Last month Intuit revealed GenOS, an operating system that helps Intuit’s technologists “design, build, and deploy breakthrough generative AI experiences through Intuit Assist”, the company’s new generative AI assistant.
Intuit is the first fintech company to talk about using generative AI as a foundation for building an operating system. It is not the only company to talk about integrating this technology into accounting software, however. In the same month, Oracle NetSuite revealed how it had built generative AI into its cloud ERP to automatically create item descriptions and fill other text fields.
GenOS was the result of years of work, said Alex Balazs, Intuit's CTO, and was built on top of a huge quantity of accounting, financial and marketing data amassed from its 100 million customers.
Intuit promoted GenOS as the engine for its AI assistant Intuit Assist, which it said would be a “huge accelerant for innovation”.
“We are building generative AI experiences at an unprecedented speed as a result of our data, our talent, our platform, and GenOS. This is a transformative leap forward that will power unparalleled benefits, and this is only the beginning of innovations using generative AI,” Balazs said.
Here’s what we know about GenOS so far.
It’s big. GenOS is built on 60 petabytes, or 60,000 terabytes, of customer data. Intuit imports 20 billion bank transactions annually, many of which are between its own customers. This is a treasure trove of information that it can mine for insights, trends, and alerts.
It’s detailed. Each business stored in the GenOS database is tracked with more than 500,000 customer and financial attributes, or data fields. Each individual customer has 60,000 tax and financial attributes. This gives Intuit an enormous number of ways to segment its business and consumer customers, either for analysis or to make personalised offers at the right time.
It’s faster to deploy. GenOS is a “modern, fully cloud-native platform” that has accelerated its software development – Balazs claimed they had increased their velocity by nine times in the last three years.
It’s complex. GenOS has 2 million models running in production, 65 billion machine learning predictions per day. I have no idea whether that’s a lot compared to Oracle NetSuite, but it does show that Intuit is operating and testing these models at scale.
How does Intuit GenOS work?
GenOS has four core components.
GenStudio is a dedicated development environment where developers can rapidly experiment and refine generative AI experiences, including: fine-tune prompts, connect capabilities, and leverage existing data and code.
GenRuntime is an intelligent layer that can access the right data and platform capabilities, choose the right large language model in real time, and orchestrate and execute an action plan personalised for the customer
GenUX is a library of consistent customer interfaces and flows. This ensures “a clear and delightful” generative AI experience.
Financial large language models (LLMs) that are fine-tuned to solve tax, accounting, personal finance, and marketing challenges. Intuit trained its own LLMs which operate on Open AI’s foundational model, the same used by ChatGPT. GenOS uses multiple LLMs so it can pick the best model that will do the best job for the customer.
What does Intuit GenOS do?
Balazs gave an example of how GenOS works with its email marketing platform, MailChimp, to create an email campaign.
Intuit’s developers use GenStudio to write code to answer these types of requests by fine-tuning prompts, connecting capabilities, and leveraging existing data and code.
Once the request has been sent, GenRuntime creates a unique and tailored execution plan. In this case, it creates a market segmentation, drafts an email campaign, and sets up automatic reminders.
These results are presented to the customer in a consistent, clear way using common elements from the GenUX library.
GenOS selects the best large language model to execute that plan.
Embracing AI for developers
On a related note, Intuit also mentioned that it had deployed AI-powered tools for its developers to increase the speed of software development.
AI-assisted code development: Constructs code using a few descriptive prompts in natural language
AI content generation: Translates tax documents and other rules-based processes into code
AI-assisted data analysis: Facilitates quick data discovery, query generation, and query translation
These tools have had a significant impact, Intuit claims. It says it is translating tax rules into code up to six times faster, integrating new code four times faster and discovering and analysing data two times faster.
Is generative AI a fundamental technology to the next generation of accounting software? Or are we at peak hype?
Until we see Intuit Assist in action and the “revolutionary experiences” that Intuit is promising, it’s impossible to validate. However, clearly Intuit believes in this technology enough to put a lot of effort into branding itself as a leader in this space.
What isn’t in doubt are the resources Intuit is throwing at this. It’s spending a huge amount on R&D, it definitely has one of the largest and richest data sets, and all product leaders from the CEO down are talking about generative AI driving a fundamental shift in how we run our businesses.
Image credit: Intuit
Subscribe to our newsletter
Subscribe to receive the latest stories and new guides to your inbox. No spam, we promise.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.