Strategic Acquisition Brief · Confidential
Navy Federal has rebuilt the rails and the interface and declared wealth management a growth mandate under augmented intelligence. MaxiFi is the computation layer that goes underneath: for a member’s facts and assumptions, it solves — not guesses — the lifetime plan, every dollar of taxes and benefits computed under current law, and it asks first whether the family needs to take risk at all. The law layer is deterministic, reproducible, auditable. Accurate enough to stand behind with a guarantee.
In the last eighteen months Navy Federal has, in its own words, laid the foundations and started building on them. Dietrich Kuhlmann, President and CEO, May 2026: “The next chapter is cloud transformation and the adoption of augmented intelligence. We call it augmented intelligence because we’re focused on technology that enables our team members to better serve our members, not replace them.” He has challenged his teams to grow the credit union’s wealth-management offerings, “areas critical for veterans and their families,” and named the destination: “the preferred and most trusted financial institution for service members, veterans, and their families.”
On August 3, 2026 Navy Federal Investment Services relaunched Digital Investor on Apex Fintech Solutions’ real-time platform — real-time account opening, self-directed investing, and automated, goals-based portfolios. Diane Young, Chief Operating Officer of Navy Federal Financial Group: “it’s critical that our infrastructure is able to scale without sacrificing the seamless experience our members expect and deserve.” Branch, contact center, online and mobile now run on a single omni-channel platform; the core transition will end batch banking.
A goals-based portfolio is only as good as the goal, and for a military family the goal is arithmetic: pension plus Social Security, the Survivor Benefit Plan election, VA income, 42 state codes. Real-time balances invite the next question — what can this family safely spend, and does it need to take risk at all? Trust in advice is either a brand promise or a computed property. Only one of those survives an audit.
The industry’s planners — and the automated, goals-based portfolios built on them — start from a spending target, find the probability of reaching it, and raise the probability by adding equity. Spending is fixed, risk is the starting point, the plan is one period long, and the household’s own aversion to loss never enters the arithmetic. For a military family with a pension, a survivor election and a mortgage, that is not an approximation of the right answer. It is the wrong answer in the family’s own terms.
MaxiFi starts safe. It computes the most a household can sustainably spend if it takes no risk at all, and then treats risk as an option, evaluated against the household’s own risk aversion, across the whole lifetime, with spending free to adjust. Professor Kotlikoff’s August 14 case:
| Strategy — Jim, 62, $2 million, Comfort Index 7 of 9 | Versus the TIPS-only base case |
|---|---|
| 80% stocks / 20% bonds | 13% worse — like 13% of spending confiscated every year for life |
| 20% stocks / 80% bonds | 4% worse |
| The same 80/20, if Jim were highly risk-tolerant | 54% better |
| Social Security claiming and Roth timing, no risk added | +$218,077 lifetime discretionary spending |
For a member-owned institution whose members are, by the nature of their service, the most cautious savers in the country, this is the member-first question. “Most trusted” is the institution that asks it first.
A large language model is a horizontal capability. In any function where a wrong answer is catastrophic, no serious operator ships the raw model to the customer: a purpose-built application layer sits on top of it, enforcing the rules, the computation and the audit trail. Intuit runs a deterministic tax engine under TurboTax’s assistant and will not let the model guess the numbers. MaxiFi is that layer for lifetime financial planning, and no one else has it.
Integrate once and it cascades: under the assistant on the omni-channel platform; under Digital Investor, where the engine’s sustainable-spending answer becomes the goal input instead of a questionnaire — the Apex relationship is fed, not displaced; under the advisors in more than 150 branches, who produce a computed plan in the meeting; under the contact center, where “what’s my balance” becomes “what can I safely spend.”
MaxiFi is Professor Laurence Kotlikoff’s lifetime planning engine, built at Economic Security Planning, Inc. over three decades. It answers the fiduciarily correct question — what is the most a household can sustainably spend, and does it need to take risk to get there — rather than the aspirational one, and it computes the answer under the law as written.
Federal income tax and 42 state codes; Social Security (claiming, spousal, survivor, earnings test); Medicare Part B and D with IRMAA; ACA subsidies; RMDs; Roth conversions; pensions, annuities, life-insurance need, housing transitions — and consumption smoothing across every year of a household’s life.
Dynamic programming over the whole horizon, with spending endogenous and the household’s own risk aversion inside the objective; investment strategies compared by expected lifetime utility from a safe base case. The law layer is deterministic: rerun it and check.
An API and a maintained law table, plus the advisor application. Annual law updates by one engineer with two backups; any competent team can be trained on the cycle. Thirty years of real households were the validation method.
The rulebase — thirty years of encoded, test-suited federal, state, Social Security and benefit rules — and the optimization that solves them jointly. A quant team can verify it faster than it could rebuild it; the months of shipping unvalidated numbers while it tried are the cost that matters.
Military households carry the hardest arithmetic in American personal finance. A rule of thumb on any of these is not an estimate; it is a wrong number the family lives with.
The facts, plainly. Navy Federal Investment Services is a FINRA member and an SEC-registered investment adviser. FINRA’s Regulatory Notice 24-09 (June 2024) put member firms on notice that the existing rules — Reg BI, communications standards, supervision — apply in full to advice produced with generative AI; the notice forecloses “the technology was new” as a defense. The NCUA’s 2026 supervisory priorities put AI and third-party AI vendors inside the examination through the existing third-party due-diligence letters, and the examiner’s question is direct: how do you monitor the AI for accuracy?
An assistant in front of 15.5 million members that approximates a claiming age, a conversion or a survivor election produces the same wrong answer for every family that asks. After notice, that is not an incident; it is a pattern with a paper trail.
For a member-owned institution the cost is not an equity re-rating. It is restitution to the families that relied on the number, the examination that follows, and the erosion of the one asset Navy Federal has spent 93 years building: being the institution the military community trusts with money.
With an engine whose law layer is deterministic, “how do you monitor for accuracy” has a one-line answer: the assistant’s numbers are the engine’s numbers, and they can be rerun and checked against the law tables in force on the plan date. Not a better disclaimer — a different mechanism.
Because the warranted event — a computational error — is objectively decidable, a bounded Accuracy Guarantee is underwritable. A conventional planner cannot offer it: its spending target is exogenous, so there is no correct answer to warrant.
The evidence ladder. Public substantiation; a 30-minute live demonstration on one real household; clean-room verification of households the acquirer chooses, under LOI and exclusivity; integration after closing. No pilots. Possession never precedes closing.
Every one of 15.5 million member households has a pension, benefit or claiming question. The engine turns “talk to an advisor” into a computed plan in the app, the branch and the contact center — the conversion path from banking relationship to Navy Federal Investment Services. Provable in the first quarter of ownership.
“Advice you can hold us to.” For a not-for-profit that returns roughly $4.5 billion a year to its members — about $470 each, in Mr. Kuhlmann’s figure — a guarantee on the arithmetic is the member-first statement in its purest form, and the one the four big banks cannot make.
Two-thirds of Americans bank primarily with one of four big banks; Mr. Kuhlmann has named that as the opportunity. “Most trusted” becomes a property a member can test rather than a claim a member must take on faith.
Navy Federal has said it wants to help the larger credit union movement succeed. Owned inside Navy Federal Financial Group, the engine can be offered to the other 4,300 federally insured credit unions as the movement’s computation utility — a revenue line that exists only for a credit-union owner.
The documented, reproducible answer to the accuracy question, for the assistant, the advisors and the examiners, comes with the deal; it is not the price of the deal. So does an active base of households already planning on the method, and Professor Kotlikoff’s continuing national voice — including his writing on military retirement. Figures in the data room.
This is a deliberately narrow process. The engine is going to the acquirer where it does the most good for the most real people, and where the acquirer’s own structure makes correctness a dividend rather than a marketing claim. A member-owned institution serving 15.5 million military households, that measures itself by what it returns to members, is the clearest case of that we know. Professor Kotlikoff intends to keep contributing to the product, to help the acquirer integrate it, and to remain its public voice.
MaxiFi is being offered through a focused strategic process — the engine, its IP, and thirty years of R&D. The preference is an acquisition; that is where the strategic value sits. The next step is a 30-minute orientation: one real military household, our machine, in the room — the safe base case first, then the risk question — while a frontier model is asked to match it. Tuesday, September 29 or Thursday, October 1, 10:00 AM PT / 1:00 PM ET. Nothing is deployed, nothing left behind. First conversations with strategics are underway; we expect to narrow the field in the second half of October.