The Future of Substance Use Care Is AI-Native

Repost from LinkedIn.

A clinician finishes a 50-minute session. The member leaves. The clinician opens their laptop and begins the second session: 45 minutes of documentation. They reconstruct the conversation from memory, translate it into billing codes, fill templates designed for auditors, and hunt through tabs to cross-reference previous notes. Then they move to the next patient.

This happens six times a day. More time documenting than treating. It’s not an edge case. It’s the norm.

Show me the incentive and I’ll show you the outcome

Electronic Health Records (EHRs) were never built for care. They were built for billing. The systems designed in the 1990s now stand between clinicians and patients, optimized to capture every action, convert it into codes, and generate revenue. Clinical workflows became data-entry workflows – serving payers, not patients.

This design misalignment produces predictable distortions. Providers maximize billable sessions because income scales with time spent, not outcomes achieved. The financial incentive isn’t to solve the problem efficiently. It’s to extend the engagement. In that sense, the system isn’t broken, it’s working exactly as intended. The flaw lies in what it was designed to optimize.

The consequences are staggering. The U.S. spends roughly $50 billion each year on substance use disorder (SUD) treatment, yet reaches only about 10% of those in need. The remaining 90%, generate an estimated $135 billion in preventable downstream costs, from ER visits to comorbidity complications to inpatient care.

We’re spending more and helping fewer. The bottleneck isn’t money. It’s infrastructure built around the wrong objective function.

Continue reading