A conversation with Dr. Alice Tang, clinician-educator and faculty development leader at Weill Cornell Medicine, on why coaching is hard to teach — and what their pilot of SIMCoach, an AI-powered simulation that gives faculty a realistic space to practice their own coaching conversations, revealed about how to change that.
Disclosure: Dr. Tang serves as a consultant with Sketchy and was involved in the collaborative development of SIMCoach described in this interview. Dr. Tang receives consulting fees from Sketchy Group LLC, the developer of the DDx platform discussed in this article.
Dr. Alice Tang is a clinician-educator and faculty development leader at Weill Cornell Medicine, where she has spent the past decade studying the culture of feedback in medical training. She is the founder of the hospital medicine Master Coach program at Weill Cornell, which has trained over 40 faculty in coaching skills over the past five years. Her work focuses on building the conditions for honest, trust-based developmental relationships between clinical supervisors and their learners.
Dr. Ben Muller is a member of the team at Sketchy, where he works on DDx by Sketchy — an AI-powered clinical readiness platform designed to help medical educators teach and assess clinical reasoning and communication skills. He has been closely involved in the development of SIMCoach from its earliest stages, working alongside Dr. Tang and the Weill Cornell team to bring the project from concept to pilot.
Some context
Dr. Muller: Hi, Alice. It's great to have you and to be talking about the SIMCoach product that we've been working on together. We talk about this a lot, and so I'm glad to be talking about it with you in this setting. So I just want to jump right into it and kind of get your background and some context on how you got into coaching and how you've been working on it over the last few years.
Dr. Tang: Thanks so much for having me. I've been involved in faculty development in the realms of feedback and coaching for about a decade now at Weill Cornell Medicine. I've always been very interested in the culture of feedback within medicine. It’s something we know is critically important for learning and growth, yet so many people avoid these conversations or genuinely dread them.
What got me into coaching specifically was noticing that what tends to stick in people's minds are the times feedback didn't land well – the awkwardness, the fallout from uncomfortable situations. Coaching actually reframes that relationship between the person who's giving feedback and receiving feedback. There's a lot of intentionality that's put into establishing a safe relationship and trust and aligning both people towards a shared goal. Rather than being the bearer of bad news, you're engaging in honest reflection together, talking about vulnerabilities, and serving as a support figure helping someone brainstorm ways to work on what they're not yet confident in.
That shift in framing matters enormously in clinical training environments, where learners typically see supervisors as the person there to judge them, grade them, and determine if they'll get honor on a rotation. Being deliberate about changing that relationship is really important.
Over the past five years, we developed a coach training program within hospital medicine — where I practice — that trained over 40 faculty using workshops, role play, journal clubs with case discussion, standardized learners in the sim lab, and peer observation in authentic coaching contexts. We had really high engagement.
Logistically it's actually very difficult to do what we do in hospital medicine and other departments with different clinical schedules and less schedule flexibility, so we are really kind of struggling with how do you take what we learned and did within one context and scale it. So recently in the age of generative AI, there's just so much that's being done to transform the way that we can do education and do clinical care. In faculty development, it's the same. It was exciting to think about how we can use technology to overcome some of the logistical barriers we are facing and try to scale faculty development.
Dr. Muller: Genuinely difficult conversations. And then you also mentioned some goals that you have as sort of the coach of the coaches, but I'm curious sort of where you see those skill gaps or developmental gaps in your coaches. What are the things that they really need to spend their time working on?
Dr. Tang: We certainly did a lot of programming to help our coaches get better in this skill set, but the thing that we need the most to develop any skill is practice and feedback. If you think about coaches, faculty in general, those things are surprisingly difficult.
There's only so much role play that you can do. People don't like it, so you can't really practice that much. People end up trying to apply this coaching framework and practice in the context of real life learners which is a high stakes situation where you don't want to mess up with the learners, so there's not a lot of opportunity to practice.
Then in terms of feedback it's quite limited as well because of fear of retribution and the hierarchy within the clinical training environment.
When you ask a learner, “how can I improve?”, “am I being an effective coach?” it's mostly met with “no, you're great”, “everything's wonderful.” And I think it's really hard to get that feedback to actually improve.
The biggest gap is how do you practice a skill and how do you actually get meaningful feedback to get better in the skill?
Dr. Muller: That makes total sense. So you alluded to the opportunities afforded by generative AI and you've talked about some of the need for safe practice and honest feedback. So put all that together and tell us sort of what SIMCoach actually is.
Dr. Tang: Yeah, so SIMCoach is essentially a set of coaching cases that were developed by medical educators, peer reviewed, and then put onto this DDx platform where we have an AI essentially simulate the learner that is written in the case in a context that a faculty member would be coaching them. So at the end of a rotation or at the end of a clinic session, that dialogue that you as an educator would have with a learner.
So faculty enter the platform and they get some background information about the context in which they're interacting with the learner, whether it's clinic or inpatient setting and which rotation. Then they get some background information about the learner, like what goals they had shared previously that they are working on and what your observations had been about that learner over the course of whatever time period you were interacting with them, both strengths and opportunities for growth.
When the faculty member is ready, they essentially proceed to the next screen and then begin the coaching conversation with the simulated learner through typing. And so essentially interacting with this chat bot, which is the simulated learner, and there's a back and forth dialogue where we're hoping that the faculty member is applying this framework for coaching that they learned. And then after the coaching scenario, they can actually advance to the next page.
They get feedback on how that conversation went and that's through an AI that is trained on a coaching rubric that we had developed based on the coaching frameworks that are in the literature. It’s actually very neat that the feedback that they get is so specific in terms of language that they used that was very effective and also suggesting language that they might want to try next time to be able to do better.
How the pilot came together
Dr. Muller: You mentioned the frameworks and the rubrics and the feedback. Tell us a little bit more about the frameworks that you're using here to evaluate your coaches.
Dr. Tang: There is a framework called Prepare to Adapt that was published by Tyra Fainstad and colleagues. We essentially took that feedback framework and deliberately applied coaching concepts to each step. The reason we picked Prepare to Adapt was because we really liked how it specifically designated each phase of the relationship that's necessary to set things up for a successful conversation about feedback.
So in the Prepare step, there's a lot of intentionality in creating that trusting relationship between the feedback giver and receiver. They're able to elicit the goals and really reflect upon what is meaningful and what they should target as the objective for the coaching. Then the coach is able to say, "I'm in it with you" and "I'm going to commit to supporting you in these goals."
Then during the Perform step, really being able to be intentional as a coach to gather high quality observations about the learner.
The Adapt part is essentially what we think of as the feedback conversation, using coaching skills like asking questions, stimulating reflection to be able to identify what are the strengths and what are the areas for improvement that the learner might want to target, set new goals and action plans to get them closer to those goals.
We ended up creating something that was new, but in looking at some of the published frameworks for clinical coaching, like R2C2 or RxOCR, they're variations on the same type of concept, but we liked how discrete it was with the Prepare to Adapt framework, so we call it Coach to Adapt.
Dr. Muller: This isn't just theoretical, right? You've run this now with multiple sort of sessions of faculty at Weill Cornell, so tell us a little bit about those who came to those workshops, who participated, what was their coaching background and how'd they go overall?
Dr. Tang: Yeah, it was so exciting to be able to have colleagues try this from all different areas of the institution. We've run the coaching pilot three times now at Weill Cornell Medicine. The first two times were within hospital medicine, where a large percentage of those who participated had been involved in my coaching development program, the Master Coach program, so they brought that perspective of being able to compare different modalities.
The most recent pilot was actually institution wide and we had PhDs who were non-clinical and faculty from the veterinary school join. It was cool to see how people from different disciplines were able to apply lessons from this coaching framework in their own context.
The workshop was essentially a one hour session where we introduced some core coaching concepts, introduced them to this Coach to Adapt framework, and then spent a large portion of the time actually having them in the platform, doing a SIMCoach case, and being able to practice the Coach to Adapt framework within that coaching case.
Early insights
Dr. Muller: What did you find interesting from either the case results or from faculty feedback on the case experience?
Dr. Tang: It was interesting, especially with faculty who had experience with other modalities for coaching faculty development, that a lot of faculty actually found it to feel less artificial coaching this AI-powered standardized learner compared to the usual modalities we use, which is role play. I guess I shouldn't be surprised, because whenever I say we're going to do role play to practice skills, there's an audible groan and nobody really wants to do it. People tend to have a love-hate relationship with role play — they know they need to practice the skills, but the thought of doing it in this artificial environment with colleagues watching is really a barrier to engagement. So it was surprising that people actually felt it was less artificial and felt true to life, being able to engage with a learner in this platform.
The other thing that was surprising to me was the typing feature. I thought it might be a barrier, but some people really liked it because of the intentionality involved in slowing down and thinking about what you're going to say next. That was also surprising to me, but it makes sense when you're learning a new framework.
Dr. Muller: I'm curious in thinking about the platform, you've talked about a number of strengths already. What are some of those opportunities for improvement or things that didn't work quite the way you'd anticipated?
Dr. Tang: When we were rolling out the first case in this pilot, we worked with a variety of faculty with different coaching experience. We aimed to create an entry-level case where it wasn't super difficult to coach, but we heard from a lot of faculty that our coachee Pat was too insightful. That could be a barrier depending on your level of training — they might not find it challenging enough. But we're actually in the process of taking that case and iterating on it. We spent some time thinking about what makes coaching difficult, what makes it complex, and we're essentially taking each case and varying different parts to make it more difficult. So in version 2.0, having the learner not buy into the feedback or be defensive, and then in version 3.0, perhaps having the learner not know how to create goals or have difficulty coming up with them. Taking the same case and iterating on it. And I think it's actually really exciting thinking about what kind of customizable experience faculty can have down the line — perhaps being able to sort by the type of coachee, like one who doesn't know how to set goals or one who doesn't buy into your feedback.
Dr. Muller: How do you see SIMCoach fitting into a broader coaching development curriculum for faculty?
Dr. Tang: I think this is really a wonderful tool that can fill a gap we currently have with practice and feedback. I'm not saying it's meant to replace all other modalities — one thing that people really love about faculty development is engaging in a larger community of educators, having the chance to debrief challenges, learn from one another, observe one another. That's just such a rich learning environment. So ideally this is going to be used as an adjunct, one part within a larger faculty development program. With my prior program, I found that every modality really has its strengths and drawbacks, and a combination of different modalities is probably where we should land.
In particular, this could be great practice before real life coaching. It feels uncomfortable to have a colleague observe you coaching a learner when you're just going in cold. Practicing these skills in a low stakes environment — where nobody can see what you say, where there's no collateral damage that can happen — might be a way to make people feel more comfortable with that real life peer observation of teaching that is really valuable.
Looking ahead
Dr. Muller: You mentioned some aspects of the product that you're continuing to work on. You mentioned maybe making a more difficult coaching case than the first one that you've introduced. What are some other things that you're still thinking about as you think about where we might go from here?
Dr. Tang: One thing that keeps coming up in our program evaluation is that coaching is a communication skill at the heart of it. We heard that being able to hear how phrases are landing or see the reaction of the learner is kind of hard to gauge from a typed encounter, things that are picked up through facial expressions, body language.
Even the cadence of their voice is really important to know how you should respond to the learner, so I’m thinking about ways that we can make it more high fidelity. Your team is already actively thinking about features like the voice function of having it be an actual conversation where you can better gauge if the learner is totally disengaged versus coming along with what you're trying to get them to engage in.
Dr. Muller: Yes, absolutely. We're trying to make the most realistic path that we can. So let me just close by foreshadowing sort of our next encounter, which is our upcoming webinar for other educators who are interested in coaching and exploring SIMCoach. So do you want to give a quick preview of that session?
Dr. Tang: The webinar will essentially be a version of what we've done for the pilot — we'll introduce the coaching framework, giving some foundational information on Coach to Adapt, and then log into the SIMCoach platform and have a chance to actually coach Pat, applying the framework in that context. Then we'll spend real time debriefing amongst this broader community, because what we're really trying to identify is who's going to benefit the most from something like SIMCoach, how do we get it to those people, and how do we implement it in a way that gets people genuinely engaged and able to use this to improve their skills.
For people joining the webinar, I hope there'll be a twofold impact — as an educator, thinking about how you might take these skills and apply them with your own learners, but also from a broader view, thinking about how something like SIMCoach should be used to improve clinical coaching and the learning experience for trainees across different contexts. I'm really excited to hear perspectives from that broader audience.
Dr. Muller: Yeah, very much looking forward to hearing from them and to that session in general. And so for now, I'll just say thanks so much for joining us today and for working with us on this project. I think SIMCoach has been an amazing experience both in sort of the product development side, but also in seeing how faculty have interacted with it and enjoyed it and getting their feedback has been a really great experience. So thanks so much.
Dr. Tang: Thank you. I really appreciate the opportunity and look forward to where this goes from here.
