What will programmers do when programs program themselves?

Everybody likes to say that AI is the future and that it's going to change everything. But ask a follow-up question or two, and the answers tend to be vague and light on concrete examples of what AI is actually doing to them today.

But for Intuit CTO Marianna Tessel, AI is more than just a buzzword; it's something she's trying to imbue into the core of the business. Tessel said she sees the shift to designing products and building with AI to be as fundamental as the shift from assembly programming languages to the modern higher-level programming developers use today.

In the future, as we've started to see with projects like OpenAI's GPT-3, developing software could be as easy as having a conversation with a computer. But how does the role of the developer change when programs can program themselves?

Protocol spoke recently with Tessel about how AI will change the way developers work, how companies run, and how customers are treated.

This interview has been edited for length and clarity.

How have things been going during the pandemic?

We are coping with the reality of working from home, and just like everybody else, we've been working with the trust that comes with it, and our employees trying to figure out how to work in this environment. We obviously shifted very quickly into this world, but what was really nice to see — because we work with small businesses and consumers and their financial lives — was the rally the company had around it. And much of it was coming from the employees. People were like, "What are we doing for small businesses? What are we doing for employees?" I counted over 40 efforts across the company that had to do with anywhere from the PPP made available to our customers (and they have raved of how fast it was to get it) to GoFundMe campaigns.

We actually use some of our knowledge engineering capabilities to be able to present to users what they're eligible for. We launched in April [what] we call Intuit Aid Assist. It applies to the CARES Act and hundreds of pages of regulations and very quickly directs you to what you're eligible for. We built the knowledge engine, which takes complex rules and turns them into code.

And we're just like everybody else; whatever stress the world feels related to the uncertainty of needing to work from home, we feel.

Has this time and what you're hearing from those small-business customers affected the way you're thinking about products moving forward?

Yes, absolutely. Like everybody, the pandemic set off innovation that would have happened anyway, but now it's like they're on fast-forward. We're definitely looking at how this is changing our customers' reality and then how are we going to be there to help them.

We are realizing that everybody is going virtual much faster, and that's something that's very much on my mind. We already had a service where we allow you to connect with experts through video and chat. We're leaning into those capabilities even more: We have something called the virtual expert platform. We helped a lot of our tax customers, providing them help online from the comfort of their home. It's a service we already had and we just made it easier for customers to get access to it. The reality is, everything is going virtual. The need for money here and now is becoming more important.

I'd love to hear more on your thoughts about this kind of AI approach to app development. I've heard a lot recently about opening up development to other parts of a company beyond the IT department, through the low code-no code movement. And then there's things like what's happening with GPT-3. It seems more like sci-fi than real life. Do you see us getting to some future where computers write their own programs or less work is required by the IT departments to get this work done?

You're right, there's very much a sci-fi narrative that has to do with AI, but AI is important, and we're just now seeing the impact it has on everything. This is an area that I myself started to be more and more fascinated with, because the more I see how we can use it and how it changes how we do things, the more I understand the potential.

I'll give an example: We're using knowledge engineering to take hundreds of pages of regulations, turn them into code, and then on the results, what we're able to do is just ask you the super relevant questions for you, and then all the information that we can deduct ourselves, we just do it ourselves. So the experience as a customer is a lot more personalized. And then as a developer, the way I write the application needs to be completely different because I can no longer write that rigid flow basically thinking about the customer going from A to B to C to D. Because for this customer, maybe it makes sense to go from A to D and for this customer it makes sense to go from B to Z. So the experience could be very different for different customers. We need to think about how we write the front-end experiences in a way that is a lot more dynamic and allows hyper-personalized experiences for users.

Another thing you've touched upon is that there's a lot of efficiencies [with AI], and the back end of these applications is also changing. As an example, let's take something like observability. With observability tools, we're able to understand different back-end flows and to predict incidents before they happen and do things that I would think several years ago would seem like sci-fi.

And it's almost as dramatic in my eyes as the shift from needing to write assembly to higher-level programming languages. We're going to see more and more of that shift.

Do you think that the role of the developer changes when AI can code things itself? Does it broaden who can build things, or is there always going to be a specialized cadre of people that do the development work for a company?

I think there are certain parts that you are going to need knowledge. For example, now, to write AI algorithms, you need an understanding of these algorithms, you need understanding of specifics, you need data scientists. So I would say yes, the role and the skills of developers is going to evolve in that they need to understand AI. People now say "mobile first" — we've been saying there's "AI first." I think all developers will need to have certain knowledge of AI, and we need to have this AI-first mindset.

But what you've touched upon is that with the introduction of AI simplifying and automatically generating things, you could see, and we ourselves are experimenting with it, the creation of different flows actually done by not engineers. If you're able to tell a computer, create this flow for me, that looks like this, and just say like that in English and it's created, that's not something you can do today. You have to program it. That is going to be available to a lot more people.

I feel like the shift has already begun, but when do you think that this will hit the mainstream? When do you think this sort of thing will be as common as a front-end developer knowing JavaScript today?

I wonder if we are already there to some extent. Maybe there's different companies that feel differently, but for us, we actually declared our strategy as an AI driven expert platform. And we are having conversations about the impact of AI already within every team.

But I do think that this process will take years. It reminds me of Moore's law. These things have this kind of exponential shape to them where the pace is picking up as you go. If we have this conversation five years from now, I think we would be blown away by what's available.

Are there other areas where you're pushing ahead with AI more than others?

We actually feel that this is a match made in heaven type of thing, because we feel like AI can really help our customers in a meaningful way. The problems and needs customers come to us [with] are complicated and highly personalized to them. A QuickBooks or a small-business customer has needs that have to do with their financials, that's highly personalized to them. What's true to them is not true to another company. And we might do a disservice if we just do things that are generic.

One example in this area would be eliminating drudgery in work. We know that our customers have to enter receipts, get forms, categorize transactions — a lot of that needs to happen as part of your day-to-day running of your financial lives or your business. We have used AI to automate a lot of that and eliminate work for our customers so they can focus on what they love to do. If you open a flower shop, you want to focus on selling flowers: Your passion is around flowers, it's not around paperwork, right? [Another] area is actually using AI to help you manage your business. Being able to predict your cash flow is a lifeline for any business, especially at the start, and we've been using AI to help customers understand where they stand and what they can and cannot do.

We also use a lot of AI in our back-end system, making our products more reliable, helping experts be expert on our platforms, helping them be more accurate and more productive. And this is evolving. We're not completely there. We still see a way to go, but we're creating more and more experiences that allow you to really understand where your business stands and what you can do.

One of the good things we've done at the company is actually define what AI means for us. It means machine learning, knowledge engineering and natural language processing. And that kind of definition really helped us just be very focused where we apply AI and the relevancy.

The deadline for the extended tax year recently passed. From a tech perspective, how does Intuit get beyond the issues it's had in the past with obfuscating the free filing service? How do you build a more transparent structure moving forward?

Intuit is committed to offering the free tax preparation service as demonstrated by more people filing their taxes for free with TurboTax than all other tax-prep software companies combined. So that's where we stand.