The AI proficiency mirage on Wall Street’s margins

In boardrooms from Manhattan to Minneapolis, senior leaders at registered investment adviser (RIA) firms are telling themselves a confident story about artificial intelligence.

They have rolled out AI tools. They have drafted policies. Staff have completed “Prompting 101” modules. Adoption appears strong; vendors promise efficiency gains measured in hours, even days.

A new AI proficiency report, based on a survey of 5,000 knowledge workers at companies with more than 1,000 employees in the United States, United Kingdom and Canada, tells a more sobering story. Employees are using AI, but not in ways that reliably change how work gets done.

For RIA employers, the problem is not a lack of technology. It is a shortage of meaningful, repeatable use cases; uneven access for frontline staff; and a leadership class that appears convinced deployments are succeeding while the rest of the workforce remains unconvinced and underpowered.

From “Can You Prompt?” To “Can You Produce Value?”

In 2025, AI proficiency meant something modest: Did your people know how to use AI safely, avoid obvious data risks and write a decent prompt?

Corporations, including financial firms, spent the last year investing in that basic fluency. Employees learned what AI is, how to use it “responsibly,” and how to ask a model to summarize emails or rewrite client communications.

In 2026, the bar has shifted.

Now, AI proficiency means incorporating AI into meaningful, value‑adding work tasks every week. Not a one‑off experiment, but consistent use in real workflows: research, portfolio operations, compliance checks, reporting, client communication and business development. That is the gap firms must cross to achieve genuine enterprise return on investment from AI.

So far, most have not crossed it.

ChatGPT reports nearly 900 million monthly users, and 56 percent of Americans say they use AI, yet 85 percent of the workforce does not have a value‑driving AI use case and 25 percent do not use AI for work at all. Even in tech‑forward environments and language‑intensive functions, most AI use remains surface‑level.

For RIAs, whose business models depend on repeatable processes, scalable client service and rigorous controls, that disconnect goes straight to operational leverage and competitiveness.

A Workforce of Experimenters, Not Practitioners

Three years after the launch of ChatGPT, the report finds that most people are still AI beginners.

Roughly 70 percent of the workforce fall into the “AI experimenter” category – people who use AI for basic tasks such as summarizing meeting notes, rewriting emails or getting quick answers. The second‑largest group, “AI novices,” accounts for 28 percent of workers: those who do not use AI, or tried it a few times and bounced off.

Only a tiny fraction sit at the levels that matter for real productivity change:

• AI practitioners: 2.7 percent of the workforce
• AI experts: 0.08 percent

In other words, less than 3 percent of workers are either practitioners or experts – people who put AI to work in their workflows and see significant productivity gains.

The report’s summary is plain: 97 percent of the workforce are using AI poorly or not at all. A quarter of workers say they save no time with AI. Forty percent say they would be fine never using AI again.

For RIA employers, that raises an uncomfortable question: if your staff report “using AI,” is it actually changing the economics of your firm, or just shaving a few minutes off an email?

The “Use Case Desert” In Financial Work

The report names the central obstacle: employees are in a “use case desert.”

The biggest challenge in using AI is not learning to prompt; it is knowing what to use AI for.

Across thousands of workers, a pattern emerges. Employees understand how to use a large language model in principle, but they bounce off when they cannot identify a concrete, role‑specific use case.

• 26 percent of respondents say they do not have a work‑related AI use case at all.
• 60 percent of the workforce say their existing use cases are beginner‑level.
• After analyzing 4,500 reported work‑related AI use cases, the authors judged only 15 percent likely to generate return on investment for the business.

The impact on time follows. Fewer than one‑third of knowledge workers report saving four or more hours per week with AI, even though most organizations would need to target 10 or more hours per employee per week to see meaningful ROI.

For RIA firms, this “use case desert” is especially damaging. Many of the highest‑value workflows – portfolio rebalancing prep, KYC documentation review, surveillance, performance and attribution reporting, client review packs – are complex, regulated and cross‑functional. They do not lend themselves to simple “summarize this paragraph” type prompts.

When staff are left on their own to “experiment,” they tend to use AI on the periphery – writing assistance, meeting notes – rather than in the core of their work.

Use Cases That Will Not Move the P&L

The report’s analysis of what workers actually do with AI shows just how far reality is from the automation‑heavy future described in conference keynotes.

Among the top 10 work‑related use cases, by share of knowledge workers:

1. Google search replacement – 14.1 percent
2. Draft generation – 9.6 percent
3. Grammar and tone editing – 5.7 percent
4. Basic data analysis – 3.8 percent
5. Code generation – 3.3 percent
6. Ideation and brainstorming – 3.2 percent
7. Meeting support (for example, notes) – 2.7 percent
8. Document summarization – 2.0 percent
9. Learning and skill development – 1.6 percent
10. Task and process automation – 1.6 percent

Viewed another way:

• 59 percent of reported AI use cases are basic task assistance.
• More than 25 percent have no relevant role in larger processes or workflows.
• Only 2 percent are judged to be advanced use cases.

Grouping by category shows a similar pattern. Research use cases account for 19.6 percent of workers; writing accounts for 18.1 percent. Both are heavily skewed toward beginner‑level usage: one‑off copy suggestions and basic informational searches, not embedded workflow redesign.

For RIA employers, that matters because surface‑level use rarely shifts core metrics such as assets per adviser, clients per service team, time to onboard or cost to serve. Those numbers move when AI is integrated from end to end: from document ingestion to decision, not just paragraph polish.

Thin Slices of Time Saved

Because most workers are using AI for basic tasks, its impact on productivity is modest.

The report’s time‑saved breakdown looks like this:

• 24 percent of workers report zero time saved.
• 21 percent save less than two hours per week.
• 23 percent save two to four hours.
• 18 percent save four to eight hours.
• 8 percent save eight to twelve hours.
• 6 percent report saving more than twelve hours.

More proficient workers do reap more benefit. AI practitioners are 1.8 times more likely than AI experimenters to save four or more hours a week, and 20 times more likely than AI novices to reach that threshold.

In an RIA environment, those practitioners might be the rare employees who have, for example, automated elements of suitability checks, built internal copilots for repetitive client reporting, or systematized research synthesis. They are, by the report’s count, a tiny minority.

Investment Is Up. Proficiency Is Not.

The report does not suggest that companies are idle. If anything, they are accelerating their AI investments.

• 63 percent of respondents say their company has an AI policy.
• 50 percent have access to an AI tool.
• 44 percent receive AI training from their company.

Those investments are not meaningless.

• Employees in companies with a formal AI strategy are 1.6 times more proficient than those without one.
• Employees with access to AI tools are 1.5 times more proficient than those with no access.
• Employees who have been trained on AI are 1.5 times more proficient than those who have not.
• Employees whose managers expect AI usage are 2.6 times more proficient than those whose managers discourage it.

Since March 2025, access to a formal AI policy is up 17 percent. Clear guidelines for AI usage are up 16 percent. Investment in AI tools and platforms is up 2 percent.

Yet despite this, most employees remain stuck at the experimenter level. Workers who have undergone AI training score, on average, 40 out of 100 on AI proficiency.

The likely explanation is straightforward: most programs remain focused on access, safety and prompting. They give people an LLM, define the guardrails and provide a framework for writing a good prompt. That is a necessary foundation. It is not a bridge from usage to value.

For RIAs, the distinction is the difference between advisers who occasionally ask AI for email drafts and operations or compliance teams that systematically redesign workflows to strip out days of cycle time.

Executives Believe AI Is Succeeding. The Rest of the Firm Does Not.

Perhaps the most consequential finding for investment employers is the leadership perception gap.

C‑suite respondents overwhelmingly believe their AI deployments are going well. A large majority say their company has a clear AI strategy, that tools exist with a clear access process, that there are clear, enforced policies connected to that strategy and that employees are encouraged to experiment and build their own solutions. Nearly half say they have widespread adoption with open sharing of use cases and best practices.

Individual contributors – the people processing forms, reconciling data, preparing client materials – tell a different story.

• Eighty‑one percent of C‑suite leaders say, “We have a clear, actionable policy that effectively guides AI use,” compared with 28 percent of individual contributors – a 53‑point gap.
• Eighty percent of leaders say tools exist with a clear access process, compared with 39 percent of ICs.
• Seventy‑one percent of executives say they have a formal AI strategy, against 32 percent of ICs.
• Sixty‑six percent of the C‑suite feel encouraged to experiment and create their own AI solution, compared with 25 percent of ICs.
• Forty‑eight percent of leaders say there is widespread adoption with open sharing of use cases and best practices, compared with 8 percent of ICs.

At the same time, 75 percent of C‑suite members say they are excited about AI’s implications, and 94 percent report almost complete trust in its contributions. The majority, 57 percent, use AI for work daily; only 2 percent do not use AI for work at all.

For RIA firms, whose senior leaders are often the most exposed to vendor narratives and conference stages, this optimism is understandable. It is also dangerous. If leadership measures success through adoption and access metrics, they can remain insulated from the operational reality inside their operations, compliance and service teams.

Individual Contributors: Those Who Could Benefit Most Get the Least

The report singles out individual contributors – knowledge workers without direct reports – as the group that benefits least from their company’s AI resources.

• Least likely to have clear access to an AI tool.
• Least likely to receive AI training.
• Least likely to be reimbursed for AI tools.

The breakdown is revealing:

• Thirty‑two percent of individual contributors report clear access to AI tools, compared with 57 percent of managers, 72 percent of directors and VPs, and 80 percent of the C‑suite.
• Twenty‑seven percent of ICs have received company AI training, compared with 48 percent of managers, 68 percent of directors, 70 percent of VPs and 81 percent of executives.
• Only 7 percent of ICs are reimbursed for AI tools, compared with 26 percent of managers, 44 percent of directors, 33 percent of VPs and 63 percent of the C‑suite.

As a result, individual contributors are more likely to be anxious or overwhelmed by AI, less likely to trust it and least likely to say it is having a transformative impact on their work. Manager support is weak and worsening: support for AI use among ICs has fallen 11 percent since May 2025. Only 7 percent say their managers expect daily AI use, and only about one‑third receive encouragement to use it.

In the RIA world, many ICs sit in precisely the roles where AI could have the greatest leverage: operations analysts, client service associates, compliance staff, research associates. If they lack tools, training and expectations, the firm’s AI strategy will stay trapped in slide decks.

Leaders and Laggards – And Where Finance Sits

The report ranks industries by AI proficiency on a 100‑point scale. While it is not dedicated to financial services alone, the placement is instructive:

• Technology: 42
• Finance: 36
• Consulting: 35
• Manufacturing: 34
• Media: 33
• Real estate: 32
• Food and beverage: 29
• Education: 29
• Healthcare: 28
• Retail: 27

Finance is near the top, but the scores are low across the board. A 36 out of 100 is not a position of strength; it is an early‑stage capability.

Across functions, the pattern is similar. Engineering and strategy roles lead with proficiency scores of 41 and 39. Business development and human resources follow at 37. Marketing, finance and legal, product and operations trail close behind. Customer service and support sits last at 27, despite being a domain with obvious potential for AI transformation.

Many functions are not using AI for their most obvious use cases:

• Fifty‑four percent of engineers do not use AI for writing or debugging code, scripts or formulas.
• Fifty‑six percent of marketers do not use AI for creating first drafts of content.
• Eighty‑seven percent of product managers do not use AI for creating prototypes.

Transposed into an RIA setting, those numbers invite basic questions: How many research teams systematically use AI for idea generation and cross‑document analysis, versus occasional summarization? How many compliance teams embed AI in surveillance, testing and documentation, versus rewriting policies? How many client service teams use AI to support personalization at scale, versus fixing typos?

What RIA Employers Should Do Now

The report’s leadership imperatives can be read as a practical agenda for RIA firms.

First, stop measuring AI success by access and adoption rates alone. If 55 percent of your workforce uses AI weekly but only 15 percent have value‑driving use cases, your adoption metrics are misleading. For RIAs, success should be tracked in time saved per role, quality and maturity of use cases, and concrete business outcomes such as reduced errors, faster turnaround and higher adviser capacity.

Second, treat use case development as a core competency, not a personal side project. The workforce is not stuck because it cannot prompt; it is stuck because people do not know what problems AI can solve in their specific role. RIA leaders can sponsor function‑specific use case libraries – for research, trading, compliance, operations and client service – and create role‑based playbooks that move employees from basic to intermediate and advanced use.

Third, bridge the individual‑contributor gap quickly. The people doing the most repetitive, automatable work have the least access to tools, training and support. This is backwards. Firms should prioritize IC enablement in operations, compliance, research and service, standardize access to approved AI tools and require managers to identify and track at least three AI use cases for each direct report.

Fourth, recognize that training got you to the starting line, not the finish line. A 40 out of 100 proficiency score after training suggests that current programs are teaching the wrong things. Shift emphasis from how to use AI safely to how to identify and redesign workflows where AI can eliminate or transform work.

Fifth, close the executive awareness gap. If C‑suite members believe deployments are succeeding while ICs report minimal impact, you have a data visibility and culture problem. Implement regular skip‑level sessions focused on AI adoption barriers, and ask executives to shadow employees as they attempt to use AI in their daily work, not just during polished demos.

Sixth, accept that the proficiency bar will keep rising. The gap between experimenters and practitioners will widen as AI capabilities advance. Build continuous learning infrastructure now – internal academies, communities of practice, coaching and clear progression paths from basic to intermediate to advanced use cases in each function.

The Real Question for RIA Firms

The report’s deeper point is simple. AI transformation in the workplace is not primarily about buying technology. It is about redesigning how work is done: task by task, workflow by workflow, role by role.

For RIA employers, that places AI squarely in the territory of operating model, skills and culture. It is no longer enough to ask whether advisers, associates and analysts can use AI.

The question is whether the firm has shown them what to do with it – and whether leaders are willing to measure the answer honestly, not through adoption dashboards but through the hard numbers of time, quality and risk.

Until that happens, much of the industry will remain where this report finds corporate America: optimistic at the top, experimental at the edges, and far from the returns many have already pencilled into their plans.