Digits and CLA will train firm-specific accounting AI for thousands of clients

Why it matters

CLA gives Digits distribution into live accounting workflows and the corrections needed to improve a firm-specific model. The test is whether those gains arrive with auditability and client-level data controls.

Jeff Seibert (@JeffSeibert) has secured a large accounting-firm test for Digits, the ledger software company he founded in 2018. The bet is that accounting firms will replace generic bookkeeping automation with AI trained on their own judgment. CLA and Digits said in a Tuesday announcement that they will co-build a proprietary model, starting inside CLA's Client Accounting and Advisory Services practice and expanding to thousands of current and future clients over three years.

Seibert founded Digits in 2018 after building two developer-focused startups with successful exits. Box acquired Increo in 2009, and Twitter acquired Crashlytics in 2013. He graduated from Stanford with a computer science degree, according to a biography published by Stanford's Technology Ventures Program.

That technical background shaped Digits' founding thesis. In a 2024 essay, Seibert described running startups with live product and engineering dashboards while waiting two or three weeks after month-end for static financial statements. Digits set out to rebuild the underlying ledger so transactions, reconciliations and reports could update continuously rather than adding an assistant to older accounting software.

The deal gives Digits a large accounting channel

CLA gives Seibert access to the operating workflows that matter most for training accounting software: repeated transaction reviews, categorization, corrections and closes. CLA says its client accounting practice represents roughly 15% of more than $2 billion in annual revenue and works with tens of thousands of privately held businesses. The initial rollout will be offered within that practice's service levels, with CLA supplying the accounting knowledge and Digits supplying the ledger and model-training technology.

The arrangement also puts Digits inside a service line where software adoption can spread across many client books through one institutional decision. Each correction made by a CLA professional can sharpen the CLA-specific model. Digits describes its accounting-firm product as learning from a firm's prior work, then applying those patterns across its client portfolio.

Digits publicly launched its Autonomous General Ledger in March 2025 and expanded accounting-firm platform partnerships in June 2026. The June release said Digits managed more than $850 billion in financial activity across hundreds of accounting firms and thousands of businesses. CLA brings a national professional-services workforce and a client base measured in tens of thousands of privately held businesses.

For CLA CEO Jen Leary, the partnership is a bet that CLA can package its institutional knowledge into software without removing professionals from the client relationship. Leary previously served as CLA's chief strategy officer and became CEO in 2021, according to CLA's CEO appointment announcement. "Most of the companies we evaluated had tacked AI onto decades-old software," Leary said in the announcement.

CLA says the goal is to shift employee time from transaction categorization and document collection toward review and advisory work. Automating preparation could let CLA serve additional clients without increasing bookkeeping labor at the same rate. The announcement says CLA will continue hiring and developing professionals, though it does not provide staffing projections or productivity targets tied to the Digits rollout.

What CLA's proprietary model means

CLA and Digits are using "proprietary" to describe a firm model trained on CLA's client base and workflows. The announcement says Digits will provide the platform and model-training technology while CLA supplies accounting knowledge, but it does not say whether the CLA layer is a separate model, a fine-tuned component inside Digits' ledger, or another implementation.

Digits' public positioning is that its AI-native ledger supports business- and firm-specific models and month-end close automation. The CLA announcement provides no independent security or model-performance assessment.

The CLA announcement does not explain the model architecture, client consent or opt-out controls, retention rules, cross-client learning boundaries, mandatory human-approval points, accuracy benchmarks, error rates, auditability measures, or commercial terms.

Those questions will determine whether Digits can turn CLA's professional judgment into repeatable automation while preserving the controls expected of accounting providers. Digits says corrections sharpen future predictions, which creates a useful feedback loop. It also puts governance at the center of a deployment expected to learn from work performed for restaurants, contractors, manufacturers, nonprofits and other privately held organizations.

Accounting firms are becoming AI distribution partners

Digits is competing in a well-funded market where accounting providers increasingly serve as distribution channels and trust anchors. Rillet announced an alliance with EY in April 2026 for AI-native finance and controls, while Basis raised a $100 million Series B in February 2026 to expand accounting agents used by professional firms. Puzzle and DualEntry are also building AI-driven close and ledger products.

Digits entered that contest with substantial backing before the current AI accounting rush. In March 2022, Digits raised a $65 million Series C led by SoftBank Vision Fund 2, with 20VC Growth, GV and Benchmark participating. Digits said the financing brought its total funding to $97.5 million and valued it at $565 million. No financing was announced alongside the CLA agreement.

CLA gives Seibert something capital alone cannot supply: a recurring stream of real accounting decisions made inside a practice that serves tens of thousands of privately held businesses. Over the next three years, Digits will have to show that those decisions can improve automation at scale without turning professional judgment into an opaque output that clients and accountants cannot inspect.