The forward-deployed health plan

When we examine why a health plan's departments are shaped as they are, a useful distinction emerges: some functions exist because the underlying decisions are genuinely difficult, and others exist principally because the work is voluminous. A utilization management department employs dozens of nurses not because each authorization requires deep clinical judgment; most are routine. It employs them because thousands of requests arrive every week, through five separate intake paths (the care platform, the behavioral health delegate's portal, the post-acute vendor's queue, fax, and phone), each on its own regulatory clock.

The same distinction, applied across the building, sorts the organizational chart into two lists. On the volume side sit claims-edit resolution, appeals and grievances intake, enrollment reconciliation, member correspondence, HEDIS abstraction, and the retrospective chase for risk documentation; this is work that arrives in thousands of instances, follows policy, and must be documented against a deadline. On the judgment side sit actuarial pricing, network strategy, benefit design, vendor selection, and the minority of clinical reviews that genuinely require a medical director's attention. The volume list is where the headcount is; the judgment list is where the margin is made or lost. It is worth pausing on that asymmetry, because the traditional plan's organizational design treats the first list as its core and staffs accordingly.

Given that arrangement, consider where the plan's scarcest expertise actually spends its time. Nurses rekey fax attachments into the care platform; pharmacists reconcile files from the PBM; compliance officers draft boilerplate correspondence against regulated clocks. In engineering organizations, this pattern has a name and a known correction: the forward-deployed model, in which scarce expertise is embedded directly at the points where consequential decisions are made, while standardized systems carry the volume behind it. The traditional health plan is organized in precisely the opposite way. Its experts are stationed behind the queues, consumed by the volume list, and its consequential decisions (pricing, network, benefit design) receive whatever attention remains. This is not a criticism of any plan's leadership; it is the accumulated result of staffing to volume for decades because there was no alternative.

What has changed, and recently, is the feasibility of separating the two lists. Examined closely, the volume work shares a small number of repeatable shapes: documents arrive and are read against policy and history to produce a decision with a deadline and an audit trail; regulated language is generated from case facts and reviewed by a person who owns the signature; evidence is located in a chart and tied to a code or a measure. Current AI systems, properly governed and in several cases hosted entirely inside the plan's own environment, now handle those shapes reliably enough to carry most of the volume list. The judgment list they do not touch, which is rather the point.

For a plan's leadership, the practical exercise follows directly. Take the organizational chart and sort every function into the two lists, honestly; where a department contains both kinds of work, as utilization management does, split it on paper. The volume side becomes a portfolio of candidates for governed automation, each with a number it must hit before anything is built. The judgment side becomes the question of where the plan's experts should sit once the queues no longer consume them; forward, we would argue, at the decision points the traditional design has always starved. The contrast between those two pictures (the plan as staffed today and the plan as sorted) is, in our experience, the most useful single exhibit an executive team can produce this quarter.