Improving the Beam UX with Generative AI

There are a lot of headlines in the news (and on LinkedIn) about AI to the point where it feels the technology is being used as a gimmick to capture searches than delivering value to users and customers. To see if others were feeling the same, I spoke to several VCs and entrepreneurs last week about how they think AI will impact their businesses. 

This space is immensely crowded. Founders are having difficulty building AI features that can be tested and trained with users. Tons of AI companies have formed in the last four months. YC announced that 70% of their cohort are generative AI companies. Google, Microsoft & Amazon offer cloud credits to entice founders to build AI applications. 

Even Healthcare giants like Epic have partnered with Microsoft to integrate GPT into their EHR. While the headline certainly sounds engaging, the press release is ambiguous. The main functionality comes from data visualization in its internal tools. Healthcare executives still need to figure out GPT in technology that interacts with patients and providers. 

Ultimately – we need to ask ourselves where the value of this technology comes from. Is it a marketing tactic for large companies to appear technically ahead? Is it replacing admin-level jobs or a way to make products more human? 

I believe it’s the latter. There is value in this technology, but most companies talking about integrating this value are late to the game. At the pace these companies typically move, we won’t see any meaningful integration for a while. 

So now, to this article’s point – how do you differentiate? It starts with educating customers and users about what AI could actually do and sitting with them as they use the technology. At Beam, for example, we’ve logged the time it takes to perform several actions on our platform. We take these benchmarks and compare them to trends in the market – for example, studies correlate EHR use to physician burnout. This is consistent with how healthcare providers use Beam and its EHR integrations. In the past, we’ve had users use Beam for telehealth and clinical operations – we would have our team manage the integration with the EHR. We can now transcribe asynchronous and synchronous telehealth and in-person consultations using OpenAI’s Whisper API. 

The most significant current limitation is that Whisper still can’t natively differentiate between who’s speaking (patient vs. provider). Once addressed in the next Whisper update, we can give our providers more than 95% of their time back.