Auto-adjudication is the outcome, not a separate process. Every claim runs through the same sequence — eligibility, benefit matching, pricing, cost-sharing, edit checks — and a claim that clears every step cleanly comes out the other side auto-adjudicated: priced, cost-shared, and finalized with no adjudicator involved. A claim that trips any single check gets routed to a human instead. The rules engine doesn’t decide “should this be automatic” as a separate question; automatic is just what happens when nothing stops it.
What makes a claim pend instead
A claim pends when the system hits something it isn’t authorized to resolve on its own — a missing or unmatched prior authorization, eligibility that doesn’t clear cleanly, no fee schedule match, a suspected duplicate, an unconfirmed coordination-of-benefits question, or an edit flag on the billed data. See pended claims for the full list of triggers, how the reason code travels with the claim, and how the adjudicator queue is worked once a claim lands there.
A worked example
Two claims for the same code, CPT 99213 (established-patient office visit), same day, same plan. Claim A comes from an in-network provider with a contracted rate of $95 for that code, the member’s eligibility is active with no gaps, and no prior authorization is required — it clears every check without an adjudicator touching it. Claim B is the identical CPT code, but from an out-of-network provider with no contracted rate on file. It pends: someone has to price it against a usual-and-customary benchmark — say $140 for that code in that geography — before it can be decided. Same code, same plan, two different paths — the difference is entirely in what each claim brings with it, not in the system’s tolerance for automation.
Why this matters operationally
The size of your pended queue is a direct function of plan design and data quality, not a fixed property of the software. Clean, current fee schedules reduce pricing pends. Real-time eligibility feeds reduce eligibility pends. Authorization data that’s actually synced with the plan reduces auth pends. None of that is something a vendor can promise you in a single number, because it depends on your provider mix, your plan’s benefit complexity, and how current your source data is — which is exactly why Claimaro doesn’t publish a benchmark auto-adjudication rate. Any number quoted without your plan design and claim mix attached is marketing, not information.
What you can control is what the pended claims cost you once they land. A claims adjudication engine that attaches the pend reason to the claim — not just a status flag, but why it stopped — means your adjudicators are working exceptions instead of re-diagnosing routine claims from scratch. For a TPA running multiple client books at once, that difference compounds: a rules engine tuned per client keeps the straight-through rate as high as each plan’s own design allows, instead of forcing every client through one generic threshold.