Puzzle

Rather than treating bank feeds as an input to manual coding, Puzzle runs AI agents that draft the work and hold it for approval. Categorization, reconciliation and close entries arrive already prepared, and the accountant reviews exceptions instead of touching every line. The AI Close agent builder lets a firm encode its own month-end checklist so the same steps run for every client on the same schedule. Feeds come in natively and read-only from Stripe, Brex, Mercury, Ramp and Gusto, which is a deliberate choice of stack: this is a ledger designed around how venture-backed companies actually spend money. Cash and accrual are both supported, with revenue recognition on top.

Puzzle is explicit that firms are its channel rather than its competition, and the positioning is a direct swipe at Intuit. Its accounting-firm page argues that "unlike QuickBooks, which offers its own bookkeeping and accounting services that directly compete with accounting firms, Puzzle is built exclusively to support firms and their clients." Firms get a revenue share and client discounts, a listing in a find-an-accountant directory, a partner rewards program and certification training. The percentages are not published, so a firm has to ask. Puzzle cites 7,000-plus customers and partner firms scaling from 10 to 30 clients on the same headcount, with month-end close reported as 25 percent faster.

Published pricing is per entity and billed annually: Starter $25, Core $60, Complete $100 and Scale from $150 a month, with monthly billing roughly 20 percent higher. Complete carries a money-back guarantee tied to cutting close time by half within 60 days, which is a confident claim and also a signal that onboarding effort is the real risk. Accounting-firm pricing is quoted separately.

The gaps matter. Puzzle's own materials do not describe a true multi-tenant console for running 100 client files with staff assignment and workload views, which is what a firm past a certain size needs. The integration list is narrow and startup-shaped, so a construction client on a regional bank with no Ramp card is a poor fit, as is anything needing inventory or job costing. The security page documents AES 256-bit encryption, US servers and read-only integrations but does not state a SOC 2 report, so that is unconfirmed here. And the whole model depends on a firm trusting AI-drafted entries enough to review rather than redo them, which is a cultural change more than a software one.

Kick is the nearest competitor and undercuts Puzzle with a free tier and cheaper multi-entity pricing, but is younger and smaller. Digits sits in similar territory. Against QuickBooks Online the trade is automation versus ecosystem: Puzzle removes more manual coding, QuickBooks connects to everything and every hire already knows it, which is why Puzzle invests in a white-glove QuickBooks migration.

Strong-option is right and it stays. The install base, the published prices, the explicit firm channel and the QuickBooks migration path are more than most AI ledgers can show. It is not a top pick because the client profile is narrow and the firm-side console is thinner than the marketing implies.