Something happened in a client meeting that I have kept thinking about.
We were sitting with the CFO of a 400-person professional-services firm to discuss its digital platform. Before we began, she showed us a financial scenario model she had created that morning with help from Claude. It had sensitivities and a structure she found useful. Then she asked:
"If I can do this, why is my auditor still sending me a PDF summary that takes three weeks?"
She was not claiming that an AI-generated model and an audited conclusion were equivalent. She was asking why one part of her working life felt immediate and iterative while another arrived later with little visibility of the process.
An answer about methodology and sign-off was true and incomplete. Assurance, professional judgement and accountability take time. The service provider still needs to explain where that time creates value, which parts of the process could be faster and why the final output deserves more trust than a twenty-minute draft.
That is the expectation shift worth examining.
Clients compare processes as well as outputs
AI tools give many clients direct experience of rapid drafting, summarisation, analysis and exploration. The quality varies, and some uses create serious confidentiality, accuracy or governance risks. The experience still changes the questions clients bring to advisers.
They may ask:
- Why does gathering and structuring information take this long?
- Which part requires expert judgement and which part is administration?
- Are we paying for the firm's learning curve or for a better outcome?
- How did you verify this advice or analysis?
- Are you using our information in an AI system, and under what controls?
- Can we work with the intermediate evidence instead of waiting for a final PDF?
The expectation is not simply "use AI". It is greater clarity about pace, process, evidence, data and the human contribution.
Expertise and relationships remain the foundation. The definition of credible expertise now includes the ability to discuss where AI changes the domain, where it fails and how the firm governs its own use.
Do not confuse generation speed with service speed
The easiest response is to promise faster everything. That would replace one misunderstanding with another.
An AI tool can produce a plausible summary quickly. A professional service may also require source validation, legal or regulatory analysis, challenge, consultation, independence, professional standards and accountable sign-off. The elapsed time includes more than typing.
Break the service into stages and explain them:
- obtaining and validating the inputs;
- structuring or analysing information;
- applying domain judgement and considering exceptions;
- reviewing evidence and challenging conclusions;
- agreeing action and communicating it;
- maintaining an appropriate record.
AI may accelerate parts of stages two and four. It may also introduce new review work, especially where outputs are difficult to reproduce or contain plausible errors. A faster first draft has no value if a senior professional spends longer correcting it.
Clients will accept necessary assurance more readily when they can see what it protects them from and receive useful progress in the meantime. Status, early questions and working hypotheses can improve the experience without lowering the quality threshold.
Six expectations that may be changing
1. Faster routine work
Clients may reasonably expect less delay in gathering, formatting, comparing and summarising material. Start with repeatable, lower-ambiguity tasks and measure the complete workflow.
Do not benchmark professional turnaround against an unverified client experiment. Establish a baseline, identify waiting and rework, then decide whether technology, process, resourcing or clearer scope is the right intervention.
Where the firm becomes faster, be prepared to discuss pricing and value. Clients may challenge a fee based on hours when visible production time falls. The answer should focus on responsibility, expertise, outcome and the operating investment required, without pretending efficiency is irrelevant.
2. Better access to evidence
A static summary may feel inadequate when clients can explore data and scenarios themselves. Give them useful access where it is safe and meaningful: source references, assumptions, scenario ranges, status and a route to ask questions.
Avoid dressing an interactive dashboard up as insight. Current data can still be incomplete; more dimensions can obscure the decision. The adviser adds value by framing what matters, testing explanations and identifying consequences.
3. Proactive relevance
Clients may expect the firm to notice a material change before the annual review. AI and analytics can help scan information or prioritise signals, but the design needs restraint.
An inferred risk is not a fact. A relevant alert for one client can be intrusive marketing for another. Define which events the firm is entitled and equipped to monitor, when a professional reviews the signal and how the client controls communication.
Proactivity means bringing a useful, considered observation at the right time. It does not mean increasing the volume of automated messages.
4. Transparency about AI use
Clients increasingly have reason to ask whether their work or information passes through AI tools. A credible firm can answer plainly:
- which approved tools and suppliers are involved;
- what information is permitted;
- whether data is stored, used for training or accessible for monitoring;
- where human review occurs;
- how errors and incidents are handled;
- whether the use affects the service, fee or client choice.
The answer depends on contract, professional duties, regulation and tool configuration. "Enterprise-grade" is not a data-flow explanation.
5. Evidence of quality
AI increases the supply of polished language. That can make provenance, source and accountable judgement more valuable.
Show how conclusions were reached and checked at a level suited to the task. Keep records of source material, model or tool configuration where relevant, versions, review and material changes. The evidence should allow the firm to investigate a failure and defend a professional decision.
6. Advice about the client's own use
Clients may want an intelligent conversation about AI in their sector, even when they did not instruct the firm for an AI project. They may be experimenting with sensitive documents, AI-generated analysis or automated decisions without understanding the risk.
The adviser does not need to pretend to be a machine-learning engineer. They should know how AI affects their domain, which obligations and common failure modes matter, and when to involve a specialist.
The credibility gap
The risk is rarely that a client immediately leaves because a partner cannot discuss the latest model. It is a gradual change in how the client judges the relationship.
"We are exploring AI" carries little weight if nobody can explain what the firm has learned, where it is using the technology or why it has declined a use. Empty transformation language can be less credible than a careful early-stage position.
One managing partner at a 180-person consultancy expected a major client to demand sophisticated AI deployment. The client actually wanted evidence that the firm was thinking seriously about the implications. The partner described two internal experiments and explained why client-facing outputs remained outside scope until quality controls improved. The client valued the judgement more than an inflated strategy claim.
Credibility comes from informed specificity:
- what the firm knows;
- what it is testing;
- what evidence it has;
- what it will not do;
- which uncertainty remains.
Build sector-specific literacy
Senior client-facing people do not need a stream of product announcements. They need a maintained view of how AI affects the problems they advise on.
A practical rhythm might include:
- a short monthly briefing on sector uses, failures and regulatory change;
- case discussion in practice meetings;
- hands-on use of approved tools with non-client material;
- a clear route to security, legal, data and technical specialists;
- questions relationship owners can use to understand client experimentation;
- an updated firm position and client FAQ.
For legal services, the SRA's current warning on AI misuse is a more useful starting point than generic productivity claims. It addresses competent and ethical use, including output checking and client information. Other professions and jurisdictions need their own authoritative sources.
Training should cover limitations and refusal as well as capability. A partner who can say why a proposed use is unsafe, then offer a bounded alternative, is demonstrating value.
Ask clients what is actually changing
Do not infer an expectation shift from headlines. Research your clients.
Relationship owners can ask:
- Where are you already using AI in this part of the business?
- Which tasks feel materially faster or different?
- Where has the output created extra checking or uncertainty?
- What would you want us to explain about our own use?
- Which parts of our service now feel unnecessarily slow or opaque?
- Where is human judgement most important to you?
Listen for differences among roles and client types. A CFO at a PE-backed firm may have different expectations from an owner-manager, general counsel or regulated operations lead. Some clients will prioritise innovation; others will prioritise assurance and data control.
Record themes without turning experimental use into a sales trigger. Client curiosity is an opportunity to understand the relationship, rather than permission to push an AI product.
Audit one service journey
Choose a service where clients experience delay, repeated information or limited visibility.
Map the stages, elapsed time, professional time, waiting, data and decisions. Mark where AI might assist and where accountable human judgement must remain. Include the client's actions and uncertainties.
For each proposed change, define:
- the client and service outcome;
- the baseline;
- the tool and data flow;
- quality and risk controls;
- human responsibility;
- evidence and review point;
- what the client should be told.
Pilot a bounded workflow. Measure total turnaround, review effort, error, exceptions and client experience. Do not announce an efficiency percentage from a handful of unusually simple cases.
The AI readiness checklist is a useful starting point for comparing the firm's capability with the expectations emerging in its market. It should prompt evidence and decisions, rather than produce a decorative maturity score.
The opportunity beyond the credibility floor
Firms that move from literacy to disciplined AI-assisted delivery may improve responsiveness, analysis and access to knowledge. The value should appear in a client outcome, not only an internal demonstration.
The companion article on the AI adoption curve explores different organisational positions. The article on what using AI in a business means covers the foundations before client-facing applications.
The shift is a relationship story as much as a technology story. Clients' tools change how they think about pace, evidence and usefulness. Service providers need to understand that experience while explaining the extra value of assurance, judgement and accountability.
You do not need to become an AI company. You do need to be credible: able to discuss the change, show how your service is improving, explain how client information is protected and distinguish a fast output from a sound professional outcome.
That is how expertise expands without becoming hype.
