🏥 Sloppy AI

Your takes on vetting AI comms solutions, Lumexa's IPO, and more

PARTNERED WITH Hey Hospitalogists,   Special shoutout to everyone who's gone the extra mile to answer questions asked by your peers in the Hospitalogy community. We've received such great feedback - positive, helpful, and full of great insights. Please keep asking your questions and answering them, so we can keep sharing them.   Let's get into it. Was this email forwarded to you? Subscribe Here SPONSORED BY WELLSKY Want to maximize your organization's capabilities in value-based care? The latest case study from WellSky is your blueprint for how top-performing ACOs and physician practices are doing it.  Successful risk-bearing entities don't rely on theories. They rewrite the playbook, using advanced technology, tighter collaboration, and dynamic provider networks to enhance both patient outcomes and financial success.  This leads to results that help them excel in the modern healthcare environment: seamless care transitions, lower readmission rates, better patient care, and improved cost management.  If you're running an ACO or VBC program, see the framework that helped drive a 15% reduction in SNF days, a 14% increase in patient engagement, and $8M in annual savings. Read the Case Study ASK HOSPITALOGY Vetting AI-Enabled Patient Communication and Call Center Vendors "We've been testing AI-enabled patient communication and contact center vendors, and we're coming across a lot of sloppy products. Has anyone vetted the space and can speak to what to look out for? Any general feedback about AI vendor vetting?" THE PRODUCT VETTING TAKE Anonymous "Certainly convos around cost and case studies will be important, but other things to look for or ask about: HIPAA compliance and clear BAAs (Business Associate Agreements) are non-negotiable. You also want to see evidence of encrypted data at rest and in transit, access control, and audit logging practices. Red flag: Vendors who "can be HIPAA compliant if needed" or dodge questions about data handling. Where will your data be stored and how long will conversation/interaction data be retained? Having any of your patient data used to train their models should be a dealbreaker. Red flag: They fine-print optional data usage for "quality improvement." What is their healthcare regulatory and compliance awareness? What guardrails do they have in place around prompting, clinical advice? What are their escalation protocols, such as routing to a human agent when medical risk is detected, and their safety classifiers for suicidal ideation, emergencies, or abuse disclosures? How do they mitigate "hallucinations" and what visibility do they have into logs and decision traces? You want to know their ability to surface why the AI took certain actions. Fit to healthcare workflows, such as integration with operations (e.g., your EHR, CRM, call center/telephony systems, ticketing systems, and patient engagement platforms) and fit to your use cases (e.g., scheduling, insurance verification, pre-visit prep, post-visit followup, no-show recovery, medication refill workflows)? What channels are supported (phone, SMS, patient portal messaging, email, web chat)? Red flag: Their solution sounds generic and not healthcare-specific. Fit to IT and operational requirements, such as security architecture (cloud provider, VPC isolation, penetration testing frequency, SOC 2 Type II certification), access and identity controls (role-based access, MFA requirements, audit trails), and operational reliability (performance metrics, uptime guarantees, first-call resolution rates, real-time AI to agent handoffs)?" THE COMPLIANCE TAKE Shawntea Gordon - CEO  "AI-enabled communications can be very helpful to triage patient calls, to support scheduling, and other low acuity, low risk tasks. The most critical component is a really well defined and intentional framework around safety, security, and compliance." THE EXPERIENCED TAKE Patrick Allen - VP of Business Development  "The bottom line is that most of the off the shelf AI products around communication are going to be kinda sloppy at this point. This is why you hear about many of the larger systems building their own LLMs to support this effort.   That being said, I suggest sticking with the established players (i.e., Nuance, Abridge, Epic, Oracle, Copilot Chat, Claude, etc.) or whatever vendor(s) your organization has already built relationships with.   In terms of things to consider:   1. Ensure you have a segregated instance. We're talking about PII, PHI, etc. and you must maintain compliance.   2. Be specific on what you want the tool to do. This may sound obvious, but literally everyone skips this step and wants to go straight into dev and LLM training work. Knowing what, exactly, you want the model to do, what you want it to replace, how you want it to behave, etc. are important decisions before you begin any of the technical work. Don't underestimate this step; skipping it is the biggest reason why AI implementation fails in healthcare.   3. There will be some of your patients who don't want to use AI. Make sure you accommodate patients who do and don't want these tools.   4. Look outside HC for use cases. A great example is the Podium app. It's an agentic AI employee that can coordinate communication from multiple sources and take action. I've deployed it in the Medi-Spa space, and I'm working with a car dealer to incorporate it into their website. In both of these examples, my client is using this tool to help schedule and confirm appointments. But see #2; this AI employee requires training and performance management.   Good luck and happy to talk further."   Hospitalogy members can join this discussion here. Not a member yet? Apply to join here. Lumexa's IPO "After months of rumors, looks like Lumexa's finally trying to go public.   Quick stats: Looking to raise up to $500M One of the largest imaging/radiology providers in the US High debt load, but profitable on an adjusted EBITDA basis Backed by WCAS Anyone think this is gonna work?" MY TAKE Blake Madden Just days after this question was posted, Lumexa Imaging officially announced the launch of its IPO. Looks like the diagnostic imaging player will look to raise up to $500M to get rid of their debt load, at a valuation just south of $2B. Some key facts and figures:   Lumexa Imaging operates an outpatient diagnostic imaging platform. As of Sept 30, 2025 they claim to run 184 centers across 13 states — making them reportedly the second-largest outpatient imaging center footprint in the U.S. by freestanding location count.   If we assume the total post-IPO shares (existing + new) run somewhere between 100–140 million (a typical range for a mid-sized roll-up IPO in healthcare) — then: At $17 → $1.7B–$2.4B enterprise equity value At $20 → $2.0B–$2.8B equity value The IPO being registered is for 25,000,000 shares of common stock.   Anticipated price range per share is $17.00 to $20.00.   If fully exercised at the high end, that's 28.75M shares total. So: At $17/share → gross raise ≈ $425M At $20/share → gross raise ≈ $575M Debt load is the biggest lever here IMO. If its IPO is successful, Lumexa will materially lower its debt on the balance sheet moving from ~5.5x+ debt to EBITDA down to ~3.5x.   Here's the amended S-1. Hospitalogy members can join this discussion here. Not a member yet? Apply to join here. SPONSORED BY ABRIDGE Ambient AI has moved well beyond the hype, and Abridge has a new report that shows one of the reasons why: health systems are seeing results. A Guide for Evaluating ROI Impact of Ambient AI at Enterprise Health Systems is a playbook for how health systems are measuring real operational and financial lift from ambient AI, from wRVU gains to cleaner claims and reduced after-hours work.  If you're trying to separate signal from noise in the AI boom, this is required reading. Download the Report ON YOUR RADAR Read: This white paper suggests that combining billing compliance departments and revenue cycle teams under a unified "revenue integrity" umbrella will help healthcare orgs better protect cash flow and avoid costly write‑backs. Read the full strategy here. Breakdown: Behavioral health is a blind spot that can wreck your population health strategy, quality scores, and VBC performance. Read my breakdown to see who can help you solve that.*  Resource: PitchBook says the next few years won't be business as usual for health systems and health‑tech investors. Their new 2026 Outlook reveals the deal trends, funding vectors, and structural shifts that could reshape who wins vs. who struggles in healthcare. Download the report. Roundtable: Join us for the January Roundtable for Plus Members, Friday January 23 at 1pm EST. *This read is brought to you by one of my brand partners who help make this newsletter possible! RECENT HEADLINES  HHS released its AI strategy, the next phase of the department's initiative to make AI available to the federal workforce and integrate it across internal operations, research, and public health. Mark Cuban in talks with Humana. During the recent 2025 Forbes Healthcare Summit, Cuban said that Cost Plus Drugs and Humana are exploring a potential partnership to help lower the cost of prescription drugs for employers. Read the full story. Mercy and Wellvana just inked a 20‑year deal to bring independent PCPs and APPs into a shared clinically integrated network. The plan? Give community providers who aren't employed by Mercy the tools, from EHR integration to care‑coordination support, to succeed in value-based care. Read more. That's all for today - I hope you enjoyed! Let me know your thoughts.   — Blake   Share Hospitalogy, Earn Rewards Have friends who'd love Hospitalogy too? Click the link below to share Hospitalogy with your friends and earn awesome rewards! PS: You have referred 0 people so far Share Hospitalogy! 🏥 Hospitalogy Membership I'm building a community of leaders in strategy, finance, and ops at hospitals and health systems to help us connect, learn, and grow together. Apply for free membership here. Get your brand in front of 50,000+ executives and healthcare decision-makers. Let's Partner → Workweek Media Inc.  1023 Springdale Road, STE 9E Austin, TX 78721   Want to ruin my day? Unsubscribe.