We Tested QuickBooks AI for 90 Days

Every accounting software vendor is racing to badge their product as “AI-powered”. But QuickBooks has gone further than most.

Since mid-2025, Intuit has been rolling out AI-powered bookkeeping experiences across QuickBooks Online. These features go beyond chatbots and suggestions. They assist with transaction categorisation, reconciliation, anomaly detection, invoice drafting, and automated client follow-ups.

The pitch is familiar and compelling:

• Save hours each week
• Reduce manual bookkeeping
• Focus on higher-value advisory work

But once you move past the marketing, UK accountants are asking very different questions in practice forums, Reddit threads, and private peer groups:

• ⚠Does this actually reduce workload, or just shift it?
• ⚠How much checking is really required?
• ⚠What happens to client data when AI is involved?
• ⚠And who’s liable when it gets things wrong?

This isn’t another think-piece about the future of AI in accounting. It’s a practical, grounded assessment of what QuickBooks AI does today, where it helps, where it creates risk, and what UK practices should be doing right now to stay compliant and protected.

Question 1: Does QuickBooks AI Actually Save Time – or Create More Work?

What Intuit claims

Intuit positions its AI features as a way to save “hours per week” by automating categorisation, reconciliation, and routine workflows – while keeping accountants “in control”.

What we’ve observed in practice

The reality is mixed. Very mixed.

In limited trials and practitioner feedback, QuickBooks AI performs best when transactions are highly repetitive and predictable. It struggles when judgement, context, or tax nuance is required.

The honest verdict

If most of your client base consists of sole traders and micro-businesses with clean, repetitive data, AI assistance can save time.

If you deal with complex SMEs, property portfolios, or group structures, you should expect to spend a meaningful amount of time reviewing and correcting AI outputs.

AI doesn’t remove work. It moves it from data entry to quality control. Whether that’s a net win depends on how disciplined your review process is.

Question 2: Is Client Data Safe – and What Does the OpenAI Partnership Actually Mean?

This is where clarity starts to fade.

In November 2025, Intuit announced a multi-year partnership with OpenAI, reportedly valued at over $100 million, to support AI capabilities across QuickBooks, TurboTax, and Credit Karma.

The question UK accountants should be asking isn’t “Is this clever?”
It’s this:

When QuickBooks AI processes my client’s financial data, how – and where – is that data handled?

What Intuit says publicly

Intuit states that its AI features are built with a commitment to privacy, security, and responsible AI governance, and that customers remain in control.

What isn’t always made explicit

From a UK GDPR perspective, practices need clarity on:

• ⚠Whether any AI processing involves international data transfers
• ⚠Which sub-processors are involved in AI features
• ⚠What contractual safeguards apply to AI-related processing
• ⚠How long AI-processed data is retained and for what purposes

These aren’t academic concerns. Under UK GDPR, you remain the data controller, even when using cloud software with embedded AI.

Our recommendation

Until vendors provide clearer, plain-English documentation on AI data flows, UK practices should:

1. Conduct a DPIA covering AI features in bookkeeping software
2. Review Intuit’s Data Processing Agreement and sub-processor disclosures
3. Clearly disclose AI use in engagement letters and privacy notices
4. Be cautious about enabling AI features for higher-risk clients without documented oversight

If you need help mapping AI data flows or completing a DPIA, this is exactly what we support firms with at PPCS through our ISO 27001 readiness services.

Question 3: How Do You Know When AI Gets It Wrong?

This is the most underestimated risk.

Traditional accounting software makes mistakes when humans make mistakes.
AI can make mistakes autonomously – and unless you’re actively looking, you may not spot them until much later.

The audit trail question

QuickBooks Online does include an audit log. But AI-assisted workflows increase the volume and speed of changes, which makes disciplined review essential.

Ask yourself:

• ⚠Can you easily identify which entries were AI-assisted?
• ⚠Do you have a defined review cadence for those entries?
• ⚠Can you evidence that professional judgement was applied if queried?
• ⚠Would your PI insurer be satisfied with your oversight process?

Question 4: Should You Trust AI – or Stick with Humans?

Here’s the uncomfortable truth:

AI is faster. Humans are more accurate.

AI is excellent at pattern recognition and volume processing. It’s poor at context, nuance, and professional scepticism.

The best model is hybrid: AI as an assistant, humans as the final authority. But only if you’re honest about the time needed for proper review.

Question 5: Can You Turn QuickBooks AI Off?

This is where many firms feel caught out.

There is no single “turn off all AI” switch in QuickBooks Online. Some AI-related features can be adjusted or disabled, depending on region and rollout phase, but others are increasingly embedded in workflows.

You may be able to:

• Disable conversational AI assistants
• Control automated customer emails
• Adjust some automation rules

You generally cannot:

• ⚠Fully disable AI-assisted categorisation suggestions
• ⚠Switch off background anomaly detection entirely
• ⚠Remove AI from reconciliation workflows altogether

The direction of travel is clear: QuickBooks is becoming AI-first. Practices need to decide whether that aligns with their operating model and risk appetite.

The Security and Liability Risks Vendors Don’t Emphasise

1. Accuracy and “hallucination” risk

AI systems can generate outputs that look plausible but are wrong. In finance, that can mean misapplied tax codes, incorrect assumptions, or misleading explanations.

Mitigation: Never rely on AI-generated explanations without verification. For more on this, see ICAEW’s guidance on managing AI risks.

2. Third-party AI risk

Using AI features means relying not just on Intuit, but on its AI supply chain. Outages, breaches, or policy changes upstream can affect you downstream.

Mitigation: Ensure your contracts clearly cover sub-processors, breach notification, and data deletion rights.

3. Professional liability

If AI contributes to an error, responsibility doesn’t disappear.

The question isn’t “Did the AI do it?”
It’s “Can you demonstrate reasonable oversight?”

Speak to your PI insurer now, not after a claim. Learn more about protecting your practice in our article on common cybersecurity mistakes accounting firms make.

The ISO 42001 Opportunity: Turning AI Risk into Advantage

While many firms are using AI informally, very few are governing it properly.

ISO/IEC 42001 is the new international standard for AI Management Systems. It provides a structured way to show that AI is used responsibly, transparently, and under human control.

It covers:

• ✓AI inventories and use-case mapping
• ✓Risk assessments for AI systems
• ✓Human oversight and escalation procedures
• ✓Data governance and privacy controls
• ✓Supplier and third-party AI risk management

Our Verdict

QuickBooks AI can be useful – in the right context, with the right controls.