When Intuit sent AI agents to 3 million customers, 85% returned. The reason, according to the company’s EVP and GM: Combining AI with human expertise proved to be more important than anyone expected – not less.
Mariana Tessel, EVP and GM of the financial software company, calls this AI-HI combination a “massive demand” from its clients, noting that it provides another level of confidence and trust.
“One of the things that we learned that has been fascinating is really the combination of human intelligence and artificial intelligence,” Tessel said in a new VB Beyond the Pilot podcast. “Sometimes it’s a combination of AI and HI that gets you better results.”
Chatbots alone are not the answer
Intuit – parent company of QuickBooks, TurboTax, MailChimp and other widely used financial products – was one of the first major enterprises to move to generative AI with its ZenOS platform Last June (long before the fear of "SaaSpocalypse" Are SaaS companies struggling to rethink their strategies).
However, the company soon recognized that chatbots were not the only answer in enterprise environments, and focused on what it now calls Intuit Intelligence. The dashboard-like platform includes specialized AI agents for sales, tax, payroll, accounting, and project management, with which users can interact using natural language to gain insights on their data, automate tasks, and generate reports.
Customers report that invoices are being paid 90% in full and five days earlier, and manual work is reduced by 30%. AI agents help close the books, classify transactions, run payroll, automate invoice reminders, and uncover discrepancies.
For example, an Intuit customer exposed the fraud after interacting with AI agents and asking questions about amounts that did not add up. “Initially it was like, ‘Is this an error?’ And as they investigated, they discovered very significant fraud,” Tessel said.
Why are humans still in the loop?
Still, Intuit operates on the principle that humans are “always accessible,” Tessel said. The platforms are built so that users can ask questions of a human expert when they aren’t getting what they need from an AI agent, or they want a human to bounce ideas off of them.
“I’m not talking about product experts,” Tessel said. “I’m talking about a real accounting expert or tax expert or payroll expert.”
The platform is also designed to suggest human involvement in “high-risk” decision-making scenarios. AI gets to a certain level, then human experts review and classify the rest. According to Tessel, this provides a level of confidence.
“We really believe that there is a greater need and more power for this at the right time,” she said. “Experts still provide things that are unique.”
The next step is to give customers the tools to perform next-generation tasks like Vibe coding – but with simpler architectures to reduce the burden for customers. “The idea we’re testing is that you can be coding without really realizing that’s what you’re doing,” Tessel said.
For example, a businessman who runs a flower shop wants to make sure they have the right amount of items in stock for Mother’s Day. They can vibe code an agent who analyzes previous years’ sales and creates purchase orders where stock is low. That agent can then be instructed to automatically perform that task for Mother’s Day and other major holidays in the future.
Some users will be more sophisticated and want the ability to delve deeper into the technology. “But some people want to express what they want,” Tessel said. “Because they just want to run their business.”
Listen to the full podcast to hear about it:
- Why might a first-party create data? "Chasm" For SaaS companies.
-
Why does showing an AI’s logic matter more than a sophisticated interface?
-
Why 600,000 data points per customer change what AI can tell you about your business.
You can also listen and subscribe beyond the pilot But spotify, Apple Or wherever you get your podcasts.
<a href