DC consultants stopped debating AI and started using it on the paperwork

The fence emptied fast. Only 14% of the defined contribution consultants and advisers who responded to T. Rowe Price's sixth annual Defined Contribution Consultant Study, published Tuesday, said they were still evaluating artificial-intelligence tools or unsure how they would use them, against 44% a year earlier, and none of the respondents reported avoiding AI altogether, down from 8% in 2025.

Where that use has landed is narrower than the collapse of the undecided bloc suggests. The gains are concentrated in meeting notes, which 75% of respondents now process with AI, up from 36% a year earlier, and in research, data analysis and report generation, which two-thirds now automate, up from 25%. Client outreach has risen to 47% from 22%, and the most common daily applications are drafting or summarizing email, cited by 69%, and meeting preparation and summaries, cited by 35%. Jessica Sclafani, who heads the retirement strategist team at T. Rowe Price, credits supply as much as appetite: "There's an availability component to it. More firms are rolling out approved AI resources and tools and making them available to larger swaths of their employee base."

The gains have landed in the documentation layer, the least contested ground in the business. No plan sponsor picks a consultant on the strength of its meeting minutes, and the tasks AI has absorbed—notes, summaries, first drafts—are tasks no firm ever sold as its edge. The harder analytical work has not moved: 41% of respondents never use AI for plan design and benchmarking analysis, 33% never for participant engagement or educational content, and 23% never for client prospecting, while the shares for forecasting market trends, identifying investment opportunities, and making better investment decisions are 6%, 8%, and 8%—close to rounding error.

Consultants have automated their overhead and shielded their judgment, a rational instinct that looks temporary from here: benchmarking, plan design and proposal support are structured-data chores these tools already handle well, and the firm that automates the middle of the job answers more searches with the same headcount. The sorting in defined contribution consulting will be decided not by who writes better notes but by who lets software touch the work clients believe they are paying for—and that work has been migrating toward the adviser's desktop becoming the point of assembly for retirement capital.

T. Rowe Price's own crosstabs complicate the usual story about careful firms falling behind: consultants and advisers at firms with formal AI governance and guardrails were more likely to report frequent use than those with less governance, a result Sclafani calls counterintuitive. "Oftentimes, when you present people with lots of rules and guardrails, it can make people less likely to embrace the change … [but] with AI, it's the opposite," she says. "The rules actually encouraged adoption."

That finding makes the 14% still evaluating a more awkward group than the headline number suggests. A firm with a policy has decided and trained its people to it; a firm without one is left watching whether staff use unapproved tools on client material or use nothing at all, and the seventh study will likely show governance and adoption moving together, because at this point they measure the same thing: whether anyone at the top has decided.

There is a commercial logic to an asset manager paying for this research: T. Rowe Price reaches plan menus through the same gatekeepers it surveys, and its capital has followed the same route—its $19 billion deal for F/m in August brought fixed-income and managed-account capabilities into the retirement business. A study of what the consultants who sit in front of plan sponsors do with their tools is, in effect, a map of the channel that decides which funds reach a menu.

The quietest number in the survey carries the most commercial weight: 14% of respondents said their plan sponsor clients had no opinion on retirement income, down from nearly 60% in 2021, which moves in-plan income from a conference panel to a standing agenda item. Sclafani describes the expectation that has replaced the blank stare: "Plan sponsors expect their consultants and advisers to be able to offer them an annual review of the in-plan retirement income landscape." The build-out is already visible in the asset data, with income-linked target-date strategies up 18% in the first half of the year, per Sway Research.

As this publication has argued, the durable retirement-income default is the required minimum distribution, and the sponsor data here cuts toward that reading. What sponsors want, on Sclafani's account, is an annual scan of what has come to market and what fits a plan, so the consulting work runs on a calendar instead of a shelf review. Menus will keep adding income options, while billable hours shift to withdrawal design and product fit.

The 41% is the number to watch. If plan design and benchmarking are still AI-free when the seventh study lands, the tools will have been absorbed as a productivity gain for firms and a discount nobody handed the client. If they are not, the profession has a pricing question it has not had to answer before: what an hour of consultant judgment is worth when the first draft takes seconds.