Strategic Acquisition Brief · Confidential
On May 5, Dario Amodei and Jamie Dimon shared a stage at “The Briefing: Financial Services,” and Anthropic shipped ten financial-services agent templates — already in production at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa. The one domain where a wrong number does lasting harm — how much a household can safely spend, for life — is one a model that samples tokens can only approximate. MaxiFi is the deterministic engine that computes the provably correct lifetime plan — the answer Claude can incorporate and stand behind. Built over 30 years by BU economist Laurence Kotlikoff.
On May 5, 2026, Anthropic and JPMorgan co-hosted “The Briefing: Financial Services,” with Dario Amodei and Jamie Dimon sharing a stage for the first time. Anthropic used the moment to ship ten ready-to-run financial-services agent templates — pitch builders, credit-memo drafters, KYC screeners, month-end closers — and disclosed a $1.5 billion services joint venture with Goldman Sachs, Blackstone, and Hellman & Friedman to bring Claude into the daily operations of mid-market companies. Claude is already in production at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa. That is a deliberate, well-funded move into the surface where a confident-but-wrong number does the most harm: how much can a household safely spend, for the rest of its life?
As the market writes the growth story for Claude in financial services, the hardest question it will ask is not can Claude answer a money question — it already does, for institutions and, increasingly, their clients. The question is whether the product can be right about someone’s money, and prove it. That question is now live at the institutions Claude already serves — and it is exactly the question MaxiFi was built to answer.
MaxiFi resolves it — a validated, deterministic engine that computes the mathematically correct lifetime plan, rather than approximating one. It is the substance behind a vertical that has already shipped to the largest names in finance.
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 user — a purpose-built application layer sits on top of it, encoding the domain’s rules and holding the model to them, turning raw generation into an action that is correct, defensible, and safe to deploy. It is already how high-stakes AI gets built: Intuit runs a deterministic tax engine under TurboTax’s AI and won’t let the model compute the numbers; patient-facing healthcare AI runs inside a safety layer, not on a raw model; and no one boards a plane flown by an unverified black box.
Personal financial planning is exactly such a function, and getting it wrong is its own kind of disaster: the retiree who runs out of money at 82, the family under-insured by a million dollars. It is also precisely where a model, left alone, fails — because it reaches for the same rules of thumb the incumbents use, and in this domain approximation is not “close enough”; it is wrong, in ways that compound every year to the household’s detriment.
Built from the ground up over thirty years, MaxiFi is the proprietary workflow that computes the correct, auditable answer under the actual tax and benefit rules — the function users actually want performed. The model doesn’t have it. The user doesn’t have it. The incumbents approximate it — and, in this domain, approximating it means getting it wrong, to the household’s cost. Value accrues to whoever owns that trusted workflow: the model layer commoditizes, while durable value moves up to the layer that owns the user’s trust and performs the function.
MaxiFi is the financial-planning platform of Economic Security Planning, Inc., built over more than three decades by Professor Laurence Kotlikoff of Boston University. It uses consumption smoothing and dynamic programming to compute the single, mathematically optimal lifetime plan — solving simultaneously across Social Security strategy, federal and state taxes, Roth-conversion sequencing, withdrawal order, estate planning, and upside investing.
Goals-based tools and rule-of-thumb calculators answer “What is the chance you hit your number?” MaxiFi answers “What is the optimal path, and how much can I spend today without jeopardizing tomorrow?” It is not a better simulator. It is a different class of engine — the computed answer Claude can incorporate and stand behind.
Prof. Laurence Kotlikoff — William Fairfield Warren Professor at Boston University; Harvard Ph.D.; former Senior Economist on the President’s Council of Economic Advisers; named by The Economist among the 25 most influential economists. Larry Kotlikoff intends to stay on with the acquirer in whatever capacity best serves the product — architect, spokesperson, advisor.
MaxiFi’s economics build on Nobel-laureate work, and Nobel laureate Robert Merton teaches with MaxiFi at MIT Sloan as an “outstanding science-based lifecycle and retirement management platform.” Featured in Bankrate’s “Best financial planning software of 2025” roundup, cited as best for near- and long-term tax planning and the decumulation phase.
Patent-winning algorithms refined over 30+ years, built from economic theory rather than scraped text — exactly the kind of intellectual property a probabilistic model cannot reverse-engineer by sampling tokens.
MaxiFi is already in the wealth-advisor channel via its Pro subscription — fee-only planners use it as top-of-funnel lure and client-retention glue. The acquirer inherits a paying installed base that extends naturally into Claude’s institutional and enterprise reach.
Every planning tool — and every AI model trained on them — answers a lifetime-money question by approximation: rules of thumb, replacement ratios, withdrawal heuristics, simplifying assumptions. As models improve, they converge on one another and the industry mistakes that agreement for accuracy. But a perfect mimic of an approximation is still an approximation — still wrong — and the error compounds a little further every year, to the household’s detriment.
MaxiFi does not approximate. It computes — iteratively, multivariately, and simultaneously across taxes, benefits, longevity, and cash flow, year by year for a whole life — the one optimal plan. The same inputs produce the same answer, every time, with a full audit trail from data to result. It is provable, not merely confident: the only answer that holds up when someone with an adverse interest checks the math. That claim is about the computation — the optimization and the tax and benefit math are exact and inspectable — not about predicting markets. MaxiFi does not claim to know the future; it claims that, given the facts, its arithmetic is reproducible, auditable, and holds up to scrutiny.
The path is straightforward: MaxiFi incorporated into Claude — a computed, correct core surfaced wherever people ask about their money. How that incorporation is engineered is Anthropic’s call to make; what MaxiFi brings is the one input that is provably correct going in, so what comes out can be, too.
SEC Regulation Best Interest and FINRA Rule 2111 (suitability) govern the substance of a financial recommendation regardless of the interface that delivers it. A model that generates a plausible-sounding number is making a recommendation in substance, whatever it is called in the interface. The rules do not pause for new technology, and being “AI-generated” is not a liability shield.
At the sector level, AI-generated money advice offered without a fiduciary safeguard is beginning to draw its first suits across the industry. That is a category-wide dynamic, not a claim about any one company’s litigation history. The direction of travel is toward accountability for the substance of the answer, not the label on the product that delivered it.
A correct-by-construction engine addresses the underlying exposure directly: if the math is right, reproducible, and auditable, the answer holds up to scrutiny on its own terms — the same assurance a security vendor gives its customers when it lets them stand behind what it ships.
CBS MoneyWatch (May 7, 2026) ran an identical retirement question — a 50-year-old single woman retiring at 65 — through Claude and ChatGPT. The verdicts diverged. MIT’s Andrew Lo was quoted on the underlying structural point: today’s consumer AI carries no best-interest duty. Kotlikoff was quoted describing the risk that AI “may do more harm than good” when it mishandles claims like Social Security timing or substitutes an average for a maximum life expectancy.
A concrete, checkable example: AI engines trained before the One Big Beautiful Bill Act (enacted July 2025) told users the federal estate-tax exemption would “sunset” on January 1, 2026 — reverting to roughly half its level. In fact, the Act permanently raised the exemption to $15 million per person starting in 2026. A model repeating pre-2025 training data would confidently tell a household to rush an irrevocable estate move it no longer needs — a costly, hard-to-reverse error delivered with total confidence. A computed engine, fed current law, does not carry stale assumptions forward as fact.
Neither example is about any single company’s brand. It is the same structural point twice: confidence is not correctness, and an answer’s value depends on the currency and correctness of the computation behind it — not the fluency of the sentence delivering it. The May 7 test named Claude directly; the gap it found is the one MaxiFi closes.
Larry’s Economics Matters Substack — 137,000+ subscribers — has run a six-post sequence testing named frontier engines against MaxiFi on dollar-specific household problems. Two posts name-test Claude; three name-test ChatGPT. The variance across engines on identical, checkable prompts is the proof.
“The AI said John and Jane can spend approximately $52,000 per year in discretionary spending. MaxiFi’s demonstrably correct answer — verifiable by inspecting its reports — is $63,382.”
Read the head-to-head →“AI’s best hope of providing accurate economics-based planning is by pairing a conversational front end with MaxiFi’s computed results — precisely correct, not clearly pretend.”
Read the structural argument →Estate-planning head-to-head naming Claude’s output against MaxiFi’s computed result — the same structural gap, applied to estate and gifting strategy.
Read the estate test →“The median household leaves $182,370 of lifetime Social Security on the table. AI tells Jane a job change adds at most $35K in lifetime benefits when the right answer is $168K.”
Read the Social Security test →Head-to-head against Gemini and ChatGPT on the shape of lifetime spending — the “retirement smile” — comparing generated narrative against MaxiFi’s computed trajectory.
Read the retirement-smile test →Claude’s Roth-conversion sequencing tested against MaxiFi’s optimized path — MaxiFi’s computed strategy came out 72.7% better on the same household facts.
Read the Roth-conversion test →Acquiring MaxiFi acquires the megaphone these pieces ship from — pointed, with credibility no one in the category can match, at the financial-services vertical Claude just staked out on stage with JPMorgan. Larry Kotlikoff intends to stay on with the acquirer in whatever capacity best serves the product, turning a category critic into the correctness narrator for whichever engine incorporates MaxiFi. The CBS finding is the named, neutral proof; the Substack series is the dated, dollar-specific record behind it.
The same mechanism — being right about a client’s money — is revenue today and a durable, higher-quality earnings stream tomorrow. It runs as one funnel: attractant → adhesive → loyalty → continuity. Correctness draws the client in on the highest-stakes question they will ever ask an AI; correctness keeps them, because a wrong answer on money is disqualifying and a right one is not replaceable with a cheaper competitor; retention becomes recurring, lifelong engagement; and that same engagement is what a market rewards with a stronger equity story, because it is durable and hard to copy.
Questions about one’s own money are the highest-intent, highest-trust, highest-willingness-to-pay questions a client ever asks an AI. Being right is the attractant, the adhesive, and the source of loyalty — recurring, lifelong engagement, and qualified demand for every other agent Anthropic sells into financial services.
A correct-by-construction engine retires the largest overhang on the money-advice story: being confidently wrong with people’s money, at scale, at the exact moment AI money advice with no fiduciary safeguard is drawing its first suits, sector-wide. And there is exactly one MaxiFi — in a rival’s hands, it strengthens their answer and weakens Claude’s.
The market rewards a dollar of earnings that is defensible and low-risk with a stronger equity story. MaxiFi makes the same revenue from Prong 1 proprietary, un-copyable, and trust-based — and removes a tail risk in the same motion. A moat plus a removed risk is what a premium private valuation is built on.
The attractant→adhesive→loyalty→continuity funnel is the same mechanism twice: it is the revenue engine today, and it is the durable, high-quality earnings stream that strengthens Anthropic’s equity story tomorrow. One asset, financing both sides of the story.
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. Continuity de-risks it: Larry Kotlikoff intends to stay on with the acquirer in whatever capacity best serves the product — architect, spokesperson, advisor. The next step is a 30-minute live demonstration: MaxiFi solves a real household’s plan while Claude is asked to match it. The gap is the thesis.