Using AI in Client Work: What to Tell Clients and What to Keep Private

Every professional services firm is navigating this question right now: how much do you tell clients about your use of AI? The answers range from "full proactive disclosure" to "say nothing and they'll never know." Neither extreme is quite right — and the right answer depends on context in ways this post will help you work through.

The second half of the question — what to keep private, meaning what confidential client data you should never put into AI tools — has a clearer answer. Understanding both will help you build AI-augmented services that clients trust rather than services that create exposure.

The Disclosure Question: A Framework for Deciding

There is no universal disclosure requirement for AI use in professional services — yet. Norms are evolving, and sector-specific requirements are emerging, but most professionals currently operate in a space where they must exercise judgment.

Here are three questions that help clarify the right disclosure approach for a specific situation:

1. Is the AI use material to the client's decision to engage you?

If a client is paying a premium for what they believe is intensive human expertise, and you're delivering AI-generated outputs with minimal human input, they would likely consider that material information. The relevant test: "Would this affect their decision to engage me or pay this price?" If yes, disclose.

2. Does the AI output go directly to the client, or does it support your own work?

Using AI to research an industry before writing your analysis is different from sending a client an AI-generated report with minimal editing. The former is a workflow tool. The latter is a deliverable — and the client may have legitimate interest in knowing how it was produced.

3. Does your sector have emerging disclosure norms or requirements?

Legal, financial services, healthcare, and publishing are all developing more specific disclosure expectations. If you're in one of these sectors, proactively research what's expected and err toward disclosure where norms are developing — it's much better to over-disclose before requirements formalise than to be caught under-disclosing after they do.

What to Tell Clients: The Practical Approaches

Most clients who understand AI — which is an increasing proportion of decision-makers — don't object to AI use per se. They object to surprises, to undisclosed risks, and to poor quality. Proactive, confident disclosure typically lands better than you'd expect.

"We use AI tools as part of our research and drafting process. Every deliverable we produce is reviewed, edited, and ultimately signed off by our team — you're getting our expertise and judgment, supported by AI-powered efficiency."

That framing — AI as a capability enhancer, human expertise as the accountable layer — is both honest and reassuring. Most clients find it persuasive, not alarming, when stated with confidence.

For clients who specifically prohibit AI use in their supplier agreements, you need to be aware of those terms and comply. These clauses are becoming more common, and violating them is both a legal and reputational risk.

Protecting Client Data: The Non-Negotiable Rules

This section is simpler than the disclosure question — and more critical.

Never input client-identifying information into public AI tools. Names, company names, financial figures, strategic plans, personnel information — none of this belongs in ChatGPT's public interface, the free tier of Claude, or similar tools. The data handling terms of these tools typically allow them to use inputs for training, and their security posture is not appropriate for confidential client data.

The practical workaround: anonymise. Describe the scenario without identifying details. "A retail company with $50M revenue in Southeast Asia" is different from "Retailer X based in Singapore." You can often get the AI assistance you need without exposing client identity.

If you're doing substantial AI-assisted work with client materials, invest in tools with appropriate enterprise data handling: Microsoft Copilot (with Microsoft's enterprise data commitments), Claude for Enterprise, or your organisation's private AI deployment. These are designed for professional use with appropriate data governance.

Quality Control: The Standard You Can't Drop

Using AI in client work raises the quality control stakes, not lowers them. The fastest path to damaging client trust is sending AI-generated work with obvious errors — hallucinated facts, wrong client names (especially in templatised content), inappropriate tone, or generic content where specific insight was expected.

Non-negotiable quality controls for AI-assisted client deliverables:

• Every factual claim independently verified against primary sources
• Client-specific details checked and confirmed (no wrong company names, wrong figures)
• Tone reviewed by a human who knows the client relationship
• Overall coherence assessed — does this actually answer the client's question?
• Final review by the person who is professionally accountable for the work

AI draft quality varies significantly. First drafts often contain plausible-sounding errors that a subject matter expert would catch immediately. Don't let speed optimisation override quality assurance.

Pricing Considerations

If AI makes you faster, should clients pay less? This question makes many professionals uncomfortable, but it's worth thinking through rather than avoiding.

Value-based pricing is the clearest framework: if the client is paying for an outcome (a strategy, a contract, a financial plan), the time you took to produce it is secondary to whether it meets their needs. AI that lets you produce better work faster increases your value delivered — it doesn't decrease it.

If you bill by the hour, the question is more complex. Charging full hourly rates for work that took 20% of the time it previously would have done raises questions of honesty. The industry is moving toward value-based engagements partly for this reason. If you bill hourly, be honest with yourself about whether your rates reflect the actual time invested.

Building AI-Augmented Services Clients Trust

The firms winning with AI in client work aren't hiding it — they're building it into their value proposition. "We use AI-powered research tools to deliver insights faster, and our senior team's judgment to make them actually relevant to your business" is a stronger pitch than either "we do everything manually" or "AI does most of it."

Clients are looking for results, speed, and confidence. AI helps with the first two. Confidence comes from the quality of your work and the transparency of your process. Get all three right and AI becomes a competitive advantage rather than an ethical minefield.