Humans and AI
Humans and AI

Features

How to safely use AI in patient-facing situations.

How to safely use AI in patient-facing situations.

Puppeteer combines AI with fix responses and human's step in, ensuring patient safety with features for sensitive situations.

Luca Spektor

|

Aug 2, 2024

|

4

min read

Our Approach:

Puppeteer is a patient-facing solution designed to handle sensitive, human-delicate situations, making our product uniquely capable of addressing serious scenarios.

In the AI boom, where everything seems to be best handled by AI, at Puppeteer we see the most potential in the collaboration between humans and AI. So much so, that the name “Puppeteer” comes as a description of our product’s capability of pulling the right set of strings to bring the best “Puppet” into the scene. In simple terms, this would describe the ability of our AI Assistants to understand in which situation they are and answer accordingly with the best solution: in the majority of cases an AI response, however, when needed a fixed message or a human alert to step in.


Pay Pal's Hybrid Solution For Frauds

This substitution or collaboration dilemma is not new regarding technological advances and human jobs, and the thought of joint work is not a consolation for humans, it’s the best solution for companies to implement. To set an example: early in its operations, PayPal faced significant losses due to fraudulent charges, costing millions per month. Co-founder Max Levchin initially deployed sophisticated software to detect and cancel bogus transactions, but the adaptive tactics of fraudsters quickly rendered this approach ineffective. Realizing that human analysts were better at identifying these evasions, Levchin, and his team developed a hybrid solution: software flagged suspicious transactions for human review, combining the strengths of both computers and people. (Read more on this case here) This approach highlighted that computers should be seen as tools, not substitutes for humans, making the hybrid model more effective than relying solely on one or the other.

So, how would this hybrid situation look at Puppeteer?

As mentioned in the article "What is Puppeteer and How Do We Create AI Assistants?", we customize our AI Assistants according to the behaviors needed in each case. However, we have two key features that focus on patient safety:


Guardrails


Like physical guardrails along a highway, our guardrails ensure that users stay within the intended scope of the AI's capabilities. If irrelevant or off-topic topics are mentioned, the AI guides the user back to the relevant conversation, preventing deviation from the intended purpose.

This is the first security element to ensure every conversation follows the designed path. However, users can sometimes send messages that divert from this structured path, so in those cases, we resort to:


Safeguards


Our safeguards detect serious messages where the conversation can't continue and act as configured to follow the necessary next steps focused on the patient’s safety.

For instance, if a mental health clinic receives a message from a patient who says they are trying to harm themselves, a fixed message to call 911 would be sent to the patient along with an alert to a doctor and the patient’s family.


Mental Health Chat Sample



Conclusion


Puppeteer combines AI’s potential with humans on the loop to handle sensitive situations effectively. By using guardrails and safeguards, our AI Assistants ensure patient safety and relevance, maintaining the essential human touch in healthcare. This hybrid approach optimizes outcomes, demonstrating that the best solutions come from leveraging both AI and human strengths.

Request a demo

Request a demo

Get custom pricing and become a part of the Puppeteer community

© 2024 Puppeteer. All rights reserved.

© 2024 Puppeteer. All rights reserved.

© 2024 Puppeteer. All rights reserved.