Being Vulnerable About AI

To build on last week's post about the January hell in healthcare, let's see why we need to work on ridding the ecosystem of phone calls altogether. Also, why short term fixes like RPA and Agent Assist just a temporary band-aid on a much deeper problem.

Phone calls are terrible. Especially in a world of COVID / WFH. Check out this example of an agent working with a rooster crowing in the background.

Many AI companies work on agent assist & automation tools with two key success metrics: Call Deflection Percent & Cost Reduction. For executives buying into this, reducing staff by 5% while saving 7-10% of cost is a big enough win to push for the holy grail of interoperability.

So the question is: Why does healthcare interoperability matter? It can drive accurate, instantaneous data access while unlocking healthcare operations staff to do so much more meaningful work like supporting patients & their families instead of waiting on hold and copying & pasting data. Patients would get to get on therapy faster. Providers would have a much better downstream experience with Accounts Receivables. Payers would not have an unnecessary but large line item called ‘Provider Contact Centers’ to read out data from different pdfs and databases.

A temporary band-aid is exactly that. It makes us feel like we have a win, but doesn't move the world forward in the way we need it to. Increasingly, healthcare thought leaders are realizing and vocalizing the fact that the current madness has to end. They are realizing that we need to adopt technology to solve the deeper rooted problems. But they are often blocked by the same pushback.

"In a pilot of the AI, there were errors 1-2% of the time".

Demos of all AI tools look beautiful and perfect. Most companies won't tout where AI falls apart and due to this, as a society we compare AI to perfection. Instead of comparing AI to humans. Let's change that.

My team at Infinitus has been fortunate to work with some trailblazing partners who not only wanted to adopt our technology, but were ready to measure its efficacy side-by-side with humans performing the same tasks. They compared AI to humans, not to perfection. And this is ?? .

Jeff Buck at AmerisourceBergen recently said that Infinitus' Eva digital assistant “can get through calls around 30% quicker and the quality is around 10% higher than humans since there are fewer miscommunications or typos.”

In our sales presentations, we openly talk about when our AI fails. Here's an example where the system misinterprets the agent's name:

It isn't all bad though. AI, when trained correctly and used in the right environment, can be magical. Here's an example of Eva pushing back on an agent when provided incorrect information.

We’ve had to build logic to do real time checking of the data we receive back because Eva talks to humans who are doing routine tedious work (i.e. answering phone calls). These agents are prone to making mistakes (due to being tired, being temporary, or just being careless).

We talk about being vulnerable in our day to day lives. Let's do that for our AI products as well. It'll enable everyone to understand the power, pitfalls and progress that our computers are making.

While we are solving some exciting NLP/AI problems, the biggest problem we are solving is the lack of data interoperability in healthcare. We're using Eva to standardize the way in which data exchanges hands between healthcare entities with a hope of replacing these phone calls with API calls.

We’ve grown our business 5x in the last year, grown the team 2x, and are looking to grow our impact infinite-x (get it?) in the coming years. If you are interested in joining us on this journey, DM me or apply online at https://jobs.lever.co/infinitus