AI alone won't transform your coding. Change management does.
Eight field-tested fundamentals from the Coding Quality Council panel at AHIMA26
Walk the floor at any coding conference this year and half the sessions are about AI. Far fewer are about what happens after the contract is signed.
"The conversation your executive suite is having is very different than the conversation you're having with your coding teams," Lindsey Reilly, CodaMetrix VP of Solution Quality, told the room at AHIMA26. "It's not a communication problem you fix in one email. It's not a training problem you solve in one kickoff meeting. It's a change management problem."
That gap is why our Coding Quality Council chose change management as the focus of its AHIMA26 panel. The Council is a forum of health system coding leaders formed, as CodaMetrix's Jamie Noorlander explained on stage, because "there's no shared definition of quality other than the 95 percent standard." Leaders from Mayo Clinic, Cedars-Sinai, Allegheny Health Network, and OHSU shared what resistance actually looked like, how they brought skeptics in, and how they used data to build trust. Here are the eight fundamentals they landed on.
1. Clean up your house before you invite AI over.
Map your workflows and know what's truly coding work. Don't retrofit a new tool to an old mess.
"You can't allow AI to step into a process that you yourself don't understand," said Monica Watson, Corporate Coding Director at Allegheny Health Network. Document your workflows, understand every connection, and build a data dictionary: "What something means to you as a system may not have the same meaning to your vendor."
That includes IT. "Your partnership with your IT team has to come before your partnership with your vendor," Monica said. "Otherwise you won't have a successful outcome."
Even thorough preparation surfaces things. Catherine Finnie, Senior Director of Coding at Mayo Clinic, mapped workflows across multiple sites ahead of go-live and still uncovered a radiology coding team her group hadn't been connected to. Her advice is to budget for discovery: "Think about it like a house remodel. They tell you six months. Plan for 12 to 18."
2. Elevate, don't replace.
Redirect coder expertise to the complex work where human judgment matters most, and make that role clear to your team.
"I've implemented a few autonomous coding solutions," said Tammy Bickle, Senior Director of Revenue Cycle at OHSU. "I've not lost one coder."
The goal is a career path. "Instead of just being production coders on low-hanging fruit, we want them to have a succession plan and a path forward," Monica said. Catherine borrowed a phrase from an earlier session: the aim isn't human in the loop. It's "human in the lead."
3. Data before drama.
Baseline your coding quality, denials, and lag now. Then let the data tell you what's real.
Colleen Stalvey, Director of Coding Services at Cedars-Sinai, hears "we have all of these denials" and asks the same question every time: how many is all? "It's show me before you tell me anything else."
Catherine added a caution: "There's an assumption that if there's an error, it wasn't there before. But the data will actually tell us that. It might have been something we weren't capturing."
4. Connect leadership goals to coding-team reality.
Translate the CFO's and CIO's goals into a clear "why" for your coding team.
Colleen uses a visual with her team: "A packed house full of elephants hanging out of the windows. For everything we give up, there's a line of elephants on the front porch. There will always be work. It's just going to look different."
Tammy makes the why personal: "You can take a vacation and not come back to a mess in your work queues. You've earned that."
5. Start where AI is strong.
Begin with high-volume, repeatable work, not your hardest niche cases as a test.
"We're using AI to do that production-based work that's simple without thought," Monica said, "and then enhancing our coder skill set" for the complex cases and payer challenges that need it. A good partner should be upfront about where AI delivers fastest and where it needs more time.
6. Treat the tool like a new coder.
Validate, refine, and improve performance over time, and keep your experts in the lead.
Catherine sees the risk on both sides. Coders who grow too comfortable with any tool start accepting auto-suggested codes without question. "We need to be cautious of that and pull our coders and leaders into the conversation," she said, "because we want that skill and expertise building out something that is going to evolve and improve in quality." The tool learns from the team. The team has to stay engaged for it to learn well.
7. Communicate through trusted voices.
Pair enterprise announcements with direct-manager conversations so teams know what is changing and why.
"I can talk about what these tools are. I can talk about what it's going to mean to them," Catherine said. "But the trust they have is with their direct leaders." Senior leaders set the stage; frontline managers carry the message. "When you're implementing these tools, you have to be implementing change management at the same time."
8. Find your coding champions early.
Empower skeptics, trusted voices, and influential team members to be part of your core, with the right guardrails.
Colleen's team keeps a standing, open invitation for coders to raise what the system got wrong and what's frustrating them. "Creating that safe place is helping us know where our gaps are," she said, "so we can turn our attention to those places and remove those hurdles."
Monica added the guardrail: bring coders and auditors to the table, "but do it responsibly, so we don't go off the road and lose sight of what our goal is." And for the holdouts, Catherine's question is simply: "How do we recruit them into the process?"
These eight fundamentals are a place to start. Change management isn't a phase of an AI implementation. It is the implementation.
Want the full playbook? Download Going Beyond 95% Accuracy: Rethinking Quality in Medical Coding
