Eric Dreyfus introduced a cool AI tool that analyzes PBM contracts. It worked. And Anthem is not very happy…………….
PBM Ethicist | AI x Strategy x Big 3 PBM Contracts | SiriusB.Ai | former Fortune 200 SVP | former Aon Middle Market Practice Leader & Food,Ag, and Bev Practice Leader
June 5, 2026
THE KINCAID RX DEFENSE PBM CONTRACT X-RAY IQ INTELLIGENCE SERIES
That is not the behavior of a market leader ignoring a nuisance. Shame on you, Anthem, Inc. How dare you try to intimidate a leading, actuarially-focused broker. He is an honest man. What is wrong with you?
That’s the behavior of an incumbent responding to a threat.
For decades, healthcare has hidden extraordinary economics behind extraordinary complexity.
Entire business models depend on the assumption that employers, patients, and even fiduciaries will never have enough information to ask the right questions.
Eric Dreyfus introduced a cool AI tool that analyzes PBM contracts. It worked. He is the kind of producer that global consulting houses like Aon and Marsh McLennan Agency stopped attracting 15 years ago.
In Eric’s case, suddenly, Anthem lawyers appear.
Maybe the cease-and-desist is justified. Maybe it isn’t. Time will tell.
What is already clear is that innovation has advanced beyond theory and entered territory important enough to provoke a response from one of the industry’s most powerful organizations.
In business, attention is a currency.
A cease-and-desist from a giant is sometimes the market’s way of acknowledging that a challenger has become impossible to ignore.
A Cease-and-Desist Is Not a Defense. It Is a Disclosure.
Anthem’s lawyers responded to an artificial-intelligence engine that reads pharmacy benefit contracts. The response told us more than the contracts ever did.
Anthem’s General Counsel sent a cease-and-desist letter to one of the most innovative brokers in the industry. For more PBM insights, visit SiriusB.AI.
The target was not a competing carrier. The target was not a rival consultant pitching a lower fee. The target was a piece of software — an artificial-intelligence engine that reads pharmacy benefit contracts and renders the unreadable legible.
That is not how a market leader responds to a nuisance.
That is how an incumbent responds to a threat.
For a generation, American healthcare has protected extraordinary economics with extraordinary complexity. The complexity is not an accident of a regulated industry. The complexity is the product. It is the asset that has to be defended, because the moment a contract becomes readable, the margins inside it become arguable. And margins that can be argued are margins that can be lost.
So when a tool appears that can read the contract — quickly, cheaply, and at scale — the rational incumbent response is not to compete on price. It is to retain counsel. The letter that followed should be read for what it is. Not a refutation. A receipt.
I. The Moat Was Always Made of Complexity
Begin with the structure, because the structure explains the lawyer.
A self-funded employer hires a pharmacy benefit manager to administer drug claims. The plan sponsor is told, in effect, we will use our scale to get you better pricing than you could get alone. This is a reasonable promise. It is also a promise that is almost impossible to verify, because the instruments that govern it are written to be unverifiable.
Consider the vocabulary a plan sponsor must master simply to know what it is buying. Average Wholesale Price, a benchmark that no one actually pays and that has been described in litigation, for decades, as a number unmoored from acquisition cost. Wholesale Acquisition Cost, a different benchmark, applied selectively. Maximum Allowable Cost lists, which the administrator controls and is frequently permitted to revise without notice.
Brand and generic effective-rate guarantees, blended across an entire book so that a shortfall in one category can be cured by an overage in another. Rebate definitions that quietly exclude the categories where the real money moves. Spread pricing, the simplest mechanic of all — bill the plan one number, pay the pharmacy a lower number, keep the difference — and the one most reliably obscured by the surrounding architecture.
None of this is illegal. That is the point worth sitting with. The opacity is not contraband. It is craftsmanship. The contract is engineered so that every individual clause is defensible in isolation and the aggregate is unreadable in combination. The plan sponsor signs because the alternative is to not have a pharmacy benefit, and because the document arrived with the implicit assurance that sophisticated people had already vetted it.
The economic model rests on a single load-bearing assumption: that the employer, the patient, and even the fiduciary will never have enough information, time, or technical fluency to ask the precise question that unwinds the structure.
For most of the system’s history, that assumption was sound. Not because anyone was negligent, but because reading a pharmacy administrative services agreement properly — cross-referencing every defined term against every pricing mechanic, reconciling the guarantee language against the reconciliation language, tracing where a dollar of rebate is recognized and where it is routed — is genuinely hard, and genuinely expensive, and historically required the most expensive professionals in the building.
The moat, in other words, was never primarily legal. It was cognitive. It was built out of the cost of understanding.
II. The Variable That Changed Was the Cost of Reading
Here is where the engineering matters, and where the incumbent’s anxiety becomes rational rather than reflexive.
For thirty years, the binding constraint on PBM accountability was not the law. The fiduciary duties existed. The benchmarks were knowable. The spread was, in principle, calculable. The constraint was the cost of the analysis.
A forensic read of a single administrative services agreement — the kind that actually finds the basis points hiding in the definitions — was the province of senior ERISA counsel and specialized actuaries billing at rates that priced the exercise out of reach for all but the largest plans. A mid-market employer with two thousand lives was never going to spend eighty thousand dollars in professional fees to audit a contract it had been told was standard.
So the contract went unread. Not because no one was capable of reading it, but because reading it did not clear the cost-benefit threshold for the people who had signed it.
Artificial intelligence does not change the law. It changes the cost curve. And in a market whose entire defensibility was built on the cost of understanding, collapsing that cost is not an incremental improvement. It is a phase change.
What does it actually mean to “X-ray” a contract? Strip the metaphor and the mechanism is concrete. A properly built system ingests the executed agreement and its exhibits as structured data rather than prose. It extracts every defined term and builds a dependency graph — which definitions feed which pricing provisions, which guarantees are blended against which, where a term defined narrowly in Section 1 quietly governs an entire pricing schedule in Exhibit C.
It identifies the clauses that limit the plan’s audit rights, the ones that cap look-back periods, the ones that route data ownership to the administrator at termination. It compares the contract’s benchmark language against the mechanics those benchmarks are supposed to discipline, and it flags the gap between what the guarantee appears to promise and what the reconciliation provision is structured to deliver.
The work is not magic. It is parsing, classification, and cross-referencing performed at a speed and a marginal cost that human review can never match. The model does in seconds what the eighty-thousand-dollar engagement did in weeks — and it does it for the next contract, and the one after that, at a cost that rounds to zero.
This is the part the cease-and-desist understood perfectly, even if the letter never said so. The threat was never one clever broker with one sharp insight. A single sharp insight can be litigated, settled, or simply outlasted. The threat is automation. It is the permanent and irreversible collapse of the cost of reading the contract — for every plan sponsor, in every market, at the same time.
You cannot send a cease-and-desist to a cost curve. So you send it to the firm that operationalized one.
III. The Risk Does Not Stay With the PBM
Now layer in the law, because this is where the structure stops being an interesting margin story and becomes a question of personal exposure.
ERISA imposes on plan fiduciaries a duty of prudence that is, by design, demanding. The standard is not what the fiduciary actually knew. The standard is what a prudent person, acting under the prevailing circumstances and familiar with such matters, would have done. Courts have long held that prudence is a process measured against the information reasonably available at the time. The duty is procedural, but it is not soft. A fiduciary who fails to investigate what a prudent fiduciary would have investigated has breached the duty, regardless of whether the underlying decision happened to turn out well.
Notice what happens to that standard when the cost of investigation collapses.
For decades, a plan sponsor could credibly say that a clause-level forensic audit of its PBM contract was not reasonably available — too costly, too specialized, beyond the practical reach of a plan its size. That defense was never airtight, but it was plausible, and plausibility is often enough. The technology dissolves the plausibility. When a forensic read of the contract becomes fast and inexpensive, the universe of “information reasonably available to a prudent fiduciary” expands to include exactly the analysis the incumbent would prefer no one ran.
The fiduciary who could have known inches toward the fiduciary who should have known. And under ERISA, should-have-known is the doorway through which liability walks.
The relevant provisions are not obscure. Section 404 sets the duties of loyalty and prudence. Section 406 prohibits certain transactions between the plan and parties in interest. Section 408(b)(2) conditions the reasonableness of service-provider arrangements on adequate disclosure of compensation — direct and indirect — and the indirect compensation in pharmacy arrangements is precisely the category the architecture is built to obscure. The functional-fiduciary doctrine extends these duties to anyone exercising discretionary authority over plan administration, which is a longer list than most plan sponsors assume. And the Consolidated Appropriations Act of 2021 layered a statutory documentation requirement on top of all of it, obligating plans to obtain and scrutinize compensation disclosures that, in many arrangements, were never meaningfully examined.
The regulators have been narrating the underlying economics in plain language. The Federal Trade Commission’s second interim staff report, released in January 2025, found that the three largest pharmacy benefit managers generated more than $7.3 billion in revenue above the drugs’ estimated acquisition cost on a set of 51 specialty generic medicines dispensed by their affiliated pharmacies between 2017 and 2022. The report separately estimated roughly $1.4 billion attributable to spread pricing — billing plan sponsors more than the amount reimbursed to pharmacies. In one instance the agency cited, a pulmonary-hypertension generic acquired for about $27 was reimbursed at an average of roughly $2,106, a markup exceeding 7,700 percent. The excess revenue was concentrated, with the top ten specialty generics accounting for $6.2 billion of the total, and it was growing at a compound annual rate of 42 percent.
Those are not allegations from a plaintiff’s brief. They are findings from the nation’s competition regulator, on the public record, describing the mechanics of an industry whose contracts most plan sponsors have never read.
Place those two facts beside each other. The economics are now publicly documented. The tools to find those economics inside any individual contract are now cheap. The duty to investigate is measured against what is reasonably available. The exposure, in other words, does not stay with the pharmacy benefit manager. It migrates upstream — to the plan sponsor, to the benefits committee, to the individuals who signed and renewed without looking, because the obligation to look was theirs all along.
That migration is the real reason the letter went out. The incumbent is not only protecting its own margin. It is protecting a market in which its clients have a powerful incentive to never read what they signed — because reading it converts a comfortable arrangement into a documented fiduciary problem.
The Only Honest Product Review
It is worth pausing to admire the situation on its own terms.
A cease-and-desist letter is, functionally, the only product review written by someone who has actually read the output and understood its implications. Customers offer testimonials. Competitors offer dismissals. Only the party with genuine exposure offers a legal demand to stop. The letter is the highest form of validation available in this market precisely because it is the most expensive to send and the most revealing to file. No one retains a general counsel to suppress a tool that does not work.
There is a particular comedy in watching an industry that markets itself on transparency respond to a transparency tool by reaching for an injunction. The pitch decks promise visibility, alignment, and partnership. The legal department promises consequences for anyone who builds the visibility the pitch deck promised. Both documents are signed by the same organization. They simply travel to different rooms.
And there is a quieter absurdity underneath. The contract was always available to the plan sponsor. It was sitting in the file the whole time. The only thing that changed is that it can now be read in an afternoon instead of a quarter. An entire defensive posture has been constructed around the proposition that a document the customer already owns must not become legible to the customer who owns it. That is not a market. That is a magic trick objecting to the house lights.
The humor is not that any of this is false. The humor is that it is all precisely, demonstrably true, and that the system functioned beautifully right up until someone built a machine that reads.
The Litigator Take
Set the satire aside and speak the way the lawyers actually speak when the door is closed.
If the contract language and the claims data line up the way the public record suggests they often do, the question in the room is not whether there is exposure. The question is how it is quantified, who carries it, and how far back it runs.
The senior litigator’s instinct here is not directed at the pharmacy benefit manager. The PBM’s contractual position is, in most cases, carefully constructed and well defended; that is what the complexity bought. The instinct is directed at the plan sponsor — the fiduciary who can no longer claim that the analysis was beyond reach.
In a fiduciary-breach posture, the defense that wins is process. The plan that investigated, documented its investigation, asked the hard questions, and memorialized the answers is in a strong position even if the underlying numbers are ugly, because ERISA polices the prudence of the process, not the perfection of the outcome. The plan that cannot produce that file is in a different posture entirely. Discovery does not ask whether you knew. It asks what a prudent fiduciary in your position would have known, and then it asks why you did not.
A composite of the firms that try these cases — the Wachtell instinct for where the real leverage sits, the Skadden discipline on the paper trail, the Kirkland appetite for the aggressive theory, the Quinn Emanuel willingness to take it to a jury — would converge on the same unglamorous conclusion. The decisive evidence in the next wave of these matters will not be a smoking-gun email. It will be the absence of a document. It will be the file the fiduciary should have built and did not, in a year when building it had become cheap.
The candid version, then, is this. The plaintiff’s bar does not need the technology to be widely adopted. It only needs the technology to have existed and been affordable, because that is the moment the duty to use it arguably attached. The cease-and-desist did not slow that moment down. It memorialized it.
Enter the X-Ray: Rx Defense PBM Contract X-Ray
Which brings us to the engine the incumbents are correctly worried about — not the early tremor that drew the first letter, but the category arriving in force behind it.
Rx Defense PBM Contract X-Ray is a clause-level forensic analysis engine for pharmacy benefit administrative services agreements. It does not argue with the moat. It renders the moat legible. And once a moat is legible, it is no longer a moat. It is a liability map.
The platform was built for exactly the problem this article describes. It is vendor-agnostic by design; it does not care which administrator wrote the contract or how the brand was marketed, because it reads the instrument, not the relationship. It ingests an executed agreement and its exhibits and performs the analysis that historically required a six-figure engagement: it extracts and cross-references every defined term, traces each pricing mechanic back to the definition that controls it, and surfaces the gap between what a guarantee appears to promise and what the reconciliation provision is structured to deliver. It isolates spread where spread is present. It decodes the benchmark machinery — the effective-rate blends, the cost-basis selections, the maximum-allowable-cost discretion — and states in plain terms what each one does to the plan’s economics. It maps the clauses that quietly limit audit rights, compress look-back windows, and route data ownership to the administrator at termination, when the plan needs that data most.
The output is not a slide. It is not a pitch. It is a fiduciary-grade documentary file — the contemporaneous, time-stamped record of a prudent investigation, produced in the register that ERISA counsel, plan committees, and boards actually need. In a market architected on opacity, the forensic file does something a sales presentation cannot: it does not ask the fiduciary to take action. It forecloses the option of inaction. The file is the pitch deck, because once the file exists, doing nothing has become the documented decision.
Read against the core problem of this piece, the function is exact. The threat to the incumbent was never a competitor’s lower fee. It was the collapse in the cost of reading the contract — the conversion of an unreadable instrument into a documented record. Rx Defense PBM Contract X-Ray is the industrialization of that conversion. It is the upstream documentation layer that resolves the single objection that has kept plan sponsors trapped inside structures they were told not to question: we cannot prove what is happening inside the contract. The X-Ray proves it. And a plan sponsor who can prove what is inside the contract is a plan sponsor finally free to leave it.
That is why the letters go out. Not because the tool is wrong, but because it works — and because a market that depends on its customers never reading the contract cannot survive its customers reading the contract in an afternoon.
The Forward View
Step back from the single letter and the trajectory is not difficult to read.
The information asymmetry that built this market is being arbitraged away, and asymmetries do not return once they break. The regulators have moved from inquiry to documentation, and documentation is the predicate for enforcement and for private litigation alike. The Consolidated Appropriations Act converted a best practice into a statutory duty, and statutory duties generate paper, and paper generates discovery. The technology that makes the contract legible is getting cheaper and better on the curve that all such technology follows, which is to say monotonically and without apology.
The market will sort into two postures. There will be plan sponsors who treat the new legibility as an opportunity — who run the analysis, build the file, ask the questions while asking is still voluntary, and use the documented record to renegotiate or to leave. And there will be plan sponsors who treat it as a threat to be managed by not looking, and who will discover, in a courtroom or in a regulator’s request, that not looking was the one option the law had already foreclosed.
The cease-and-desist will not decide which posture wins. It has already told us which one the incumbent fears.
And here the analysis earns a final, plainer observation, because the basis points are not the whole of it. Underneath every blended effective rate and every routed rebate is a person. The plan participant rationing a specialty generic that was acquired for twenty-seven dollars and billed at two thousand. The hourly employee whose paycheck quietly funds a spread he will never see itemized. The benefits manager who is also a neighbor, signing in good faith for a workforce she is trying to protect, unaware that the document she trusted was engineered so that her trust could not be checked. A benefits system is not finally a margin structure. It is an instrument of human dignity — the means by which work is supposed to translate into care. When opacity becomes the mechanism by which the people the system was built to serve are made to subsidize the people who administer it, that is not merely an inefficiency to be optimized. It is an inversion to be corrected. Transparency, in that light, is not only a compliance technology. It is the precondition for the system to do the human work it was created to do.
That is the deeper reason the house lights matter. Not because illumination embarrasses the incumbent — though it does — but because the people standing in the dark were never the ones who chose it.
A cease-and-desist from a giant is, in the end, the clearest signal the market can send. It is the incumbent’s own admission that the challenger has become impossible to ignore.
The lawyers read the output and reached for an injunction.
That is not a denial.
That is a disclosure.
Written by Jeremiah Franklin Shrack. The David character written about in this story is Eric Dreyfus and the Goliath is Anthem, Inc. Ann Lewandowski, MLS
This article analyzes structural incentives, contract architecture, and fiduciary-risk dynamics in the pharmacy benefit market. It is commentary and analysis, not legal advice, and it makes no allegation that any named organization has acted unlawfully. References to a cease-and-desist letter reflect circumstances described to the author. Figures attributed to the Federal Trade Commission are drawn from the agency’s second interim staff report on pharmacy benefit managers (January 2025). Plan sponsors and fiduciaries should consult qualified ERISA counsel regarding their specific arrangements.
