Spark Lunch & Learn Recap: What a Live Demo Shows

The best way to understand PCS.ai products is to see them in action. That's why we run Lunch & Learns: 30 minutes, a live demo, and your questions answered in real time. We kicked off the series with SimVox (catch the recording here), and this session put Spark in the spotlight.

Nick Stoick, VP of Sales & Partnerships, ran three Spark scenarios: Alex, an adult with chest pain; George, a stroke patient; and Violet, a sick kiddo who came in with her mom. Here are the questions those scenarios answered, along with what you'd have seen on screen.

Jump to a question:

• What is Spark, exactly?
• How do learners get into Spark?
• Are the conversations scripted?
• How is Spark's AI different from ChatGPT?
• Can learners do a physical exam?
• Can learners order tests and treat the patient?
• How do learners know how they did?
• What do faculty see?
• Can a virtual patient sound sick?
• Can learners talk to a patient's family member?

What is Spark, exactly?

Spark is a virtual patient platform, and it's been around since 2019. Learners talk to a 3D patient in their own words, examine them, order tests, treat them, and get scored on all of it.

55+ scenarios out of the box. They cover outpatient care, acute care, mental health, women's health, and pediatrics. You can also build your own cases, so the library isn't the limit.

How do learners get into Spark?

Two ways:

• Browser: Open Spark anywhere at app.pcs.ai.
• LMS: Launch it from Canvas, Moodle, Blackboard, Brightspace, or another major platform without another login.

Nick ran the demo on his computer, but learners can also use a tablet, a phone, or a VR headset.

Are the conversations scripted?

No. Nick opened with "What brings you into the hospital today?" and followed up on onset, severity, shortness of breath, and radiation. Alex answered with details specific to the case: the pain started the evening before while he was watching TV, he rated it 6 out of 10, and it had spread to his back and left shoulder blade.

There's no menu of questions to pick from. Learners have to work out the right line of questioning themselves, which is exactly the skill you're trying to build.

How is Spark's AI different from ChatGPT?

PCS.ai built and tuned its own model in-house. That focus pays off in three ways:

• Standardization. Every learner meets basically the same patient and gets very similar answers, so you can compare scores across a group with confidence.
• Speed. The model is smaller than general-purpose chatbots, so the patient answers faster and the conversation stays lifelike.
• Realism. The model plays a patient, not an encyclopedia. It understands everyday medical terms but not clinical jargon. If a learner asks something a typical patient wouldn't understand, the patient asks them to explain.

Can learners do a physical exam?

Yes, a complete head-to-toe exam. When Nick clicked Physical Exam, Alex lay down supine and a tray of tools appeared:

• Stethoscope. Listen to the heart, lungs, and bowel. Faculty can pick from a library of sounds to expose learners to different findings.
• Visual exam tools. Check the eyes, ears, nose, mouth, throat, or skin to see a finding and an image. Upload your own images to change the pathology, such as a patient with conjunctivitis.
• Glove. Palpate the abdomen by quadrant, or do a back or breast exam, with findings for each area.
• And More...

Adjustable difficulty: To make a case harder, remove the written "no abnormal findings" notes so learners have to read the images on their own.

Can learners order tests and treat the patient?

Yes. Alex's monitor showed low SpO2 and a high respiratory rate from the start. In the Diagnose & Treat window, Nick:

• Ordered a chest X-ray and a metabolic panel, with results back right away
• Set a primary diagnosis of STEMI, added and ranked differentials, and added comments explaining his reasoning
• Gave Alex oxygen, and watched his SpO2 begin climbing from 91%

Ordered tests stay listed for review, and treatments change the vitals, so learners see the effect of what they do.

How do learners know how they did?

They get feedback the moment the simulation ends. Nick's demo run scored 40%, and the assessment showed exactly why: strong history of present illness, but no past medical, social, or family history. Learners can open each category to see what they asked and what they missed.

• Scoring: Built-in criteria for both medicine and nursing. Change the criteria, adjust the weighting, or write your own to match your curriculum.
• Communication feedback: How learners phrased questions, whether they used open-ended questions, and whether they showed empathy and asked about the patient's perspective.
• Quizzes: Some scenarios include a multiple-choice or free-text quiz to run before or after the simulation, for prebriefing or debriefing.

What do faculty see?

• LMS gradebook: If Spark is connected to your LMS, the overall score goes straight to the student's gradebook.
• Sim Log: Faculty can log in to Spark to see the full session, in order: what the learner said, how the patient answered, every exam action, and when the learner earned credit.

Can a virtual patient sound sick?

Now they can. George Callahan, one of Spark's newest scenarios, is having a stroke, and you can hear it in his slurred speech as he explains that his students said his face looked like it was drooping.

• New AI voices: Launched at the start of this year, after PCS.ai spent a year bringing its voice engine in-house. Spark previously used Amazon and Microsoft speech synthesis, which couldn't carry emotion or pathology.
• Slurred speech and more: A crying, hysterical voice has since been added too, so learners practice getting information from a patient who is hard to understand.

Can learners talk to a patient's family member?

Yes. Violet is a pediatric patient in a hospital bed with her mom at her side. Learners can direct a question to either one. Otherwise, whoever is the more natural person to answer will respond.

Violet said she had a bad cough, her chest hurt, and she wanted to go home and watch Netflix. Her mom explained that the school nurse had sent Violet home with a 101-degree fever.

That Netflix line is intentional. PCS.ai has tuned its pediatric patients to talk like real kids. Ask one to rate their pain from 1 to 10, and you might hear "a million." Kids are so dramatic 😊

What's next?

This session covered what learners' experience. The next Lunch & Learn covers the faculty side: patient authoring on the pcs.ai platform. We'll edit a case live and show how fast the changes reach learners. Sign up here.

Want to see Spark with your own scenarios? Email sales@pcs.ai to set up a one-on-one demo with Nick, Erik, Marcus, or Read.