Medtech Teams, Pull Up A Chair For Your New AI Teammate
By Chris Danek, Bessel LLC
Do you compete through innovation? The best innovators will bring artificial intelligence (AI) into their teams.
The buzz about AI isn’t new — for the past few years, we’ve heard about “Industry 4.0,” the “fourth industrial revolution,” and technologies like high-performance computing, digital engineering, and artificial intelligence. But this change moment is much bigger than any industrial revolution we’ve seen before. We aren’t building a new version of railroads. We’re at the beginning of a technology revolution, and how we use these tools together in teams will shape the future.
Your teams can embrace this moment of change and opportunity and pull up a chair for a new, transformative teammate: AI.
How To Work With An AI Teammate: 4 Steps
Working with AI on teams isn’t really about shiny new tech. AI is just another teammate that happens to help us work in transformative ways.
Think about putting AI to work by following four steps: define what you want your AI teammate to deliver; learn how to communicate with your AI teammate; verify the output; and review and integrate that work product into your team’s deliverables.
Step 1: Define what you want your AI teammate to deliver.
Let’s say you’re working on a new product and looking to ChatGPT to help you. Before writing your first ChatGPT prompt, think about what you want to get from the AI. What are your objectives, and how will you show up in your conversations with AI to meet those objectives?
Do you want to:
- Brainstorm and ideate? Learn something new? Check for something your team may have missed? Then you should play the role of interviewer.
- Write code or create designs alongside your AI teammate? Then you should play the role of a collaborator and co-creator.
Once you know what you want and your role, you can approach your AI teammate with the right mindset and give it the context it needs to deliver.
Step 2: Learn how to communicate with your AI teammate.
Interacting with AI is a new fundamental communication skill that millions of people worldwide are learning. It’s so critical that you’ll find new certification courses on “prompt engineering.” But I’ve found that the easiest way to communicate with AI is to keep it human. Think of ChatGPT as a stakeholder and craft the conversation like you would with any other stakeholder.
(Note: You can find a guide for interviewing any stakeholder, including robots, in the Field Guide to Human-Centered Design by IDEO.org [page 39], a free download.)
Here’s a template for starting your conversation that one of my student teams developed:
- State who you are.
- State the topic and goal.
- Add specifications or another context.
- Ask ChatGPT to take the role of someone.
- Be specific in what you want from it.
Here’s an example prompt built on that template:
I’m an R&D engineer working on a design project to create a small speaker enclosure that would work in the frequency ranges of 80-200 Hz and be roughly the size of a cell phone. As an audio engineer, how would you design it and why?
The prompt checks each box:
- State who you are: "I’m an R&D engineer working on a design project..."
- State the topic and goal: "To create a small speaker enclosure..."
- Add specifications: "…that would work in the frequency ranges of 80-200 Hz and be roughly the size of a cell phone."
- Ask ChatGPT to take the role of someone: "As an audio engineer..."
- Be specific in what you want from it: "… how would you design it and why?"
Beyond the initial prompt, think about how to enrich the conversation. Simple tweaks that improve conversations with people also work well with ChatGPT. Try starting at a general level in order to allow discovery. Then dig deeper into anything interesting. If you need to, zoom out and revisit essential topics.
Step 3: Verify the output from your AI teammate.
If your team is creating a new prototype or product feature, do you assume it’s perfect immediately? I doubt it. Instead, you work as a team to test it, expecting to refine and improve it. You can think of an AI teammate’s output the same way — check it just like you do work from human teammates.
How might you verify the output?
- Have a different team member check the output. Focus on whether the output is accurate and whether it answers the prompt. Independent verification is necessary, and involving another team member helps distribute interaction with your AI teammate, further integrating AI into the team.
- Borrow from software QA and set up a workflow that mimics software developers’ practice of continuous integration and testing.
- Work empirically. Sometimes the best way to check an answer is to build and test a prototype.
- Keep working with your AI teammate. As you refine its work, feed the corrected information into the AI for refined responses.
- Ask your AI teammate to provide references and URLs to back up its output. Wharton Professor Ethan Mollick, who requires students to use AI in his entrepreneurship classes, suggests: “If you get something wrong, paste in the incorrect text and ask it to explain the error or to walk you through it step-by-step. The results can be surprisingly helpful.”
During this process, keep your objective and the prompts in mind to ensure you’re aiming at the right target. By checking every output, you’ll avoid the hallucinations (convincing but made-up answers) and mistakes we’ve all heard from AI.
Step 4: Review the verified output and integrate it into your team’s deliverables.
Finally, it’s time to review the AI’s output. Medical device development teams perform cross-functional review and approval in almost all work facets — design reviews, risk management, change control, and complaint handling. This review is essential and required by the U.S. Quality System Regulation and to meet ISO 13485 quality management standards. When an AI teammate has contributed to the deliverables under review, make sure the reviewers are informed and that you carefully document your use of AI:
- What did AI contribute to the work under review?
- How did the team verify the work?
- What risks or unintended consequences might follow from this use of AI?
AI hallucinations are a known hazard, with leaders at major AI players like Google warning about their danger. Your team might be too close to the work to recognize all of them. A formal cross-functional review of your AI teammate’s work is essential.
In summary, I’ll leave you with this challenge: AI is here. You could ignore it, but that would be to your detriment. Instead, I invite you to be one of the forward-looking teams that welcomes and integrates AI as a critical team member.
About The Author:
Chris Danek is the CEO of Bessel LLC. He is a serial entrepreneur and veteran of the life sciences industry. At Bessel, he works with entrepreneurs, startups, and established company teams to develop breakthrough medical device technologies. In prior roles, he was co-founder and CEO of AtheroMed (now Philips AtheroMed) and VP of R&D at Asthmatx (acquired by Boston Scientific). He is a visiting professor at the W.M. Keck Center for 3D Innovation at the University of Texas at El Paso, an advisor to the Santa Clara University Healthcare Innovation and Design Lab, and an inventor of more than 85 U.S. patents.