We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 – 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
Completing a customer’s purchase order correctly and answering requests, questions and comments of all types is difficult enough for an enterprise to do on an individual basis. Doing it millions of times a day seems impossible. But that’s what thousands of companies with online sales and marketing arms do on a 24/7 basis.
The capability to provide personalized experiences at scale is complicated. Doing this using unique models for each customer is even more so. However, using artificial intelligence (AI) and machine learning (ML) in the right instances can go a long way toward helping companies satisfy all their customers, Intuit CTO Marianna Tessel told the audience today at VentureBeat’s Transform 2022 conference here at the Palace Hotel.
Intuit has been able to accomplish this through accelerated – and sometimes uniquely built – AI, allowing its IT system to provide these benefits to customers and small businesses.
Categorizing data for AI at scale
“One of the things we’ve been doing a lot of lately is helping small businesses categorize their data,” Tessel said. “This helps us use what we call ‘AI at scale.’ We apply AI to a lot of our people, and it looks different in every case. We actually have 2 million AI models in production that are refreshed daily to achieve the level of categorization we need to be effective.”
Intuit started with one model years ago and kept adding data to it as more people became customers. This involves a lot of minutiae data that serve to define personalities. “For example, one person (who owns a gas station) may tend to use the term ‘gas’ and another station owner may use ‘fuel’ – it’s differences as small as that make these AI models personal that are used in our production,” Tessel said.
“Each one of our small business customers is unique and passionate about what they’re doing; they’re all individual snowflakes. They want to call things the way they want to call them – they don’t want us to force them into a particular way of categorizing,” she said.
Thus, Intuit focuses on the data language of its many customers. This is why it has so many distinct AI data models running in its systems night and day.
Why Intuit chose AWS
Intuit is an AWS shop, enabling the company unlimited cloud-computing scale.
“AWS enables us to bring in the (compute) modules that we need as we need them, which works for us,” Tessel said. “At the beginning, we were using our existing tools and servers, but it became cost-prohibitive and the complexity was heavy. We decided on a combination of automation and execution using AWS, but we also weren’t afraid to build some things ourselves as we needed them.
“Don’t think about AI as just developing the AI and the models; but somewhere you have to say, ‘How can I capture the uniqueness of customers in what I’m trying to do?’ That would mean building some of your own tools. That’s one of the lessons we learned.”
When the company does as much customization as it does, how does Intuit understand the meta trends from across its customer base?
“You can use the data to funnel it once for a very personalized model, then you can funnel it again for discovering the trends across, so you can learn from that as well,” Tessel said. “You can cut the data many different ways, even for the same use case or same customer. That’s the way we do.”
VB’s Transform 2022 continues online through July 28.
Author: Chris J. Preimesberger