Last Updated: 31 July 2026
Most people choosing a cosmetic procedure decide alone, late at night, long after the clinic has gone dark. They scroll before-and-after photos. They compare prices across three sites. They type a nervous question into a search bar, about pain, or scarring, or how long recovery really takes. Nobody answers. By morning the front desk opens to a full voicemail box and a stack of web forms, and half those people have already booked somewhere else. The lead wasn't lost to a better surgeon. It was lost to silence.
Benjamin Stover spent more than a decade in healthcare technology before that silence became his problem to solve. In aesthetic and elective practices he kept seeing the same breakdown. The marketing worked. Patients showed up online, curious and ready. Then the practice, short-staffed and pulled in ten directions, couldn't respond fast enough to turn that curiosity into a booking. The tools meant to help, generic chatbots bolted onto websites, often made it worse by giving vague or flatly wrong answers to medical questions.
That gap became AI Med Consult, where Stover is Chief Commercial Officer. The company built an AI patient engagement platform for one industry only, aesthetic and elective healthcare, instead of a general tool stretched to fit. It answers procedure questions instantly, qualifies interest, and guides people toward a consultation around the clock, while lifting repetitive admin work off the staff. The bet behind it is blunt. An AI trained on the whole internet has no business advising a patient about their own body, but one trained for this single field does. The results are loud. Some practices report returns above 1,000 percent, and in one case above 4,000.
We spoke with Benjamin because patient engagement is a problem every practice owner recognizes and few have truly fixed. Automating patient questions is easy. Doing it without giving someone bad medical information, right when they're deciding to go under the knife, is not. Stover walks through how the system decides what it will and won't answer, makes his case for a specialist AI over a general one, and explains the one line he won't let a machine cross.
1. From a patient's first question to a booked consultation, what actually happens inside AI Med Consult? Which parts run on a language model, which follow fixed rules, and which still need a person?
When a client interacts with one of our AI products, our proprietary AI is doing all the work communicating with the client, answering questions about the doctor, procedures, and scheduling. After the client schedules an appointment there are two scenarios that can occur. 1) the client has our integration with their EMR activated and the scheduled appointment automatically appears on the doctor’s calendar; no work is required by the office staff. 2) the requested appointment shows in our Lead Management Back Office System and the patient coordinator will contact the client to confirm the appointment.
2. AI MedChat is built as a medically trained large language model rather than an ordinary chatbot. In plain terms, what does "medically trained" mean here? Is the model fine-tuned, does it pull answers from approved medical sources, are the prompts carefully designed, or is it some mix of all three?
It means the model was built on medical textbooks, a specialized cosmetic and plastic surgery procedure library, and physician-led training.
Surgeons tested the platform extensively and helped refine not only the accuracy of its answers, but also how those answers are communicated. The goal was to translate complex medical information into clear, patient-friendly language without losing credibility or accuracy.
3. The platform launched in 2025 after two years of development with a physician-led team. AI Med Consult frames artificial intelligence as an exponential shift for aesthetics rather than a gradual one. What did those two years reveal about what general AI gets wrong in a clinical setting, and why call the change exponential rather than steady?
The biggest issue is the difference between a generalist and a specialist.
General AI models are designed to cover many subjects, which makes them more likely to give inaccurate or overly broad answers in highly specialized areas. AI MedConsult was built specifically for aesthetic medicine, so it can connect a patient’s goals with appropriate procedures while also reflecting the services, pricing, and preferences of an individual practice.
That is why we see the shift as exponential. It moves beyond collecting contact information and can conduct a meaningful consultation that educates and qualifies the patient.
4. Most patients research providers outside office hours, long after a clinic has closed. Beyond simply being awake at 2 a.m., what does the AI do in those late-night conversations that a basic booking form cannot?
It can engage the patient while their interest is highest.
A booking form simply collects information and asks the patient to wait. The AI can answer questions, address concerns, explain procedures, and help patients understand their options immediately.
As a result, the patient is more informed, confident, and prepared by the time they book a consultation.
5. Between a patient showing real interest online and a practice actually booking them, plenty falls through. From what you have seen, where does that process break down most often, and why have generic chatbots failed to fix it?
It usually breaks down when a patient has specific questions and receives only generic or delayed responses.
Traditional chatbots often lack the medical knowledge and practice-specific information needed to build trust. Our platform is designed to support the entire patient journey, including website conversations, after-hours phone calls, follow-up emails and texts, scheduling, and post-operative communication.
6. When the platform personalizes a conversation, what is it actually responding to? The patient's questions, their procedure interests, past chats, the practice's services, or something more? And where do you choose to hold personalization back?
It considers the patient’s questions, goals, procedure interests, previous conversations, and the practice’s services, pricing, and preferences.
The AI then focuses on the information most relevant to that patient rather than overwhelming them with every possible detail. It may also offer helpful next steps, such as viewing before-and-after photos, learning about recovery, or scheduling a consultation.
7. Which questions is AI MedConsult allowed to answer outright, and which does it have to soften, refuse, or pass to a clinician? Can you give one example of a question it is built specifically not to answer?
The AI does not provide diagnoses, determine whether someone is medically eligible for surgery, or give definitive answers that require clinical judgment.
Instead, it provides general education, uses appropriate ranges when discussing topics like recovery, and directs personalized medical questions to the surgeon.
For example, it would not tell a patient exactly when they can return to exercise. It would explain the typical range and clarify that their surgeon must provide an individualized recommendation.
8. The platform is clear that it does not give medical advice or a diagnosis. How is that line enforced in practice when a prospective patient asks a sharp clinical question after hours, with no staff watching?
When a question requires clinical judgment, the AI acknowledges it, provides limited general context, and refers the patient to the physician.
The same principle applies after surgery. AI Post-Op Nurse tracks patient-reported symptoms and escalates concerning responses based on thresholds established by the practice. Severe pain, dizziness, or other warning signs can trigger an immediate handoff, along with instructions to call 911 when an emergency may be present.
9. One module is named AI Post-Op Nurse, though its publicly described functions are onboarding, scheduling, consult prep, and reengagement. Does it also track recovery or read patient-reported symptoms? If it does, what kind of message triggers a handoff to a real clinician, and who checked that those rules are sound?
AI Post-Op Nurse is designed specifically to support patients after surgery by tracking recovery and monitoring patient-reported symptoms.
Patients can report how they are feeling through pain scales, visual indicators, and open-ended messages. Each practice sets its own escalation thresholds, with configurable defaults provided by the platform.
For example, a patient reporting severe pain, dizziness, difficulty standing, or an inability to eat may trigger an immediate handoff to clinical staff. The system also reminds patients to call 911 if they believe they are experiencing a medical emergency.
If the patient reports normal discomfort and steady improvement, the AI continues the conversation and shares the update with the practice. This helps staff monitor recovery while ensuring patients feel supported throughout the post-operative process.
10. Voice tools can mishear accents, background noise, procedure names, and symptom descriptions. How does AI Med Voice confirm what it heard before it shares information or acts on it?
I think the better AIs are better at this. Right. So they can pick up on more things. But the typical response, if it doesn't understand you, it's going to ask you to repeat yourself in a nice way. It sounds simple, but that's kind of how that works. It'll say things like, I think you said you're interested in mommy makeover. Is that correct? And then, you know, they'll go on from there.
Yes, but it's annoying if it does that a lot. Like, you know, so that is also the balance. Like if you said it and then it keeps asking you to like, active listening is basically what it does when it's unsure. But even in real life, active listening is such a great tool, but if you overuse it will annoy the living out of people. Right. So it's, there's a balance to that.
11. The platform sorts some people into a category it calls a "qualified patient opportunity." What is the AI judging to make that call? And how does the company make sure that judgment does not quietly work against people because of their language, the way they communicate, a disability, their background, or how much they appear able to spend?
A qualified patient opportunity is based on engagement and intent, not language, disability, background, or other personal characteristics.
Someone who discusses a procedure, asks questions, and provides contact information is considered qualified. Patients who also request photos, upload images, review financing, or begin scheduling may be considered further along in the process.
Pricing is presented early, and practices that offer financing can provide that option directly through the platform.
12. AI Prospector keeps re-engaging interested patients on its own through personalized email and text. How does it decide the timing, tone, and frequency so it never tips into pressuring someone toward an elective procedure? And how does a patient make it stop?
AI Prospector uses information from the patient’s original conversation to send relevant follow-ups instead of generic sales messages.
The timing and frequency are set by the individual practice, while the tone remains brief and consultative. Patients can stop the communication at any time by replying with an opt-out request or clicking the unsubscribe link.
13. Is this one central AI, or several specialized parts handling chat, voice, scheduling, follow-up, and post-op separately? If it is the latter, how do those parts share what they know without passing an early mistake down the whole patient journey?
It's a little bit of both. Everything is fed from a central AI, but each customer operates within their own self-contained environment. We push updates to all customers from that central intelligence, while each practice has individualized components tailored to their workflows. Our products,including AI voice and AI text, all utilize the same primary AI database, so every advancement benefits the entire platform while remaining specific to each customer's implementation.
14. The platform now takes actions, not just answers questions. Which of those actions can the AI carry out by itself, and which still need a staff member to approve them first?
The AI can capture patient information and write it directly into the EMR, eliminating manual data entry. It can collect demographics, procedure interests, scheduling information, and other patient details before the practice ever gets involved. When an action requires clinical judgment or staff approval, such as decisions involving patient care, those remain with the practice. Our goal is to automate administrative workflows while leaving medical decision-making to the provider.
15. AI Med Consult argues that basic chatbots give wrong medical answers between 50 and 80 percent of the time. That is a big number. What testing produced it, and how does the company measure its own system against the same standard?
We aren't making that claim ourselves, we're citing NIH research on large generative AI models. The difference is how we've built our system. We use a large language model with a closed-loop database, meaning the AI can only use the knowledge we've provided. That knowledge base has been physician-trained over more than a year and continually refined through feedback from early customers. We also have very strict guardrails in place. If the AI encounters a question that requires medical expertise, it doesn't guess, it immediately prompts the patient to schedule a consultation with the practice.
16. AI Med Consult recently announced a partnership with 4D EMR to pass the data captured from patient engagement directly into the EMR. Once that information lands in the clinical record, what new things can a practice actually do with it?
The 4D EMR announcement is one example, and we're actively building partnerships with EMR providers across the industry. The integration removes extra steps for practitioners by eliminating double entry. Patient demographics, scheduling information, consultation details, and even before images can flow directly into the EMR. The goal is for AI to fit seamlessly into the software practices already use, improving workflow efficiency while giving providers better access to patient information and ROI insights.
17. What happens when the AI's version of events clashes with the practice's calendar, CRM, medical record, or internal notes? Which system gets treated as the truth? *
It depends on the data point. We prioritize the source of the information. We always look at where the data is flowing to and from, and whichever system serves as the primary authority for that information becomes the source of truth.
18. The platform can collect contact details, procedure interests, images, and conversation data. Does any of that go to outside AI providers or get used to train future models? And what say does a patient have over how long it is kept or whether it is deleted?
We do use anonymized conversation data to continue improving the platform over time, but all patient information is stored on HIPAA-compliant servers and remains anonymous. If a patient requests deletion of their data, we support that process in accordance with applicable requirements.
19. The platform is described as HIPAA compliant. Secure hosting is one piece of that. What about the rest? Who can access the model, how are staff permissions set, what gets logged, how long is data held, and what happens when something goes wrong?
All patient data is encrypted both in transit and at rest. Staff permissions are determined by each individual practice, so they control who has access to their information. Our developer team is U.S.-based, and patient data is retained according to HIPAA requirements unless a patient requests deletion where applicable. If an issue occurs, we follow established HIPAA standards and compliance procedures.
20. Some platforms in this space only let the AI answer from approved, verified sources rather than guessing, which keeps it from making things up. Does AI Med Consult work the same way, and how does it keep answers tied to what a specific practice really offers?
Yes. Our AI is trained on a closed-loop data library that exists within each practice's private environment. Rather than pulling information from the open internet, it only answers using the approved knowledge we've provided for that specific practice. That ensures responses stay aligned with the services, policies, and information that practice actually offers.
21. Published case studies show returns above 1,000 percent, and in one case above 4,000 percent, with a separate figure of 2,000 percent cited elsewhere. How is that calculated, which costs are included, and how do you separate what the AI did from marketing spend, seasonal demand, and the staff's own work?
Our ROI calculations compare the total monthly cost of the platform against the revenue generated, using data directly from the EMR. We recognize that AI is one piece of the broader marketing ecosystem. We work alongside a practice's marketing strategy, EMR, and lead generation efforts while reducing the manual labor involved in patient engagement. The ROI we're seeing is tied to improvements in upstream lead conversion and operational efficiency, measured against actual patient and revenue data.
22. After talking with clinicians at events like The Aesthetic Meeting, which request or worry has changed the company's plans the most? And as someone who has built and scaled several companies, how do you tell the difference between an AI demo that looks impressive and a tool a practice can safely lean on every single day?
One of the biggest requests we've heard from practices is that they wanted text capabilities first, and then voice. Listening to customers has directly shaped how we've prioritized product development.
As for evaluating AI, it comes down to asking the right questions during the demo. We often demonstrate the platform using a prospective customer's actual website and test it with the kinds of questions patients really ask, including unusual or difficult ones. We also "secret shop" our own platform in real-world situations because that's how patients experience it. An impressive demo isn't enough. The technology has to perform consistently in real practice environments where patients rely on it every day.
