AI-native general ledger solutions provider Digits announced the release of its AI Firm Models—available exclusively through the new Digits Accountant Partner Program—which are meant to adapt to accounting practices by learning continuously from client data, staff workflows and firm-specific processes.
Digits CEO Jeff Seibert, in an email, described Firm Models more specifically as predictive machine learning models that learn from a firm's work across all its clients in Digits and apply those patterns to streamline and automate future tasks. Built, trained and run on Digits' infrastructure, each model is uniquely created for and available exclusively to the firm it serves. AI Firm Models are not a new product offering or an add-on to a current product but, rather, an inextricable part of the Digits platform.
Seibert said that partner firms train their models by onboarding their clients onto Digits and performing cleanup and monthly close work on the Digits ledger. As firms take actions within Digits, the model learns immediately and is able to mimic their work as new transactions arrive across their clients. While this is conceptually similar to how Digits' other models are trained, what's different is the ability to isolate specific actions taken by the firm and its downstream clients to train a dedicated model particular to the firm's own practices.

"You can think of the Digits AI architecture as a layer cake. At the top are company-level models. These train only on the financial transactions of each individual business or client. Once Digits has seen a transaction pattern for a specific client, it is extremely accurate at replicating that as new matching transactions come in," said Seibert.
Usually, when a transaction comes in that doesn't match the previous pattern on the company-level model, it is passed on to Digits' global model, which Seibert said has "vast coverage and very high accuracy," but lacks the ability to capture the unique industry or geographic expertise that accounting firms bring to the table for their clients. The Firm Model is meant to address this.
"By introducing this new tier of firm model, we offer partners what they've been asking us for. Their work on one client now automatically benefits all their other clients, imbuing their best practices across their client base without any manual rules to configure and manage," he said.
The Accountant Partner Program, through which firms can get these new models, offers ongoing training, support and enablement for partner firms, both for their model, as well as their overall usage of Digits with their clients. Beyond the models, participants in the program will also be able to access streamlined client onboarding and centralized client activity with real-time visibility and staff access controls; flexible wholesale pricing that scales with client growth, plus free use of Digits for each firm's own books; hands-on onboarding, client migration support and a dedicated success team; and AI-native workflow training for staff at all levels, plus a Coaching Certification program for firm leaders. Digits will work closely with partner firms on an ongoing basis to tune their models and drive accuracy improvements.
The firm model is available to firms of any size that serve at least 50 clients on Digits, with Seibert explaining that transaction volume ensures the model will have strong coverage and accuracy. He added that the model continues to expand and get smarter with every additional client added onto Digits, which makes the offering more powerful for very large firms. For firms that serve multiple distinct sets of clients in different practice areas, Digits will offer the option to train individual firm models for each practice area (of 50+ clients), so the patterns and expertise don't get diluted or confused across the breadth of the firm's client base.
"We're ushering in a new era of accounting," said Seibert in a statement. "Our Accountant Partner Program gives firms direct access to the latest in AI technology—such as our new Firm Models—tools designed and built to automate the tedium, unlock deeper insights, and deliver on the promise of AI-native workflows."