Why Accounting Software Ignores the Hardest Problem

The assumption every tool makes

Open any accounting software demo and you'll notice the same thing: the documents are already organized. The bank statement is already split from the T4. The receipt is already matched to the transaction. The client folder is already complete.

Every tool in the accounting tech stack — Dext, Hubdoc, QuickBooks, DataSnipper, TaxDome — starts from this assumption. Documents arrive clean. Your job is to process them.

But anyone who has worked in a firm knows the truth: documents don't arrive clean. They arrive as chaos.

The structural lock-in

It's not that these companies don't see the problem. It's that their business models prevent them from solving it.

Receipt-first tools like Dext have trained their AI on 320 million receipts and invoices — standardized, predictable formats. But the upstream problem isn't receipts. It's 35-page merged PDFs mixing bank statements, T4s, property tax assessments, and medical receipts. Different document entirely. Their training data is the wrong training data.

Practice management platforms like TaxDome spread R&D across 10+ modules — CRM, billing, workflow, e-signatures, client portals. Document intake gets 5–10% of engineering resources. It's a feature checkbox, not a product.

Audit automation tools like DataSnipper assume documents are already sourced and organized. They optimize the review step, not the collection step.

Each tool has optimized for its best customers — organized firms with structured workflows. The firms with the worst document chaos are, by definition, the worst customers for these platforms. Small deals. High support costs. Low NPS.

This is the classic innovator's dilemma. The problem is too messy, too unstructured, too firm-specific for platforms that serve thousands of customers with a single product.

Why acquisitions make it worse

When a tool gets acquired, the problem deepens.

Hubdoc was originally built for Canadian accountants. Then Xero acquired it for $70M in 2018. By 2024, the product was fully absorbed into Xero's subscription. The independent trajectory died. Now it serves Xero's ecosystem, not Canadian firms.

SafeSend was acquired by Thomson Reuters for $600M in January 2025. TR wants an end-to-end US tax workflow — SurePrep to SafeSend to UltraTax. Zero incentive to invest in Canadian TaxCycle or Profile integration. SafeSend's per-return pricing model ($13–17/return) also breaks with year-round document intake — it would cannibalize their own revenue.

The pattern repeats: acquisition optimizes for the acquirer's ecosystem, not for the original customer's pain.

What the gap looks like

Here's what happens in the space between "client sends documents" and "accountant starts working":

• 35-page merged PDFs need boundary detection — where does one document end and another begin?
• Duplicate submissions across multiple emails need deduplication
• Missing paired documents need detection — a T4 arrived but the matching Relevé 1 didn't
• Firm-specific classification rules need to be learned — "invoice" means different things to different firms
• Completeness tracking across all clients needs automation — 99% of the time, something is missing

"99 % du temps, il manque quelque chose."

— Rea Daher, CPA

No existing tool handles this. Not because it's technically impossible, but because it's structurally incompatible with how the market is organized.

The infrastructure opportunity

The gap isn't a feature. It's a layer.

The firms that figure out how to structure this layer won't just move faster. They'll change the unit economics of accounting — making document volume a solved problem rather than a linear cost that grows with every new client.

---

Someone will own the transition from document chaos to document usability. The question is whether it will be a feature inside a platform that spreads its attention across 10 modules, or a product built to solve this one problem with full focus.