Puzzle partners with Decimal to revolutionize accounting firms, demonstrating how separate software and service companies can collaborate to empower accountants while countering industry threats from AI and tech giants.
Decimal reached out to us to let us know they were looking for a modern general ledger that could support hundreds of their clients at scale, but with one unusual request beyond conversion and integration to their firm management solution: would we work alongside their team to help them design the firm of the future?
TLDR:
Until now, AI in accounting has been mostly theoretical, with few examples of deep, meaningful integrations beyond small pilot programs. Why? Accounting firms require reliable, proven automation since their reputation depends on it.
Traditional barriers (AI hallucinations, opaque third-party workflows, and outdated accounting infrastructure) were showing limited automation advances. While adding third-party startups to legacy systems seemed like an easy solution that avoided process changes, it typically led to disappointing results with only minor improvements, discouraging further adoption.
Everything changed with OpenAI's reasoning models and Puzzle's modern general ledger technology. This created an inflection point where AI became truly ready for scale. Now, workflows operate reliably, reasoning is transparent, and Puzzle's design gives accountants precise control over AI capabilities.
Signing a contract is easy, while creating customer value at scale is the hard part. The opportunity to work with Decimal was an exciting request, and we wanted to learn more. They explained they had four main focus areas for their accountants and bookkeepers:
We spent a day with Decimal demoing Puzzle, discussing roadmap, and talking about our vision for the industry. We realized that while Puzzle had hints of these capabilities within the product, Decimal had some great feedback on how we could improve the product for them and set a new industry standard on profit margin while improving the client experience at the same time. We realized they were not looking for a vendor, they were looking for a partner into the future. So we agreed to a plan together.
Puzzle and Decimal built out four phases to go from discussion to launch.
| Phase | Focus Area | Key Deliverables |
|---|---|---|
| Phase 1 | Adapt Puzzle's design for the accountant | Customizable views; improved navigation; faster transaction & reconciliation interfaces with automated rules |
| Phase 2 | Build firm management | Keeper integration; native connections to Ramp, Plaid, Gusto & BILL; bulk client conversion; standardized COA & automation mapping |
| Phase 3 | Expand accuracy reviews | Smart checks for duplicate vendors & GL mapping errors; AI validation of the month-end close process |
| Phase 4 | Expand to advisory | AI advisory labs; variance reports with change identifiers; dual cash & accrual views |
Puzzle was originally designed for venture-backed startup founders. But it turns out that the same technology worked well for accounting firms and their clients with a few tweaks.
Puzzle worked really well as a self service accounting solution for founders, but needed to build additional tools for firms to manage hundreds of clients.
Puzzle developed a set of accuracy reviews, but as portfolios scale and accountants are taking on more and more clients, additional checks and balances were needed.
Expanding margin and improving the client experience means helping accountants make the transition from preparation to advisory.
Software and services have struggled in the accounting space due to the long tail of edge cases and accounting complexity, which is a binary offering: accounting is either right or not.
The recent insolvency of Bench and historical shutting down of Scalefactor, each of whom built impressive consumer offerings and raised over $100M of venture capital, are demonstrations that building complicated accounting software and running a services business are a challenging combination. But when these two industries collaborate as separate entities, they unlock extraordinary potential.
This partnership would allow Puzzle to focus on building software tailored to accounting firms, while Decimal focuses on providing high-quality services for their clients. Each company excels in its domain, creating a symbiotic relationship that benefits everyone, most of all the accountants they serve.
The future of accounting doesn't have to be written by the industry. By choosing partners who align with their success, accountants can reclaim their rightful place as trusted advisors to businesses. The key is recognizing that software and services work best together when accountants review and finalize what the AI drafts.
We want to show that Puzzle is more than a vendor, but a trusted partner in the industry, building collaboration and community while showcasing a commitment to driving meaningful change. We plan to build in public, sharing our journey of integrating and scaling with Decimal, in the hopes to build trust, transparency, and engagement within the industry. By openly sharing our journey, challenges, and progress, we will show our dedication to creating solutions that truly tackle accounting pain points, not in marketing but with evidence.
We welcome other accounting professionals and firms to join us in this journey, and share their own story of adopting modern tools to sharpen their processes and increase client satisfaction. The accounting industry is under siege. From the looming threat of AI replacing accountants to Intuit's aggressive television campaign encouraging clients to fire their accountants while simultaneously launching their own bookkeeping services. In our view, those moves put Intuit at odds with the very accountants who helped make its software the gold standard for small businesses.
The dynamics that made this partnership urgent haven't slowed. They've accelerated. The 2026 AICPA PCPS CPA Firm Top Issues Survey found that change management driven by technology and AI ranked #1 across every firm-size group for expected impact over the next five years. Meanwhile, the average number of open accounting and finance roles per company has jumped to 17 (up from just five in 2025, per Personiv's CFO Pulse Survey) as the talent pipeline continues to narrow. Firms that waited to modernize are now competing for the same shrinking pool of people.
On the AI side, the numbers tell a different story: 84% of U.S. accountants now use AI in some form, and firms that have moved to full AI adoption report 20-30% lower staff turnover. The profession isn't waiting for a future inflection point. It's living through one. Before committing, understanding AI readiness for accounting firms is the right first step: the firms pulling ahead are the ones that paired modern software infrastructure with services expertise, exactly the model Puzzle and Decimal set out to prove.
Year-end busy season is the moment firms feel the capacity limit most. The U.S. Bureau of Labor Statistics projects 5% employment growth for accountants and auditors from 2025 to 2035, with about 115,300 openings per year. Demand for accounting work keeps growing, and the pool of people to do it is not growing at the same pace.
That gap is where the Decimal model pays off. When a firm standardizes its chart of accounts, categorization rules and accuracy reviews across every client, routine production work stops landing on senior reviewers. If your firm runs the month-end close for many clients, start by tracking how many reviewer hours go to work a rule could have handled. That number is the capacity you can recover before you hire anyone.
Puzzle was originally built for venture-backed startup founders who needed real-time financial visibility without a full-time controller. It has since expanded to serve two core audiences: startup founders who want accurate, automated books without deep accounting knowledge, and accounting firms and bookkeeping practices that manage dozens to hundreds of client entities. On the firm side, Puzzle is purpose-built for tech-forward firms that want to move beyond QuickBooks, with firm-level management tools, client portfolio views, and AI-native automation that legacy software doesn't support. If you're a founder who outgrew spreadsheets, or a firm that wants to scale capacity without scaling headcount, Puzzle is designed for you.
Modern firms like Decimal have moved beyond manual entry by pairing AI-native general ledger software with automated categorization rules and machine-learning models trained on their specific client portfolios. In practice: automated rule creation maps recurring transactions to the correct GL codes on first sight; shortcut keys and bulk-action interfaces cut categorization time for edge cases; and AI-drafted reconciliations flag discrepancies automatically, without waiting for an accountant to spot them. The result is that repetitive data-entry work that once consumed hours of a bookkeeper's week shrinks to minutes of exception review. Puzzle's architecture applies these rules across an entire firm's client portfolio simultaneously, so a rule built for one client can be standardized and deployed across hundreds.
For accounting firms onboarding multiple clients, the migration follows a validated sequence:
Timing depends on the source system: QuickBooks Online migrations take a couple of days through the direct integration, while QuickBooks Desktop and Xero migrations take about five to seven days. Migration is $149 per system moved.
Puzzle offers two distinct layers in its AI suite. Accounting AI chat is a conversational interface where you ask questions about your books in plain English (for example, "What drove the spike in software expenses last month?" or "Show me a P&L for Q3"), and the AI queries your live ledger data to answer in context. AI Chat is read-only, so it answers from your books without changing them; for tasks like splitting a transaction, AI Cowork prepares the work for your review. AI Close agents are automated workflow agents that run the agentic AI month-end close checklist: checking for uncategorized transactions, validating reconciliations, and surfacing accuracy issues without you having to log in and trigger them manually. Agents can be configured to run on a recurring schedule (nightly, weekly, or at month-end), so the close process moves forward in the background. Accounting AI chat is available on the Core plan and above. Check the pricing page for current plan details.
Cowork turns a one-time accounting request into a repeating job. You describe the task once in plain English, a saved prompt-based agent runs it on a schedule, and you review the result before anything is final. Cowork is included on all plans. Two examples show how it fits real workflows. For transaction categorization, Puzzle learns from transaction metadata, patterns across clients, and the decisions you finalize, then applies account-based rules to new transactions. About 90% of transactions are categorized automatically and about 85% are finalized, and you review and finalize them, which typically takes 15-20 minutes per week. Accuracy on a new account starts lower and climbs after the first month-end close. For revenue recognition, you set policies by product or subscription type, and Puzzle defers and recognizes revenue at the invoice line-item level, with bulk confirmation of up to 100 schedules per batch. Revenue recognition is Stripe-centric, so non-Stripe billing needs a policy attached to a manually created invoice or a manual journal entry. In every case, the agent drafts and a person approves.
Puzzle's multi-entity portal gives you one login for multiple companies at no additional charge. Each client keeps its own books and general ledger, so you can move between entities without mixing ledgers. You can also set up custom charts of accounts and automation rules once and reuse them across clients, which is the standardization Decimal relied on to scale. Puzzle does not offer native consolidations or eliminations across parent and subsidiary entities, so firms that need consolidated reporting use a third-party integration partner (Join) with separate per-entity accounts under one login. If you need to control exactly what each client can see, confirm the permission setup for your portfolio in a demo with the Puzzle team before you migrate clients.
The talent math has changed: over 300,000 U.S. accountants have left the profession in recent years (SmartVault, citing the Wall Street Journal), and the AICPA estimates that 75% of current CPAs will reach retirement eligibility within 15 years (Forbes Councils). Hiring your way to scale is no longer a reliable strategy. The firms growing capacity today are doing it through three levers: automation standardization (applying the same rules, COA templates, and reconciliation logic across every client so one bookkeeper's setup work multiplies across the portfolio); AI accuracy reviews (smart checks for common coding errors, duplicate vendors, and GL mapping changes that catch issues before they require manual review); and advisory-layer tools (variance reports and cash-vs-accrual dual views that let accountants serve more clients in advisory capacity, not on preparation tasks). Decimal's partnership with Puzzle was built around exactly this model: reducing per-client manual work while simultaneously improving the quality of output delivered to each client.
Not directly in Chat. Puzzle's AI Chat is read-only: it answers questions from your actual books, such as a P&L for any date range, without changing anything. For tasks that change the books, like splitting a transaction across two or more GL accounts or reclassifying a purchase as a fixed asset, you describe the task to AI Cowork, an agent prepares the work, and you review and approve it. Nothing hits the ledger without your approval, so you keep full control.
Automated categorization uses a combination of rules-based logic and machine-learning models to assign each transaction to the correct GL account without human input. When a transaction arrives from a bank feed or integrated payment processor (Stripe, Ramp, BILL, Gusto), the system checks it against a hierarchy of signals: exact-match vendor rules created by the accountant; pattern-matching rules based on merchant name, amount range, or memo field; and a trained model that predicts the most likely category based on how similar transactions were coded historically. Transactions that match with high confidence are categorized and moved forward automatically; low-confidence transactions are surfaced for human review. Over time, the model learns from every accountant correction, so categorization accuracy improves continuously. In Puzzle's implementation, rules are portable across a firm's entire client portfolio, so a rule built for one client can be applied as a template for new clients in the same industry, dramatically cutting onboarding time.
Puzzle supports multi-entity structures, which is one reason accounting software for bookkeeping firms managing diverse client portfolios choose it over single-entity tools. Each entity maintains its own books and GL, with firm-level views that let you see across the portfolio without mixing the underlying ledgers. For firms managing funds, holding companies, or clients with multiple operating entities, Puzzle produces entity-level reports for each set of books. Puzzle does not offer native consolidations or eliminations across parent and subsidiary entities, so consolidated reporting runs through a third-party integration partner (Join) with separate per-entity accounts under one login. Entity-level precision plus portfolio-level visibility is central to how Decimal scaled their firm capacity on Puzzle without adding headcount.
Digits bets that AI should act autonomously: the idea being that accounting can largely run itself with minimal human oversight. That model trades control for convenience: by the time you see the output, the categorization decisions are already made. Puzzle inverts that model: the AI does the work, but you approve before anything touches the books. That distinction matters most for accounting firms whose reputation depends on accuracy. On accounting depth, Puzzle maintains dual cash and accrual books natively (not as a toggle or a workaround), which is a structural advantage for clients that need both views simultaneously (common in VC-backed companies and growth-stage businesses). Kick is purpose-built for sole proprietors and small lifestyle businesses; it doesn't offer the firm-management layer, multi-entity support, or API depth that practices serving startup portfolios need. The right tool depends on your complexity: Kick for very simple single-owner businesses, Digits if full automation is the priority and you're comfortable with reduced control, Puzzle if you need AI-native automation with human-in-the-loop accuracy and the infrastructure to scale a firm.
Puzzle's headless API (read and write) is currently waitlist-only and not broadly available, and it is not a white-label or fully embeddable product in the traditional OEM sense. If you're building a SaaS product and want to surface your users' accounting data (transaction history, P&L, balance sheet, burn rate), the API is designed for programmatic access to ledger data, but you would need to join the waitlist first. What Puzzle does not currently offer is a fully rebranded, white-labeled UI that can be embedded inside another product under a different name. For firms looking to build on top of Puzzle's infrastructure, the recommended path is a direct partner conversation to understand API scope and integration depth for your specific use case.
Accounting firm partners receive a dedicated onboarding track that's distinct from the self-serve path for individual business customers. Firms go through a structured implementation process: COA standardization, bulk client migration support, integration setup (Keeper, Ramp, Plaid, Gusto, BILL), and training sessions for the firm's accounting and bookkeeping team. Ongoing support for firm partners is a top priority: firms managing dozens to hundreds of client entities need faster resolution times and more proactive communication than a single-entity founder account. Direct business customers have access to in-app support and documentation, but the white-glove migration and dedicated success track is a firm-specific offering. If you're considering Puzzle as a firm, expect the onboarding experience to feel more like an implementation partnership than a software signup.
AI-native accounting software is built with AI at the core of the ledger. QuickBooks adds AI features to an architecture designed decades before AI existed. The difference shows up in three places. When the work happens: in an AI-native ledger, categorization, matching and accuracy checks run as transactions arrive, so the books are already drafted when you sit down to review. With bolted-on features, you typically trigger AI suggestions inside a workflow that was built for manual entry. How much gets finished: Puzzle categorizes about 90% of transactions automatically and finalizes about 85%, with a human reviewing before anything is final. Who stays in control: Puzzle's AI does the work, and the accountant approves before anything touches the books. For a firm, that means rules, charts of accounts and accuracy checks get standardized once and applied across the whole client portfolio, which is the model Decimal adopted. If you are comparing tools, test one question: does the software finish the work during the month, or does it only suggest edits at month-end?
Three main challenges stand out in the accounting industry today.
First, the talent shortage in accounting.
Second, QuickBooks is eating more of the accountant's pie: The industry has spent decades building processes around QuickBooks, a tool that has now become a double-edged sword with dominant market share. For firms weighing AI accounting alternatives to QuickBooks, the gap is measurable: Puzzle categorizes about 90% of transactions automatically and finalizes about 85%, with a human reviewing before anything is final. By failing to adopt new accounting software early, the industry has allowed competition to stagnate.
Third, the fear of AI: AI paints a picture of a world where a single player with strong market share, committing billions of dollars to a "done-for-you" solution, could heavily influence the future of accounting. QuickBooks and OpenAI, together, present a vision of automation that leaves little room for the traditional accountant.This paints a grim picture for accountants. But what if it didn't have to be this way? What if accountants could shape the future instead of letting the future shape them?
At Puzzle, we believe accountants can take back control by adopting technology that works with them, not against them. To do this, software must uphold three core commitments:
Software and services are hard to combine well. They are distinct business models with different margins, workflows, and goals. Puzzle's approach is to run both on one ledger, with accountants reviewing and finalizing what the AI drafts.
As the Bench and Scalefactor collapses covered above show, building accounting software and running a services business under one roof is a hard combination. When the two collaborate as separate entities, they unlock extraordinary potential.
Decimal's four focus areas (less manual work, less client back and forth, a better digital experience, and a path from accounting to advisory) became the yardstick for every phase Puzzle built.
Want to learn more? Go deeper with our case study with Decimal
