Consultants automate the paperwork, not the fiduciary call

The population is small enough to list and broad enough to speak for the market: 36 consulting and advisory firms, more than 160,000 defined contribution plan sponsor clients, and $10.3 trillion in DC assets under advisement, roughly 72% of the $14.2 trillion DC plan market measured against Investment Company Institute data as of December 31, 2025. When firms controlling that share answer the same questionnaire—the sixth annual Defined Contribution Consultant Study from T. Rowe Price, covered by InvestmentNews—the average starts to behave like a forecast.

On artificial intelligence, the forecast has already arrived. In 2025, 44% of respondents described their use of AI as too early to evaluate; this year, 14% did, a 30-percentage-point collapse in twelve months. Thirty points in a year is fast for an industry whose sales cycle is measured in plan years. The likelier explanation is one of labeling: the tools doing the work in 2026 are not dramatically different from the tools available in 2025, but firms have stopped calling the work experimental, and 'too early to evaluate' was a statement about readiness that firms have fixed.

The work being automated is specific, and it stops well short of plan design. Seventy-eight percent of firms use AI routinely to improve internal workflows, 67% to prepare for client meetings, and 47% for outreach; the numbers then thin out sharply to 12% for plan design and benchmarking and 9% for participant engagement.

Those last two figures carry more weight than the 78%, because a model-drafted internal memo is a productivity claim a firm makes to itself, while a glidepath recommendation or an in-plan default is a decision the firm answers for later—to a client, a consultant, or a plan sponsor's counsel. The distance between 78% and 12% is not a measure of what the tools can do; it is a measure of how much paper a firm needs standing behind the answer, and the industry has been candid in the numbers, if not in words, that it can produce the memo faster than it can produce the file.

The study's most useful finding inverts the usual intuition about controls: firms operating under formal AI governance policies used the tools roughly 50% more frequently across business functions than firms without such policies. Guardrails, in this reading, work as a permit: a firm with a written policy knows which uses it can defend and stops relitigating the question every time a new assistant shows up, while a firm without one defaults to no. The governance gap, rather than the model, is what separates the firms pulling ahead, and the private-assets half of the same dataset is where that gap starts to cost money.

The other half of the study asks what consultants expect plans to own. Respondents rated how likely various alternative strategies were to be incorporated into DC plans over the next 12 to 24 months on a scale of one to four, and private credit scored 2.6, up from 1.7 in 2024, while private equity climbed from 1.6 to 2.2 over the same two years. Private credit's gain of 0.9 points is half again as steep as private equity's 0.6, and it is the only one of the two sitting clearly above the midpoint of the scale.

A 2.6 is conviction, and allocation is still some distance away. Between a likelihood score and a seat on a plan menu sit valuation spreads, daily-liquidity mechanics, capital-call logistics, and a recordkeeper willing to build the plumbing, none of which a questionnaire resolves. The scores show direction, and a nine-tenths-of-a-point move on a four-point scale is not a rounding error; likelihood questions tend to move before allocations do, which suggests the request is already arriving at the consultant's desk.

The 12% on plan design deserves more attention than the 78% on workflow because the diligence problem for private assets in DC plans is document-heavy: valuation policies, capital call histories, fee layers, quarterly reporting that a sponsor's counsel will eventually ask to see. Those are the chores firms have already automated elsewhere, and they are the chores that would matter most to the private-credit conviction the same consultants just reported. The low number reads as a sequencing decision rather than a technology verdict—governance first and the defensible uses after.

The study's framing pairs private assets with personalized retirement income as the forces reshaping how plans are built, though the reported detail does not break out separate likelihood scores for guaranteed income products. That leaves the annuity question, which has been moving from conference panels toward actual menu adoptions, outside this particular dataset.

T. Rowe Price is not a bystander here: its $19 billion acquisition of F/m, announced in August and covered in these pages, was a purchase of the fixed-income engine that collective investment trusts demand, with the next front inside the fund itself. An asset manager building toward custom, income-bearing target-date construction has a direct commercial interest in consultants reporting rising conviction about private assets, and it has now published six editions of evidence that they are. The findings may be right, but the questionnaire belongs to a distributor, and a census of the channel is worth having regardless of who writes the questions.

Two lines in the seventh edition are worth watching: whether the 9% on participant engagement has moved, which would mean consultants have built a source trail they trust for participant-facing decisions and would be the real AI story, and whether private credit crosses 3, which would mean the peer pressure has started. Until then, 2.6 is what conviction looks like before anyone has to sign.