5 takeaways on closing the gaps AI pilots leave behind

Most health plans have deployed AI, but few can point to real changes in how they operate. During our recent AHIP webinar, we moderated a panel with Dr. Nate Henderson, CMO and VP of Quality at BlueCross BlueShield of South Carolina (BCBS SC); Gigi Yuen-Reed, PhD, Chief Data & AI Officer at Cohere Health; and Dr. Brian Covino, Chief Medical Officer at Cohere Health. 

Here's what they said it takes to connect clinical intelligence across functions, so value builds instead of stalling in silos.

1. Siloed pilots optimize a workflow, not the operation

AI pilots deliver real wins. Dr. Covino highlighted that real-time approvals in utilization management (UM) streamline the authorization process for a significant volume of requests–reducing the administrative burden on clinicians and minimizing delays for members. However, because pilots are often built and measured within a single function, the clinical insights they generate are rarely shared with quality, care management, payment integrity, or appeals teams.

"If the UM process can't communicate with downstream processes, then you're essentially kicking the problem down the road." — Dr. Covino, Chief Medical Officer at Cohere Health

2. Clinical data shows up where you least expect it

One Cohere Health partner was working to close a Stars measure on osteoporosis treatment in women. Using Cohere's connected clinical intelligence, 30% of the information* needed to close that measure came from data gathered during imaging requests. That's a quality result from a UM program never designed to move Stars. 

Health plans already have the data. What's missing is the ability to use it at scale, moving clinical information across functions easily and getting it to the right team at the right time.  That starts with breaking down the data barrier: a common language and semantics across all functions. Once data is connected, insights that are used to depend on manual review come to the surface on their own. Dr. Henderson noted BCBS SC is now gaining a clearer, faster understanding of staff efficiency and provider-requesting behavior.

3. Connected context leads to more consistent decisions and better-timed outreach

Most reviewers can find the information they need if they have enough time and tools. The real gap is timeliness and precision: getting the right context in front of the right person at the moment of decision. When that context lives in separate systems, determinations can conflict across requests for the same member, and follow-up depends on someone remembering to act.

A connected patient 360 view brings that context into the workflow:

"The power of a platform like Cohere's is that it enables the reviewer… to know that something has been determined one way in a previous request, and to allow that to be part of the consideration." — Dr. Henderson, CMO and VP of Quality at BCBS SC

That same context turns every request into a chance to get ahead of what the member needs next:

• Factor prior determinations into new requests for the same member
• Engage members at the right moment. For example, if a procedure is denied because the member smokes, that's a chance to help them quit by enrolling them in a specific health plan-led intervention.

4. Govern AI by defining its scope, then owning it continuously

Dr. Yuen-Reed distilled her governance guidance into two core principles:

• Define the scope: AI works best when its role is clearly defined. It can synthesize information across a health plan while being limited to specific tasks, with humans remaining in the loop throughout.
• Never stop overseeing it: Regardless of how narrow the scope or how many guardrails are in place, ongoing monitoring is non-negotiable.

5. Sequence investments to build trust, then connect

Dr. Covino advised starting in one place, building comfort and trust in the AI's accuracy, then adding more functions. The largest value comes once the full enterprise is linked, from UM to payment integrity, care management, appeals, and quality.

At BCBS SC, an early proof point was electronic submissions through the Cohere Health platform reaching well over 80%. As plans scale, AI must be domain-aware. Processes that look administrative are often actually clinical, so plans shouldn't "just throw AI at the problem."

Every workflow a health plan runs produces clinical intelligence. The plans that pull ahead will carry it into the next decision, where it can close a care gap, inform a review, or reach a member at the right moment.

Where is clinical intelligence getting lost in your health plan? Watch the webinar highlight reel to hear more from the panel on connecting clinical intelligence, governing AI at scale, and sequencing investments to prove value early.

‍

^{* Based on Q3 2026 Cohere Health data}

_‍