# Calibrating Disclosure: When Clients Want To Know, And When They Don't

**By Dan Cumberland** · Published August 8, 2026 · Categories: AI Strategy

![Featured image for Calibrating Disclosure: When Clients Want To Know, And When They Don't](https://cms.dancumberlandlabs.com/uploads/featured_image_93749b9306.webp)

Illustration: Dan Cumberland Labs with Gemini.

> Four in ten professional services firms are getting conflicting instructions from clients— told both to use AI and to avoid it on the same project[^1].  That's...

Four in ten professional services firms are getting conflicting instructions from clients— told both to use AI and to avoid it on the same project[1](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-1)\.  That's not a client transparency problem\.  That's a communication vacuum, and the firm gets to fill it\.

The question isn't whether to disclose AI use\.  It's what your client actually needs to know\.  Most content on this topic frames it as a compliance question— did you follow the rules?  This article frames it differently: as a calibration question\.  Different levels of AI use require different responses, and treating them all the same means overcomplying on paperwork and undercomplying on actual conversation\.

> "The question isn't whether to disclose AI use\.  It's what your client actually needs to know\."

Here's the tool that answers it: a four\-tier framework that maps AI use level to disclosure requirement, with practical language for your engagement letters and client conversations\.  Three\-quarters of both corporate clients and firm respondents agree that firms— not clients— should lead the AI discussion[1](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-1)\.  So let's lead it\.

## The Gap Between What Clients Expect and What Firms Do in AI Disclosure

85% of law firm clients expect firms to disclose when AI is used on their matters[2](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-2)\.  Only 18% of law firms always do[3](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-3)\.  That 67\-point gap is where client relationships break down— not in the work itself, but in what wasn't said about it\.

These figures come primarily from law firm client surveys\.  But the directional picture holds across professional services— and in the tax sector, 77% of clients expect firms to use AI, yet 59% don't know if their firm actually does[4](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-4)\.

That's not a technology problem\.  It's a communications failure\.

> "This is not a preference\.  It is a threshold expectation, and firms that treat it as optional are accumulating reputational liability\."[2](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-2)

The reputational risk accumulates before anything goes wrong\.  In our experience, clients who discover AI use after the fact feel managed rather than served\.  And within two to three years, Integris projects AI governance documentation will move from competitive advantage to baseline procurement requirement[2](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-2)\.  Firms building that infrastructure now pay a low cost\.  Firms building it reactively pay relationship capital\.

Before deciding how to communicate, it helps to understand what clients are actually asking underneath the question\.

## What Clients Are Really Asking

Most clients asking "did you use AI on this?" are really asking two questions: "Is my confidential information safe?" and "If this goes wrong, are you still accountable for it?"  Answer those two questions directly, and you've answered the disclosure question\.

On data: 82% of consumers in Relyance AI's 2025 survey see AI data loss\-of\-control as a serious personal threat[5](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-5)\.  That figure comes from a vendor\-sponsored study, so treat it directionally— but the pattern is consistent with other research\.  Clients don't know which AI systems keep their data, train on it, or share it\.  They're not asking about your tools\.  They're asking whether their information is safe\.

On accountability: ABA Formal Opinion 512 \(2024\) settled this for lawyers— existing ethics rules apply fully to AI use in legal practice[6](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-6)\.  Duties of competence, confidentiality, communication, and reasonable fees all carry forward to AI\-assisted work\.  For non\-lawyers, the framing still holds: if AI touched the work, the professional is still the author of record\.  The client needs to hear that, explicitly\.

No matter the question, people are the answer\.  The client isn't asking about your tools\.  They're asking if a human they can hold accountable is still in charge\.

KPMG's 2025 global study of 48,000 people found that 83% would trust AI more with assurances including responsible governance and oversight standards[7](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-7)\.  The trust path isn't hiding the AI use\.  It's making the human accountability visible\.

Once you know what clients actually need— data protection and human accountability, not tool catalogs— you can calibrate your response to match the level of AI use, not the level of anxiety\.

## The AI Disclosure Calibration Framework

Not all AI use requires the same disclosure\.  A grammar checker and a strategy report generated by Claude are different things, and treating them identically means both overcomplying on paperwork and undercomplying on actual conversation\.  Here's how to calibrate\.

**AI Disclosure Calibration Framework— Four Tiers** A four-tier table mapping AI use level to disclosure requirement.  Tier 1 background tools require no disclosure.  Tier 2 internal drafts with human review need an engagement letter notice.  Tier 3 AI-generated client deliverables need per-deliverable disclosure.  Tier 4, where confidential client data enters AI systems, requires active consent before engagement.

**Tier 1— Background Tools:** Grammar checkers, spell\-check, research aggregators, formatting tools\.  No disclosure needed\.  AI functions the way spellcheck does— it's a production tool, not the author\.

**Tier 2— Internal Drafts with Human Review:** AI generates an initial draft, outline, or synthesis, and the professional is the primary author through editing and approval\.  Engagement letter notice is sufficient\.  AICPA guidance recommends language such as: "AI tools may be used in the provision of professional services; all work is reviewed and approved by \[professional title\] prior to delivery\."[8](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-8)  This isn't just a legal hedge— it's the accountability statement clients actually need\.

**Tier 3— AI\-Generated Client\-Facing Content:** When AI created or substantially shaped the final deliverable— the strategy memo, the analysis report, the recommendation— per\-deliverable disclosure in the document or cover note is appropriate\.  The line FourScore Business Law draws is useful here[9](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-9): when AI communicates directly with clients or functions as more than an assistant, disclosure is generally required\.

**Tier 4— AI Processed Client Data:** When confidential client information was input into an AI system— the details of a deal, personal financial data, project specifications— active consent before engagement is the standard\.  This is where data privacy and confidentiality duties are directly triggered\.  Consent can't be tucked into boilerplate\.  AICPA guidance is explicit that "generic, boilerplate provisions are insufficient"— the language must address quality control, data handling, and the specific nature of the AI use[10](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-10)\.  If you can't answer the data handling question clearly— which system, which retention policy— that's the gap to close before adding the clause\.

The quick gut\-check across all four tiers: if your client discovered you used AI for this specific task, would they feel cheated or impressed?  If "cheated," disclose— or do the work yourself[11](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-11)\.

## The AEC\-Specific Question No One Has Answered

Neither ACEC nor AIA has issued prescriptive guidance on how to adjust fees when AI reduces actual hours spent[12](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-12)\.  AEC firms are building AI disclosure policies— and billing policies— from scratch\.  That's both a problem and an opportunity\.

The billing paradox is real\.  If AI helped your team complete a scope in 80 hours instead of 120, what do you do with those 40 hours?  For time\-and\-materials work, the right answer is improved margins, not billing time you didn't spend— but no architecture or engineering association has codified this[12](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-12)\.  Firms that charge for hours AI eliminated are storing up a problem\.

What AEC firms can do now:

- **Add engagement letter language** at Tier 2 minimum\.  Borrow the AICPA model sentence and adapt it for AEC: "AI tools may be used in the provision of architectural and engineering services; all deliverables are reviewed and approved by \[name/role\] prior to delivery\."
- **Establish an internal AI policy before clients ask\.**  Guidance published for A&E firms puts it plainly: create the policy proactively so firms control the narrative before clients discover AI use\.[12](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-12)  Once you're explaining, you're on defense\.
- **Designate someone to own it\.**  AICPA guidance recommends maintaining an approved tool list and designating an AI lead or committee[13](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-13)\.  A governance owner— even a part\-time one— signals to clients that this isn't ad hoc\.
- **Address the billing question explicitly** in your internal policy\.  If you don't decide how to handle AI\-reduced hours before they occur, you'll improvise under pressure\.

Government and municipal clients may carry stricter procurement transparency requirements\.  Verify with your contract attorney before the engagement starts, not after\.

For AEC firms building their first [AI governance framework](/blog/ai-governance-strategy), the lack of ACEC/AIA guidance isn't a reason to wait\.  It's a reason to set the standard before clients set it for you\.

## Putting It Into Practice: AI Disclosure for Professional Services

The most practical AI disclosure move any professional services firm can make right now is a single sentence in your engagement letter\.  From there, it's about leading the conversation proactively rather than reacting when a client asks\.

**The engagement letter baseline\.** AICPA's model language[8](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-8) is a starting point, not a complete solution\.  Generic boilerplate is insufficient[10](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-10)\.  A complete Tier 2 notice addresses three things: that AI tools are used, that all work is reviewed by a named role before delivery, and how client data is handled\.  If you can't answer the third in your current setup, that's the policy gap to close first\.

**When a client asks directly\.** Don't answer the literal question\.  Answer the underlying one\.  Something like: "Yes, we use AI tools as part of our workflow\.  Here's how we protect your confidential information, and here's who reviews all work before it leaves our office\."  What a [decision framework for founders](/blog/ai-decision-framework-founders) often surfaces is that clients don't want a tool catalog— they want the assurance that a human is responsible\.

**When a client says "don't use AI\."** Scope it clearly in writing— and define what counts\.  Does "AI" include grammar assistance?  Research aggregators?  Spell\-check?  Without definition, there's future ambiguity baked into the agreement\.  The definition conversation itself is valuable\.  It surfaces what the client is actually worried about, which is almost always data security or AI making consequential decisions without human review\.

**For existing clients with no prior AI disclosure:** a brief note at the start of the next project phase— not an apology, just a policy statement— is enough to get ahead of it\.

**Proactive beats reactive\.** The 40% of firms receiving conflicting instructions are almost certainly in reactive mode— responding to whatever question a client asked most recently\.  Firms that introduce their AI policy at kickoff own the framing\.  That opening exists\.  Use it\.

Understanding [what a fractional AI officer does](/blog/what-is-a-fractional-ai-officer) can help firms assign governance ownership before it becomes an afterthought\.

## FAQ

A few questions that come up repeatedly— answered directly\.

### Do professional services firms have to disclose AI use to clients?

No federal law requires it, but ethical duties of competence, confidentiality, and communication apply across professions[10](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-10)\.  Disclosure is mandatory when: clients ask directly, AI affects fee calculations, client data is input into AI systems, or court and state rules require it[6](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-6)[14](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-14)\.  Best practice is proactive engagement letter notice regardless of jurisdiction\.

### What do AEC firms need to know about AI disclosure?

Neither ACEC nor AIA has issued specific guidance on AI disclosure or billing adjustments when AI reduces hours spent[12](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-12)\.  AEC firms should establish an internal policy before clients ask, add AI disclosure language to engagement letters, and treat AI efficiency gains as improved margins rather than billable hours not worked\.

### What should an engagement letter say about AI use?

AICPA guidance recommends language such as: "AI tools may be used in the provision of professional services; all work is reviewed and approved by \[professional title\] prior to delivery\."[8](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-8)  Generic boilerplate is insufficient— the language must address data handling and human oversight specifically[10](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-10)\.  Consult your professional association guidance or legal counsel for jurisdiction\-specific requirements\.

### What if a client tells me not to use AI?

Scope the restriction clearly in the engagement letter— and define what counts as "AI\."  Without definition, there's future ambiguity about whether grammar tools, spell\-check, or research aggregators fall within the prohibition[1](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-1)\.  The definition conversation itself clarifies expectations and surfaces the client's actual concern\.

### How does disclosing AI use affect client trust?

Short\-term, proactive disclosure may prompt questions\.  Long\-term, the research is consistent: MIT Sloan research found 84% of expert panelists support mandatory AI disclosures[15](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-15)\.  KPMG's 2025 global study of 48,000 people found 83% would trust AI more with governance assurances[7](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-7)\.  The path to trust runs through disclosure, not around it\.

## Conclusion

AI governance is a competitive advantage right now\.  Within two to three years, Integris projects it will be a baseline procurement requirement— the thing clients check before they hire you[2](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#ref-2)\.  The firms that establish a disclosure policy today own that story\.  The ones waiting will be explaining the gap\.

The disclosure question is really the trust question\.  What clients actually need— data security and human accountability— is what the [AI Disclosure Calibration Framework](/blog/blog-calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t#the-ai-disclosure-calibration-framework) gives you a structure to communicate\.  Answer it before they ask it\.

If you're working through what your firm's AI policy should say, or how to lead that conversation with clients, [our AI strategy practice](/services/ai-strategy) helps professional services firms do exactly that\.  The cost of starting is a sentence in your engagement letter\.  The cost of waiting is a client relationship\.

## References

1. Thomson Reuters Institute, "2026 AI in Professional Services Report" \(2026\)— [https://www\.thomsonreuters\.com/en\-us/posts/technology/ai\-in\-professional\-services\-report\-2026/](https://www.thomsonreuters.com/en-us/posts/technology/ai-in-professional-services-report-2026/)
2. EfficientlyConnected / Integris, "Law Firm AI Governance: Closing the Client Trust Gap in 2026" \(2026\)— [https://www\.efficientlyconnected\.com/law\-firm\-ai\-governance\-client\-trust\-2026/](https://www.efficientlyconnected.com/law-firm-ai-governance-client-trust-2026/)
3. Clio, "AI Disclosure for Lawyers: When to Disclose, What to Say, and How to Stay Compliant" \(2025 Legal Trends Report\)— [https://www\.clio\.com/blog/ai\-disclosure\-lawyers/](https://www.clio.com/blog/ai-disclosure-lawyers/)
4. Thomson Reuters Institute, "2025 Generative AI in Professional Services Report" \(2025\)— [https://www\.thomsonreuters\.com/en/press\-releases/2025/april/from\-incubation\-to\-integration\-generative\-ai\-adoption\-nearly\-doubles\-as\-professional\-services\-reach\-crossroads](https://www.thomsonreuters.com/en/press-releases/2025/april/from-incubation-to-integration-generative-ai-adoption-nearly-doubles-as-professional-services-reach-crossroads)
5. Relyance AI, "Customer AI Trust Survey 2025" \(2025\)— [https://www\.relyance\.ai/consumer\-ai\-trust\-survey\-2025](https://www.relyance.ai/consumer-ai-trust-survey-2025)
6. The AI Career Lab, "Disclosing AI Use to Clients: What Professional Ethics Codes Actually Require in 2026" \(2026\)— [https://theaicareerlab\.com/blog/disclosing\-ai\-use\-to\-clients\-professional\-ethics\-2026](https://theaicareerlab.com/blog/disclosing-ai-use-to-clients-professional-ethics-2026)
7. KPMG, "Trust, Attitudes and Use of Artificial Intelligence: A Global Study 2025" \(2025\)— [https://kpmg\.com/au/en/insights/artificial\-intelligence\-ai/trust\-in\-ai\-global\-insights\-2025\.html](https://kpmg.com/au/en/insights/artificial-intelligence-ai/trust-in-ai-global-insights-2025.html)
8. Journal of Accountancy / AICPA, "Drafting an AI policy that actually works" \(July 2026\)— [https://www\.journalofaccountancy\.com/issues/2026/jul/drafting\-an\-ai\-policy\-that\-actually\-works/](https://www.journalofaccountancy.com/issues/2026/jul/drafting-an-ai-policy-that-actually-works/)
9. FourScore Business Law, "AI Disclosure in Business: When, Why, and How to Inform Clients About Your Use of AI" \(2025\)— [https://www\.fourscorelaw\.com/resources/ai\-disclosure\-in\-businessnbspwhen\-why\-and\-how\-to\-inform\-clients\-about\-your\-use\-of\-ai](https://www.fourscorelaw.com/resources/ai-disclosure-in-businessnbspwhen-why-and-how-to-inform-clients-about-your-use-of-ai)
10. Journal of Accountancy / AICPA, "Should I disclose my use of gen AI to clients?" \(April 2025\)— [https://www\.journalofaccountancy\.com/issues/2025/apr/should\-i\-disclose\-my\-use\-of\-gen\-ai\-to\-clients/](https://www.journalofaccountancy.com/issues/2025/apr/should-i-disclose-my-use-of-gen-ai-to-clients/)
11. Bradford Tobin, "Using AI in Client Work" \(2025\)— [https://bradfordtobin\.substack\.com/p/using\-ai\-in\-client\-work](https://bradfordtobin.substack.com/p/using-ai-in-client-work)
12. Monograph, "AI for Professional Services Firms: What Architects and Engineers Should Know" \(2025\)— [https://monograph\.com/blog/ai\-for\-ae\-firms\-architects\-engineers](https://monograph.com/blog/ai-for-ae-firms-architects-engineers)
13. Journal of Accountancy / AICPA, "Drafting an AI policy that actually works" \(July 2026\)— [https://www\.journalofaccountancy\.com/issues/2026/jul/drafting\-an\-ai\-policy\-that\-actually\-works/](https://www.journalofaccountancy.com/issues/2026/jul/drafting-an-ai-policy-that-actually-works/)
14. PAXTON AI, "2025 State Bar Guidance on Legal AI: Policies, Ethics, and Best Practices for Law Firms" \(2025\)— [https://www\.paxton\.ai/post/2025\-state\-bar\-guidance\-on\-legal\-ai](https://www.paxton.ai/post/2025-state-bar-guidance-on-legal-ai)
15. MIT Sloan Management Review, "Artificial Intelligence Disclosures Are Key to Customer Trust" \(2024\)— [https://sloanreview\.mit\.edu/article/artificial\-intelligence\-disclosures\-are\-key\-to\-customer\-trust/](https://sloanreview.mit.edu/article/artificial-intelligence-disclosures-are-key-to-customer-trust/)


---

**How we made this article:** We use AI in our research and writing so our small team can share more of what we learn. We verify the sources and take responsibility for every article we publish. [Read how we use AI.](https://dancumberlandlabs.com/how-we-use-ai/)

---

## About the author

**Dan Cumberland** — Founder, Dan Cumberland Labs

Dan Cumberland helps engineering and construction firms see where they stand with AI and decide what to build first. He created Pacemark, the AI maturity model behind that work, from research on more than 300 companies.

- Take the assessment: https://pacemark.ai/signal/assessment/?track=aec&utm_source=dcl-site&utm_medium=link&utm_campaign=pacemark-assessment
- Book a call: https://book.dancumberland.com/ai-strategy

---

Source: https://dancumberlandlabs.com/blog/calibrating-disclosure-when-clients-want-to-know-and-when-they-don-t/