Yes, AI can help you with your taxes – if you know how to use it

Key stat: 76% of accountants saw increased client use of public AI tools in 2025 and KPMG reports 82% of Canadian firms are using or piloting AI in finance. Nearly half of accountants spend up to 3 hours/month fixing AI-generated errors (27% spend 4–6 hours); common issues include incorrect business-expense interpretations (44%) and incorrect tax claims (43%), and over 40% expect increased misuse or fraud. A cited hallucination falsely claimed the 2026 capital gains limit rose from about $1.25M to $1.3M, highlighting operational and compliance risk; dedicated Canadian tax AI tools (e.g., TaxGPT, CloudTax) and human oversight can still extract efficiencies and missed credits but cannot replace expert judgment.

Analysis

Specialized, jurisdiction-aware AI will be the primary value capture point over the next 12–36 months. Firms that combine deterministic rule engines (the ‘if-then’ layer auditors trust) with fine-tuned Canadian LLMs and provable data provenance will win premium pricing and higher retention among mid-market accountants, accelerating consolidation of smaller advisory shops into platforms that own client tax workflows.

Regulatory and liability shocks are the key tail risks and likely catalysts. Expect professional bodies and revenue authorities to demand audit trails, model version stamping, and vendor certification within 6–18 months; that creates a barrier to entry for general-purpose LLM vendors but a moat for vendors who can operationalize compliance controls rapidly.

Cybersecurity, data privacy, and real-time update mechanisms are secondary but enforceable demand drivers — clients will pay for on-premise or segmented-cloud deployments that keep tax data local and models refreshed with legislative changes. This produces a multi-year uplift in recurring software and cloud spend from accounting firms and large SMBs, favoring cloud infra and security vendors able to offer compliance-first LLM stacks.

For platform/marketplace businesses reliant on gig/SMB supply (e.g., ride-hailing, delivery), improved bookkeeping and low-friction tax tools are a two-edged sword: they lower onboarding friction and increase net supply elasticity, but also raise tax-compliance visibility and potential employer/contractor classification risk over 2–5 years. Net effect is neutral-to-modestly-positive for volume, conditional on regulatory clarity.

Key Decisions for Investors

• Long Thomson Reuters (TRI) — 12–24 month horizon. Rationale: incumbent tax/legal content + ability to integrate deterministic rules with LLMs creates a durable enterprise upgrade cycle. Position sizing: 3–5% of tech allocation. Target return: 20–30% upside vs downside 12–18% if competition accelerates or pricing resets.
• Long Intuit (INTU) via 12-month call spread (buy 1 ATM call / sell 1 out-of-the-money call 20–25% above) — plays continued monetization of tax-prep + bookkeeping automation. Risk/reward: capped upside but low cash outlay, expect 15–25% realized return if adoption accelerates; downside limited to premium paid.
• Long cybersecurity/stack names (CRWD or PANW) — 6–18 month horizon. Rationale: surge in compliance-grade deployments and managed LLM controls will lift security spend at accounting firms and fintechs. Trade: 2–4% portfolio weight; target 15–25% upside with defensible downside due to secular security demand.
• Pair/trade idea: Long TRI (or INTU) / Short PAYX (Paychex) — 12–18 months. Rationale: software and content incumbents that integrate compliance-first AI will grow wallet share versus legacy payroll providers slower to embed LLM governance. Target: 10–20% net alpha if regulatory certification favors vertical software vendors.