The Cost Line Nobody Puts on the Slide

Ask the Question

Walk into any large recordkeeper or TPA and ask what is really driving per-participant cost. You will hear compliance. You will hear servicing. You will hear distribution. Almost nobody says technology.

McKinsey does. In its 2025 analysis of the retirement industry, technology support is the second-largest cost category a recordkeeper carries, behind only the commissions paid to bring in new business. And the lever most firms have pulled to contain it has not worked: outsourcing technology support, McKinsey found, "hasn't resulted in notable efficiency gains." In-house technology cost per participant went down slightly. Third-party vendor cost went up.

That is the blind spot. Technology has quietly become one of the largest cost lines in this business, and it does not appear on the slide because everyone has agreed to treat it as the cost of doing business. It is not. It is a tax we have stopped questioning.

Where the Margin Went

Two numbers explain the last decade better than any strategy deck.

Over the ten years to 2023, the industry cut its average cost per participant by about one percent a year. Over the same ten years, recordkeeping administration fees fell 25 to 35 percent. NEPC's 2026 survey of plan sponsors puts the figure at 26 percent over the decade.

The cost came down a little. The price came down a lot. The gap between those two lines is where the margin went, and it is why McKinsey found that mega and large 401(k) plans now generate just 3 percent of recordkeeping profits.

The platform is the reason the gap cannot close. Everest Group's research on the largest US recordkeepers found that eight of the top ten, holding 73 percent of defined contribution assets and serving 66 percent of participants, run on custom-built systems that are predominantly mainframe-based. Across the industry, 80 percent run home-grown systems and 77 percent are mainframe-based, with at least a 28 percent shortfall in the people who can maintain them. Everest's conclusion is blunt: the high cost of running these systems "has bulked up recordkeepers' total cost of ownership and squeezed their operating margins."

None of this is news to the people who run those platforms. Which brings me to the part of this argument that matters most.

The People Who Already Know

The technology leaders I talk to at recordkeepers and TPAs understand this better than anyone in their building. They inherited platforms built in the 1990s. They have watched twenty years of integration layers accumulate around them, each one rational on the day it was added. They have asked for modernization budgets and been told to do more with less, because the core recordkeeping function is seen, in Everest's words, as "a cost center." Then, when the bill arrives, they are the ones asked to explain it.

Look at how the industry actually spends. Everest found that investments to overhaul the legacy core account for less than 25 to 35 percent of transformation budgets; the other 65 to 75 percent goes to the layers around it: customer interaction, analytics, infrastructure. The industry has been funding the veneer and starving the engine. That is not a decision technology leaders made. It is one they have been living with.

So this is not an argument against IT. It is an argument for the people in IT who have been saying this out loud inside their own firms and not being heard. Nobody in this industry should be more invested in what comes next than the people who have been carrying the weight of what came before.

We Have Been Here Before

A decade ago, hosting a recordkeeping system in the cloud was close to heretical. Compliance would never allow it. The regulators would never accept it. Too risky, too new, too much fiduciary exposure.

Today nobody blinks. The Cloud Security Alliance's most recent survey of financial institutions found 98 percent using cloud services and 59 percent storing or processing regulated banking data in them. FIS, whose platforms serve 68 million participants, has spent four years moving Omni and Relius off a COBOL backbone built more than thirty years ago and onto a cloud platform slated to launch this summer. The unthinkable became the default in about ten years.

It is worth remembering how that happened. It was not recklessness. Technology leaders did the unglamorous work: the control frameworks, the audit trails, the shared-responsibility models, the vendor due diligence that made compliance comfortable and regulators satisfied. The cloud did not win the argument. The governance did.

AI Is the Same Story, Faster

AI is on the same trajectory, and the clock is shorter. It is not a question of whether AI runs core retirement processes — data reconciliation, plan administration workflows, exception handling — but when. McKinsey's own prescription for the cost problem is that AI "offers new ways to improve productivity" and that firms should reimagine "workflows from end to end" in operations and technology. Recordkeeping executives interviewed ahead of this year's TRAU roundtable were more direct: several predicted needing only half their current staff within two or three years.

I would put it differently. McKinsey calls AI a productivity lever. We think it is the operating model. The distinction matters, because a lever gets bolted onto the existing process, and the existing process is the problem.

And just like the cloud, the firms that get there first will not be the ones that moved carelessly. They will be the ones whose technology leaders built the governance early: humans reviewing exceptions, every decision traceable, every model auditable, every plan sponsor able to see exactly what was automated and what was not. A plan sponsor who is skeptical of AI is not being unreasonable. They are asking the same question they asked about the cloud, and they deserve the same quality of answer.

When the shift lands, the firms still treating legacy overhead as a badge of stability will be explaining to plan sponsors why the fee went down 30 percent and the cost did not. The firms that backed their technology leaders will be setting the price everyone else has to match.

Our Bet

We made ours. We built MAPTIVA™ to take the manual, stitched-together work out of retirement data processes — payroll reconciliation, file validation, migration mapping — with the exception handling and audit trail that a compliance team can sign off on. It is not a bolt-on to a legacy process. It is a view of what recordkeeping operations look like when AI runs the process and people govern it.

If you are the technology leader who has been making this case inside your firm and not getting heard, I would like to talk. You are not the cost problem. You are the reason it can be solved.

---

Sources

• McKinsey & Company, The US retirement industry at a crossroads, April 2025
• Everest Group, Why Retirement Plan Defined Contribution Recordkeepers Need to Modernize Their Plan Administration Systems, June 2020
• Everest Group, Modernization in the Retirement Plan Industry and the Journey to the Cloud, 2021 (hosted by the SPARK Institute)
• NEPC, 20th Annual Defined Contribution Plan Trends and Fee Survey, February 2026
• Cloud Security Alliance, State of Financial Services in Cloud, 2023
• WealthManagement.com, FIS Global Retirement Launches Cloud System in July, May 2026, and Record Keepers Face Existential Threats From Fee Cuts, June 2026