Serious AI with Chris Duff
October 7, 2026, updated October 12, 2026
In our read, mostly from money it already pays for work that software can now take on. We would look in four places first: billing and denial rework, the hours that go into prior authorization, front-office turnover, and the tools bought to patch around the EHR. We would fund a first AI project in a medical group without a new budget line. It needs one number named before anything is built, with the money moved from one of those four places and released in steps as that number moves.
The timing makes the case for that framing. The Medical Group Management Association (MGMA), which describes itself as the premier association for professionals who lead medical practices, found in its June 23, 2026 MGMA Stat poll that 84% of medical groups reported operating costs higher than at the same point in 2025, with an average increase of about 11% among those whose costs rose. That share is down from 95% in 2023, but MGMA’s earlier polls put the average increase between 10% and 12.5% in each of the three years before. A week later, only 47% reported revenue up year to date, against 56% a year earlier. Whoever reads a 2027 AI request in a practice is looking at costs up and revenue growth narrowing, and in that room a request lands best as money moved from somewhere it is already going.
This answer comes from Serious AI Partners, an enterprise AI implementation firm run by Chris Duff, M.D., who trained as a physician at the University of Illinois College of Medicine, spent 2017 through 2019 in healthcare venture capital and built the systems Serious AI runs on in a business where being wrong cost him money. The work below is the operating side of a practice: billing, authorizations, phones and scheduling. MGMA Stat polls cited here go out by text message each week to a panel of healthcare leaders who sign up for them, so read them as what those leaders report about their own operations.
Industry benchmarks often put billing and revenue cycle cost at around 5% of collections, as MGMA noted on June 11, 2025, whether your own staff does the work or an outsourced vendor does. In an MGMA poll on January 6, 2026, 48% of respondents named denials and appeals as the biggest revenue cycle leak. In Experian Health’s State of Claims survey of 250 healthcare professionals responsible for financial, billing or claims decisions, published September 22, 2025, 41% of providers reported denial rates of 10% or higher, an issue Experian says has grown each year since its first survey in 2022. Respondents said 90% of claim denials are reworked with at least some human review before resubmission, and only 14% were using AI, automation or machine learning. Experian Health sells claims software, so weigh it as a vendor survey.
What software can take: checking eligibility and claim fields before submission, sorting denials by reason code, and pulling the records an appeal needs. What stays with your staff: the appeal argument and the payer contract conversation. The number to watch is your first-pass claim acceptance rate, or days in accounts receivable.
The American Medical Association’s 2025 survey of 1,000 practicing physicians, released May 13, 2026, found that physicians complete an average of 40 prior authorizations a week, taking about 13 hours of physician and staff time, and that 40% of physicians employ staff who work only on prior authorization. In an MGMA poll on March 31, 2026, 61% of practices said their staff access seven or more payer portals a week, and navigating those portals is the part that belongs in software. We broke down which steps move and which stay with a person in Can AI Handle Prior Authorization for a Medical Practice? The number to watch is staff hours per authorization, or days from order to approval.
In an MGMA poll on May 26, 2026, 28% of medical groups reported higher staff turnover than in 2025, 39% about the same and 30% lower, with the churn concentrated in medical assistants, front-desk and patient access staff, and nursing and clinical support roles. Revenue cycle roles appeared in smaller pockets. The US Bureau of Labor Statistics puts the median wage for medical records specialists at $51,140 (May 2025). AAPC describes itself as the world’s largest training and credentialing organization for the business of healthcare. In its 2026 salary report, built on more than 20,000 responses from its members, medical records specialists (coders, billers and others in revenue cycle management) reported earning $65,007 on average in 2025.
When one of those roles opens, ask what share of the job is phones, scheduling and eligibility checks before posting the whole of it. Handling that share first funds a project with no layoffs. The number to watch is call abandonment rate, hold time, or days to the next available appointment.
In an MGMA poll on July 21, 2026, sponsored by the EHR vendor NextGen Healthcare, three in four groups said clinicians frequently (43%) or sometimes (32%) work around EHR or practice management problems, through duplicate entry and separate tracking files, and some said one attempted fix required buying another application. 83% of medical groups now use AI in some patient visits (MGMA, August 4, 2026), up from 71% in 2025, and AI scribe pricing runs from free to several hundred dollars per clinician per month, depending on how deeply the tool integrates with the EHR and on contract terms, with budget-friendly stand-alone apps that can start as low as $40 (Medical Economics, March 30, 2026).
Before adding a tool, list what each existing one replaced and whether anyone measured it. That second question matters: in an MGMA poll on August 5, 2025, 44% of practices using AI in visits said it had not reduced workload, against 39% who said it had. A tool nobody measured is the first candidate to cancel or replace.
Each project gets one number, with today’s value written down and one named person who owns it. The examples above work, such as first-pass acceptance, hours per authorization or hold time. If nobody can name the number, the project is not ready for the budget, however good the vendor demo was. MGMA’s May 12, 2026 write-up on AI and clinician productivity says uneven use, paired with limited tracking and feedback, makes it hard to establish a productivity baseline, let alone measure change. In our view, a subscription bought without that baseline can renew for a year with nobody able to point to a number that moved.
Fund one process for the first quarter. The second quarter unlocks when that process hits the target agreed before work started. The third unlocks when the result holds without someone tending it. A ramp is an easier yes for a physician board than a lump sum, and it means nobody keeps paying for something that stopped working in March.
Whether a given AI project is worth building at all is a separate question, and it comes first. The four tests we run before any build are in The four questions: is this worth building at all?
AI can produce the analysis. It cannot be wrong, because being wrong requires having something at stake, and that stays with a person. Serious means underwritten.
=====
If your practice is planning its 2027 budget and wants to know which of these four places holds the money for a first AI project, the engagement starts with diagnosing the constraint that is gating you, then building with verification sitting where a person still owns the outcome. You leave the first call knowing what your constraint is whether or not you ever hire us.
Fill out a short form, and we read every answer ourselves. If there is a fit we will reach out and find a time that works for you, and if there is not, we will say so and point you toward someone better suited. Either way, the moat question is worth answering now, in a market AI has already reset.