millions workers for you. 

At this point, OpenAI CEO Sam Altman faded into view, albeit in a slick video two-shot with Lawson, who said:

Sam said something truly interesting. He said, ‘What if AI is able to bring down the price of intelligence by a factor of a million?’ Think about that for a second: bringing down the cost of intelligence by a factor of a million. What would be possible? Well guess what? You would suddenly be able to have that magical employee looking after every single customer, using the intelligence that human beings are able to bring to the table. And actually close the loop and spin that flywheel.”

Holy metaphor pile-up! Then came the moment:

And we call this ‘CustomerAI’. Something that can truly change the nature of business.

Cue whoops from the audience. Because CustomerAI is Twilio’s... well, its customer AI. Based on ChatGPT. 

Interviewing Altman in the subsequent onstage video, Lawson congratulated the OpenAI supremo for igniting developers’ energy and enthusiasm. Altman said:

Thanks for saying that. But it really goes the other way around. It's the developers doing amazing stuff that ignites the energy for us. And it's just been amazing to see.

I think most of all, what I would say is it's the breadth and depth of what people are building. Some areas I've been particularly excited about are what people are doing in education, building these AI tutors to help people learn different things. I find them super-compelling.

Educators seem to find generative AI less compelling, as seen recently. But Altman continued:

Healthcare, that's still early. But we're seeing really promising results. And then this idea of what I call AI-enhanced productivity. Some people call it co-pilot. The idea that you can make developers way more productive, and you can also make people in other lines of work more productive with similar approaches. But again, it's really the incredible breadth I'm so excited about.

Then the discussion became interesting, as it focused on what we should do now with ChatGPT and similar tools, especially when they don’t meet our needs and expectations. Altman argued:

Sometimes the model doesn't do something quite right. And people put a lot of effort into a very specific prompt, or into layers on top of it to get it to do something. And we recommend against doing that. Because with the next turn of the model’s crank, [GPT] will just get smarter and smarter. 

So, trying to augment the fundamental intelligence of the model is, I think, a lot of work for low reward, because you should just bet on the model’s improvement.  And because you should bet on the model’s improvement, I think it's interesting to think about things that are a couple of years away from a capability perspective. What is GPT-5 or -6 likely to be able to do? And what will you be able to do with that? How can you start thinking about that now?

So, it’s okay to do things now that don't scale – if you assume that a smarter model will fix the problem. As the network effect gets stronger, as the compounding advantage gets stronger, as long as you're thinking about that and viewing AI as an enabler, you can bet on that it's just going to keep getting smarter.

Fascinating, as the subtext would appear to be: OpenAI is in charge. Just wait for it to innovate. A company that is positioning its customers’ dependency as a sales strategy (and doing it well). At this point, Lawson asked if this moment is as big as those others, like PC to Web, or Web to mobile?

Altman answered:

My rough guess would be, when we look back at this, it [generative AI] is bigger than mobile and about the size of the Internet. I may be totally overhyping it, but that would be my guess.

My take

So, there we have it. With Altman’s appearance at another vendor’s conference – a good event and a fun one, with great work being showcased – you could feel the balance of power shift in the IT industry. The conference was Twilio’s. But the takeaway? OpenAI is a supermassive black hole. And the universe is bending towards it.

A grey colored placeholder image