The Direct Compact: What Emerges When the Current Architecture Falls

The twelve articles in this collection do not make twelve separate arguments. They make one argument from twelve angles. The bundled insurance product is the wrong architecture for the 1-to-50 employer market (TOS.01). Community rating accelerated the exit of healthy groups from that market (TOS.02). The uniform contribution norm misrepresents the employer’s actual retention priorities (TOS.03). The broker accountability framework protects brokers more reliably than it protects employers (TOS.04). The TPA exercises de facto plan sponsor authority without bearing the fiduciary consequences (TOS.05). Stop loss carriers determine what is insurable and therefore what the plan can cover, making them the actual architects of plan design (TOS.06). AI is approaching functional replacement of what the small group broker actually does (TOS.07). ICHRA and level funded are converging toward a contributory platform that renders both current products intermediate steps rather than endpoints (TOS.08). Groups below ten lives cannot be insured through any group mechanism in any actuarially meaningful sense (TOS.09). The consumer protection apparatus has become a barrier to simpler arrangements rather than a guarantee of better ones (TOS.10). The specialty drug pipeline is breaking stop loss pricing for the smallest employer segments on a five-year horizon (TOS.11). The most effective health investment a small employer can make for a low-wage workforce may not be insurance at all (TOS.12).

These failures interact. The drug pipeline (TOS.11) accelerates the exit from level funded for micro-groups, which the convergence (TOS.08) partly absorbs through ICHRA, which requires the regulatory accommodation that TOS.10 identifies as blocked. The broker’s displacement (TOS.07) removes the intermediary whose friction currently slows the transition, but also removes the advisor whose counsel might help small employers manage a more complex unbundled arrangement. The actuarial impossibility at small group sizes (TOS.09) makes the transition from group coverage to individual platforms not merely a preference question but a structural necessity for the population that cannot be served by group mechanisms regardless of product design. What looks like twelve problems dissolves, when examined from sufficient altitude, into one structural problem: the small group health benefits architecture was designed for an employment-and-insurance relationship that no longer describes most of the 1-to-50 market, and it is maintained by entities whose revenue depends on its continuation.

The Architecture It Was Designed For

#

The ESI architecture was designed in the mid-twentieth century for a specific market: large employers with stable, long-tenured workforces, offering coverage as a defined benefit that employees received passively in exchange for employment. The tax exclusion under IRC Section 106 assumed that the employer-sponsored product was the right delivery vehicle for most Americans’ health coverage. The ERISA preemption structure assumed that federal uniformity would protect employees in large, multi-state employer plans from inconsistent state insurance rules. The broker intermediation model assumed that the complexity of evaluating insurance products required professional intermediation that the employer could not provide internally.

Each of these design assumptions has eroded. The stable long-tenured workforce has fragmented into the contingent, part-time, freelance, and gig structures that the AARP and National Alliance for Caregiving 2025 report documents as the dominant direction of workforce evolution for caregivers alone — and caregivers are 63 million people, or one in four adults. The defined benefit has become a cost-shifting arrangement in which the nominal coverage the employer provides is nominally coverage and practically unusable at the deductibles the small group market requires, as the KFF 2025 survey data on 53 percent of workers in 10-to-199 employee firms enrolled in plans with $2,000-plus deductibles demonstrates. The tax exclusion creates a subsidy for the product category rather than the health outcome, which is why TOS.12 can construct a plausible case that a non-insurance health investment stack produces better actual health access at the same employer spend.

What Replaces It

#

The direct compact is not a product. It is a relationship. The employer commits a defined dollar amount toward the employee’s health, structured to vary by class, tenure, and retention priority in ways current group plan rules generally preclude (TOS.03). The dollar amount funds a set of coverage components assembled by the employee on a platform that handles eligibility, substantiation, and the compliance documentation that current law requires for tax-advantaged reimbursement. The components are not the current bundled plan’s three functions welded together (TOS.01); they are those functions separated and sourced independently.

Primary care goes to a DPC membership at $75 to $90 per member per month, giving the employee unlimited physician access, same-day appointments, direct communication, and care coordination without claims, without deductibles, and without the prior authorization apparatus that the bundled product generates for routine care. Pharmacy goes to a discount program that prices generics at what they actually cost to produce and distribute, not at the inflated rates the PBM opacity model generates. Network access for specialist and hospital care goes to a direct contract with transparent-price providers, or to the individual market plan the employee selects with their ICHRA contribution, or to both depending on what the platform offers. The catastrophic layer — the only genuine insurance function in the stack — goes to a high-attachment stop loss or individual catastrophic policy that activates above the threshold where the employer or employee cannot reasonably self-fund.

The employee engages with their health. This is the reciprocal dimension the TOS.PRE preface establishes as the third employer objective: make it honest. The employer is not purchasing coverage the employee may never use. The employer is investing in the employee’s actual health access and asking, transparently, that the employee use it. The DPC physician can only manage the employee’s chronic hypertension if the employee shows up. The pharmacy discount only matters if the employee fills the prescription. The compact does not mandate health outcomes; it creates access and asks for engagement in return. That is a different relationship than handing someone a plan document with a $3,000 deductible and a provider directory.

No TPA adjudicates claims under this model for routine care because there are no routine claims. The platform handles substantiation for HRA reimbursements. The DPC practice handles primary care on a membership model with no claims interface. The catastrophic layer generates claims only when costs exceed the attachment point — which, if the DPC and care coordination work as designed, happens less frequently than in a bundled plan where the deductible wall prevents early intervention and drives conditions to acute stages before the plan pays anything.

No broker recommends a product because the platform is the product. The employer’s decision is a contribution level and a platform choice, not a carrier evaluation, a stop loss term comparison, and a TPA assessment. The advisory function the broker performs today — and which AI is positioned to perform within five to seven years for the functional elements of that work — collapses into a simpler question: what combination of DPC membership, pharmacy program, and catastrophic protection, at what contribution level, serves this workforce?

Who Loses

#

The entities that lose revenue if the direct compact reaches scale are not hard to identify.

Bundled carriers lose the small group premium volume that they have been losing steadily to level funded for a decade. The fully insured small group market has already seen enrollment decline 7 percent in 2024 alone per Mark Farrah Associates data, with level funded, ICHRA, and coverage exit all contributing. The direct compact is the next step in that decline. Carriers that compete well in the individual market — where the caregiver’s and the independent worker’s ICHRA contribution goes — gain. Carriers whose business model is fully insured small group lose.

TPAs whose revenue comes from group claims adjudication volume lose the routine claims that no longer exist in the direct compact. Some TPAs pivot to platform operation — managing the ICHRA administration, the HRA substantiation, the catastrophic layer reporting, and the employer-level data analytics that currently justify TPA relationships. The TPAs with technology capability and capital make that pivot. The TPAs whose model is adjudication volume at scale do not.

PBMs lose the opacity that generates rebate capture, spread pricing, and the administrative leverage that independent analyses, including the FTC’s 2024 interim report on PBM practices, have documented as the mechanism by which PBM margin is extracted from plan sponsors without transparent disclosure. In a direct compact, the pharmacy relationship is direct pricing — Cost Plus Drugs, GoodRx, or direct manufacturer access — and the PBM’s intermediary function disappears.

Brokers lose the intermediary role for the segment of the market that treats broker relationships transactionally, which is the majority of the 1-to-50 market as TOS.07 argues. The relational broker — the one whose value is the phone call when something goes wrong — retains value for the employer segment that genuinely needs that relationship and will pay for it. The functional broker — the one who runs census quotes through carrier portals and presents three options at renewal — does not survive AI-assisted platform selection.

The compliance services ecosystem loses the demand that regulatory complexity generates. If the ICHRA platform handles compliance as a product feature, the employer does not need a benefits attorney to draft a wrap SPD. If the regulatory framework simplifies — which TOS.10 identifies as overdue and politically possible — the compliance advisory market contracts.

These entities will resist. Their resistance is not irrational; it is the rational protection of revenue models that current regulation sustains. Their lobbying, their contract structures, their proprietary data arrangements, and their control of distribution channels are all tools of resistance. The question is not whether they resist but whether the pressure from three directions — cost (TOS.11), technology (TOS.07 and TOS.08), and actuarial reality (TOS.09) — overcomes the resistance on a timeline that matters.

What Has to Be True

#

The direct compact does not emerge automatically from the failures documented in this collection. Several conditions must be met, and none of them is guaranteed.

The regulatory framework must accommodate the unbundled components on a tax-advantaged basis. Today, the IRC Section 106 exclusion applies to employer-sponsored health insurance premiums, not to the full stack of DPC membership, pharmacy discounts, transportation assistance, and catastrophic coverage that the direct compact requires. The ICHRA has expanded what can be reimbursed through an HRA, and the 2025 change making DPC memberships HSA-compatible under the One Big Beautiful Bill Act moves the DPC component further toward tax parity. But the full stack does not yet receive the same tax treatment as a bundled group plan. Every percentage point of tax disadvantage relative to insured coverage raises the effective cost of the direct compact and reduces the employer pool willing to make the transition before the bundled model becomes clearly uneconomic.

The technology platform must exist as an integrated product. Today the components are available separately: ICHRA administration through platforms like Thatch, Remodel Health, and PeopleKeep; DPC membership through local practices and emerging employer-DPC networks; pharmacy discount programs through GoodRx and Cost Plus Drugs; catastrophic stop loss through specialty MGUs. No single platform integrates all of these components into a seamless employer purchase and employee experience. That is an engineering problem, not an invention problem, and TOS.08 identifies the assembly as already underway at the component level.

Employers must demand it. Cost pressure is the forcing function. When fully insured premiums become uneconomic and level funded stop loss pricing for small groups follows (TOS.11), employers below the actuarial viability threshold for group coverage face a choice between ICHRA and nothing. The direct compact is the version of ICHRA with a catastrophic layer and a DPC foundation that makes it something rather than something-adjacent-to-nothing. That demand forms gradually and then rapidly, as Hemingway observed about bankruptcy — first slowly, then all at once.

Employees must accept it. Acceptance requires education that has not happened at scale. An employee accustomed to a benefits card that says coverage exists, regardless of whether they can afford to use it, must understand that a DPC physician plus a pharmacy discount program plus a catastrophic policy provides better actual health access than the card. That education burden is real and belongs partly to employers, partly to platforms, and partly to the DPC physician relationship itself — the doctor who knows the employee’s name is a more persuasive argument for the model than any comparison chart.

The Timeline

#

This synthesis does not name a year. It identifies the sequence of forcing functions.

The specialty drug pipeline (TOS.11) creates stop loss pricing pressure that becomes acute in the 2025 to 2030 window as the gene therapy and cell therapy approvals of 2023 through 2025 translate from pipeline events into covered claims. Carriers will respond with segment exits, carve-outs, and pricing that renders level funded nonviable for groups below 15 lives. That displacement is already starting. It creates urgency for the ICHRA convergence (TOS.08) among the employer population that cannot be served by level funded at actuarially viable pricing.

AI capability for broker function replacement (TOS.07) matures on a five-to-seven year horizon for the structured decision problems of small group plan selection. That maturation removes the friction that broker intermediation currently adds to any transition, by making the employer’s product selection decision a platform comparison rather than a professional advisory engagement.

The ICHRA convergence itself is already underway, with 52 percent adoption growth among small employers and 34 percent among large employers from 2024 to 2025, and coverage reaching somewhere between 500,000 and one million lives. The trajectory is not speculative. The question is whether the convergence reaches the catastrophic layer integration that makes it the direct compact, or stops at premium reimbursement without the DPC foundation and the care coordination that the model requires to produce better outcomes than the bundled plan it replaces.

The regulatory accommodation is the slowest and least predictable variable. It has historically followed market behavior rather than led it. Level funded grew for a decade before most state insurance departments had formal guidance on it. ICHRA launched through executive rulemaking after Congress declined to legislate it. The direct compact will likely follow the same path: market adoption will precede regulatory clarity, regulatory clarity will follow because the market has already established the product, and the political opposition from the entities that lose will be overcome when the employer population’s cost pressure reaches a threshold that makes the current system politically indefensible.

The collection’s twelve arguments, taken together, describe a system that is failing predictably, from multiple directions, at a pace that is accelerating. What replaces it is not fully built. The components exist. The demand is forming. The resistance is organized but not permanent. The direct compact is the name for what emerges when the current architecture can no longer bear its own weight.