The treasury and accounting software company DebtBook has launched two new artificial intelligence features, giving users a daily briefing and dashboard as well as an AI analyst that can answer questions phrased in plain English.
Insights, the AI-powered overview, allows the government and nonprofit organizations
The feature includes a daily briefing, top-line key performance indicator, or KPI, cards and domain views across cash, debt, and investments. It also offers AI-generated nudges about what areas need attention, according to a press release about the product's rollout.
The metrics within Insights are computed deterministically — or consistently and predictably, rather than probabilistically — from customer data, and are not AI-generated, according
Marty is described as an AI analyst that answers questions about an organization's own data in plain language. Its answers include as-of dates and links back to the in-app views behind them, so that users can see how they were derived.
"Most of the products that we develop, they're really built out of customer feedback," CEO Tyler Traudt said in an interview. One common theme of that feedback, he said, "is that a lot of this financial information lives in many different systems across different teams."
He added, "Think about that problem within just one function, debt management. Well, it exists across cash and investment management too, and now you really have a multitude of these individual systems (with) the siloed information."
Traudt said the team encountered some hurdles in developing the new features — "anytime you are trying to combine a probabilistic technology like artificial intelligence with deterministic systems, you can run into challenges," he said — but they understood that two certainties had to be present.
One, "our customers have to have software that they trust that performs the same calculation the same exact way every single time," he said.
And two, "you want to use artificial intelligence to maybe move faster or derive insights that you otherwise wouldn't be looking for," to capture some of the efficiency gains of AI.
One of the biggest challenges was marrying those two requirements, Traudt said.
Ultimately, "every every piece of analysis performed inside of our software is going to use audited, deterministic business logic," he said. And the AI "will highlight things that you might care about… and it will answer your questions."
Users can also audit every calculation surfaced by the AI features, he said, so that "if you want to trust but verify, which I think makes tremendous sense," you can.
Traudt said he doesn't foresee a world in which government or nonprofit finance officers turn over responsibility for decision-making to an AI agent, but public finance can catch up to the private sector more than it has.
"Those teams are quite deserving of more powerful, more modern solutions that can help them do their jobs more effectively," Traudt said.
