Aetna: provider-insurer trust stalls at 6.0/10, but anti-fraud AI wins favor with providers

By Beinsure · November 4, 2026.

Trust between healthcare providers and insurers in the US held steady in the third quarter of 2026 at 6.0 out of 10, compared with 6.1 in the second quarter and 5.4 in the first. That is according to Aetna, a CVS Health subsidiary, in its quarterly survey of 727 healthcare professionals, conducted by Morning Consult with a margin of error of ±4 points.

The most striking AI finding: according to the survey itself, nearly 90% of respondents believe AI is already helping to combat fraud, waste and abuse, and 30% want insurers to use it to flag suspicious claims. That was the most frequently chosen action. In addition, 94% think the problem needs to be addressed, and 57% put its annual cost at between $100 billion and $499 billion. Beinsure notes that these are respondents' estimates, not independently measured losses.

Administrative burden remains the top concern for 67% of providers, seven points lower than in the previous quarter. Among those who cite it, prior authorization (the process by which an insurer approves a treatment before covering it) is the main factor for 27%, down from 31%. And more than half of those who expect improvements attribute them to automation and industry commitments to speed up approvals, compared with around 40% before. Updating medical records is the largest source of paperwork (58%), and almost a third want real-time patient information.

Our take: three things need to be kept separate. First, this is a survey by an insurer about how its partners perceive it; it measures opinion, not outcomes. That 90% “believe” AI is already helping does not show that it reduces fraud or that it does so accurately. Second, the endorsement is for one specific kind of AI: fraud detection, which benefits both sides. The same technology applied to denying authorizations or claims is far more sensitive, and the survey does not ask about that. If models start deciding coverage, the key question, as with other enterprise agents, will be who decides, with what data, and who can correct it through human review.

Third, there is an underlying positive signal. Healthcare providers expect their biggest future problem to be having enough time with each patient, not paperwork. If automating the bureaucracy (records, authorizations, interoperability) genuinely frees up those hours, it would be a concrete example of AI giving time back to what matters: care. More broadly, the healthcare back office is among the areas most exposed to automation, so in the short term there will also be disruption to administrative jobs.

What to watch: whether the improvement in trust holds without AI being used to tighten denials, and whether independent data emerges on fraud prevented and errors made. Until then, the study is a good gauge of expectations, with the obvious caveat that it is published by one of the interested parties.