Long live the builders: Thê-Minh Trinh

So Thê-Minh, who are you?

I’m someone very curious, to the point of obsession sometimes. That’s probably how I got into Product Management!

I’m currently with Lifen as a Senior Product Manager. Lifen is a HealthTech company making apps to 1) help hospitals send and receive their medical documents, and 2) also accelerate research on rare diseases by structuring medical data at scale.

How would you define your relationship to GenAI in general?

I recently embraced GenAI and I’m still figuring out the best way to use it. It definitely is a fantastic piece of technology, allowing human-like interactions (written, oral and images)! But since it’s probabilistic, it requires a high level of awareness and double checks.

When did you start using it?

I tried it with the first launch of ChatGPT but I was skeptical and quite disappointed at first: a lot of hallucinations, hard to figure out a great use case, no simple way to verify sources… The ROI wasn’t there.

What changed then? What was the trigger?

The trigger: a combo of being well-informed + exposure to enthusiasts with use cases that resonated and a better understanding of how to write great prompts and tech maturity (Claude 3).

For the few remaining skeptics over there, what would be your main advice?

My advice if you are still in the frustration or indifference phase: take inspiration from your peers who find value in it, then take the time to experiment yourself!

Last time we spoke, you mentioned a blog post shared internally, why did you write it? What have been the reactions so far?

I wrote it because I wanted to share my journey to this Aha! moment, which was key to unlock all the rest. Great reactions and a lot of discussions with colleagues on their similar journeys and how to make the most of GenAI at work.

What are your most frequent use cases these days?

Most frequent use cases:

Prompt Wizard: My Master prompt. Helps me create a great prompt from scratch, criticizes it (with a 5-star rating) and follows up with questions to improve it.

Domain Jargon Helper: Explains jargon, acronyms, concepts related to the vast domain of the health system and hospitals. The way it simplifies, breaks down the topic and illustrates it is a great way to start.

Prototyper: I remember messing with Balsamiq to show my UX ideas. Now I can iterate extremely quickly on clickable prototypes from a few sentences and even a photo of a sketch!

Awesome. Let’s get into the weeds shall we; can you share your step-by-step process to create these agents?

I started from Dust’s very own “promptWriter” and borrowed this little twist from someone else:

Here’s the expected output:
1. The prompt. Give the best prompt according to the input.
2. Criticize the improved prompt, severely. Give a visual rating from 0 to 5 stars (0 for bad, 5 for optimal) and say what improvements are needed so the prompt is 5 stars. All hypotheses and/or problems must be addressed in your review.
3. Questions. List all the questions that need to be answered to improve the prompt.

I found that having a visual rating was key for my understanding. And follow-up questions are just a natural interaction.

How do you see your work changing?

For now, mostly productivity improvements. In the future, it could completely reshuffle the speed at which product teams talk to their customers, the way they work together, product teams sizes… But it’s still too early to tell.

What would you advise a young Thê-Minh starting to explore product management today?

The advice I’d give to my younger self would probably not be that different because of this piece of technology: be curious, keep exploring, be useful.

Where would you love to see GenAI evolve?

I’d love to see AI (not just GenAI) evolve to a useful, reliable, safe and frugal technology.

Thank you!