Communications Skills Training at Scale

Advertorial

Conversational AI is reshaping healthcare simulation by making communication skills practice repeatable, accessible, and easier to debrief. This article explores the problem it addresses, what “conversational AI” means in simulation-based education, and how different modes – turn-based, real-time, and phone-call – support different learners and contexts.

I’m a psychiatry doctor by training, so I’ve spent years watching how small shifts in language can change outcomes, especially when someone is frightened, angry, or overwhelmed. That’s why I’ve never been able to see “communication skills training” as a nice-to-have. It’s a safety tool.

It wasn’t in a lecture theatre. It was on a phone call. No room to breathe, no ‘time out’, no shared glance with a supervisor, no one else listening in to steady the moment. The silence felt louder than it ever does face-to-face, because on the phone silence isn’t neutral: it’s interpreted. And you can’t take five minutes. You have to choose your words in real time.

I was trying to support someone in crisis – distressed, angry, frightened, exhausted. The clinical facts were relatively straightforward. The conversation wasn’t. Every pause seemed to mean something different. Every interruption changed the temperature. The person on the other end wasn’t just listening to words; they were listening to uncertainty, impatience, reassurance, threat, care. When the call ended, the debrief wasn’t about what we knew. It was about how we sounded.

That experience (and many life-changing conversations since) left me with a nagging question: why do we accept that we can simulate physiology and procedures with sophistication, yet still struggle to simulate the everyday conversations that shape trust, escalation, complaints, safeguarding decisions, and risk?

Key Takeaways

The Real Constraint: Time, Access, and “Reps”

Most simulation educators don’t need convincing that communication matters. The problem is more practical: you can’t always get the people, rooms, timetables, and standardised patients you need.

And even when you can, learners often get one attempt – maybe two – then the day moves on.

But communication, like any complex skill, improves through repetition. “Reps” are how people build reliable habits: opening a call well, structuring a difficult conversation, checking understanding, making space for emotion without losing control of the task, and closing safely. The tragedy is that the learners who need the most practice often have the least access to it.

This is where conversational AI has started to feel less like a novelty and more like a missing piece of infrastructure.

Conversational AI in Healthcare Simulation

Conversational AI in simulation is an AI system that can hold a responsive spoken or written dialogue with a learner, so the scenario changes based on what the learner says, enabling repeatable practice without scheduling a human actor.

When I say conversational AI in medical education, I don’t mean a chatbot that fires back generic lines. I mean systems designed to hold a dynamic, responsive interaction where the learner’s phrasing and choices genuinely alter the path of the encounter.

In simulation language: it’s a way of generating a “responsive other” (patient, relative, colleague) that can remain consistent enough for training, yet variable enough to feel real. The promise isn’t perfection. The promise is availability, and the ability to practise the same scenario repeatedly with meaningful variation.

The evidence base in this space is maturing, including published work specifically evaluating SimFlow.ai in different contexts. In Education for Primary Care, “Enhancing GP consultation skills training…” they evaluated a conversational AI simulation delivered via the SimFlow.ai platform, reporting strong acceptability and perceived educational value (with clinical authenticity and educational value both median 4.5/5), alongside reported gains in accessibility/practice frequency and a cost comparison suggesting 24–84% cost reductions versus actor-based approaches.

A multisite JMIR Formative Research study, “Scaling Multimodal Agentic AI in Medical Education…”, similarly examined simulation effectiveness using SimFlow.ai, finding very high ratings for clinical plausibility/medical content (eg, medical content median 4.5/5) and positive educational value (median 4.0/5), even where conversational realism was more mixed, useful evidence that strong clinical logic can still translate into learning value at scale.

Finally, a practitioner-focused review in Current Treatment Options in Psychiatry highlights SimFlow.ai under AI for clinician training, noting hundreds of AI-powered patients simulating youth presentations and post-session feedback to support diagnostic and communication skill development, while also calling for continued evaluation of fidelity and real-world applicability. Although the examples in that section are youth mental health, SimFlow.ai itself supports communication simulation across wider healthcare domains and other sectors.

So Where Does SimFlow.ai Fit?

SimFlow.ai was built around a simple observation: we were asking people to learn high-stakes conversation skills in low-access conditions, and the typical workarounds (more workshops, more actors, more scheduling) weren’t scaling.

At its core, SimFlow.ai is a conversational simulation platform for communication skills training that lets learners practise with AI-powered characters, including patients, relatives, and colleagues aligned to specific learning objectives. It’s voice-based, and characters can be given voices that match age, gender, and nationality, because those details quietly shape realism and rapport.

But what matters educationally is not just that the conversation exists. It’s that the platform offers three distinct simulation modes, each designed for a different training moment and a different learner.

Three Modes for Different Learners

Here’s the simplest way to think about it:

Three simulation modes

SimVoice is deliberately more structured: a turn-based format that reduces cognitive load and increases psychological safety. It gives learners time to think, formulate, and reflect before responding, particularly helpful when educators want to focus on micro-skills (opening questions, signposting, summarising) without the pressure of real-time pace.

RealTime is the high-fidelity mode: low-latency conversations designed to feel like real interaction. This is the mode for practising the hard, messy middle of communication, where learners interrupt, get interrupted, pivot when new information lands, and manage emotion while holding structure. All while experiencing ambient background noise of a real environment.

And then there’s SimCall. SimCall exists because simulation sometimes underestimates a simple truth: the channel and conditions shape the skill. Many of the hardest conversations in healthcare happen on the phone: triage, escalation, relatives, out-of-hours, safeguarding, duty of candour, handover clarifications. Audio-only strips away visual cues and forces learners to manage tone, silence, interruption, pace, and uncertainty with fewer anchors.

So SimCall leans into a deliberately “low-tech” front end: use the actual device people use in real life. A phone call. Your habitual posture. Your normal pacing. Under the hood, the back end remains high tech: speech-to-text, conversational generation, text-to-speech, orchestration, and then structured outputs that can support reflection and educator debrief.

That design choice is not just convenience. It’s educational theory in disguise: match the training modality to the performance modality. If we train the wrong channel, we shouldn’t be surprised when transfer is limited.

Debrief and Assessment with AI Feedback

One reason I think conversational AI belongs in the sim toolkit is that it can help with a second, quieter constraint: faculty time. Educators don’t need more dashboards. They need better starting points for debrief: moments to return to, patterns to notice, language to examine.

The best use of AI feedback here isn’t as a judgement. It’s as a scaffold: highlighting structure, clarity, missed checks, moments of escalation, and opportunities to repair. SimFlow.ai’s outputs also include an emotional sentiment layer. I’m careful about this: sentiment is not “truth,” and it shouldn’t become surveillance. But as a reflective prompt – where tension rose, where reassurance landed, where empathy dropped under pressure – it can help learners notice patterns they genuinely can’t hear in themselves the first time.

Where I Think This is Going…

SIMZINE’s readership lives at the intersection of realism, scalability, and safety. Conversational AI sits right on that fault line. It won’t replace human faculty. It shouldn’t replace high-fidelity standardised patient work where nuance and relational depth are the point. But it can dramatically increase access to practice, create safer “first reps,” and make debrief more consistent.

And perhaps most importantly: it invites us to take a new question seriously.

Not just, “Can we simulate this scenario?”

But, “Are we simulating it in the way people actually live it, and at the right level for the learner?”

Because sometimes the hardest part isn’t the patient.

It’s the conversation, under real conditions.

FAQ

• Does conversational AI replace standardised patients?
  No. Its strongest role is increasing practice frequency and supporting debrief at scale; standardised patients remain essential when relational nuance and human unpredictability are the learning objective.
• Why simulate phone calls specifically?
  Phone calls remove visual cues and support; silence, tone, and interruption behave differently, so skills don’t always transfer from face-to-face role-play. Simulating the channel improves transfer.
• Where does AI feedback help most?
  As a debrief scaffold: surfacing structure, clarity, missed checks, and escalation points, without replacing educator judgement.
• How can I request a demo of SimFlow.ai?
  You can book a demo via the SimFlow.ai website, or email the team at info@simflow.ai to arrange a walkthrough for your simulation programme.

READ ALSO

total words:1848 - positions {"top":0.3,"middle":0.6,"bottom":0.9}