Skip to main content

Use Case Challenge #1: Healthcare AI 

 

 

We're thrilled to kick off the first challenge of Bot Games Season 4. This season, we're encouraging users to explore the potential of Generative AI and how it can enhance the functionality of your automations. Use cases that were previously impossible to automate can now be automated thanks to the Automation Success Platform paired with Generative AI. We can't wait to witness the innovative solutions that our Automation Anywhere Developer Community will develop for this challenge. It's time to get your creative juices flowing and reimagine the way you've previously thought about automation!

If you’ve never used ChatGPT before, you may want to create an account here. The OpenAI platform provides a trial version to test it out for a limited period. You can get the API for your OpenAI account here.

Note - you aren’t limited to only using OpenAI’s ChatGPT - you should be able to solve these challenges with any number of Generative AI model providers - we’re just using ChatGPT since that was our primary tool for most of our testing.

OpenAI Generative AI Package is available in the Bot Store. You can read the included documentation or check this quick start video showing the full setup process: 

 

If you are using your own enterprise version of Control Room, you can download the OpenAI Generative AI Package from the Bot Store and import it. If you are using the Community Edition, it is already pre-loaded.

Note: This is a forward-looking use case, demonstrating the power of integrating Generative AI within the Automation Success Platform using strictly fake data. If you plan to consider automating use cases that contain PHI or PII with Generative AI, please consult your legal team before proceeding.

Like experimenting with any emerging technology, you may encounter a few things that feel wonky in these Generative AI challenges. Part of the "challenge" in this case is learning how to effectively roll with it, write better prompts, and solve interesting problems using the newest and best of AI + automation.

 

Challenge Objective

 

A healthcare company, Medikorps, provides treatment for a wide variety of patient conditions and wants to standardize its report-generation process. Once a patient visits with a doctor, the doctor analyzes the patient and shares his/her notes about the patient's condition along with prescribed medications if needed. The doctor then sends an email to the medical assistant to generate a medical report to be shared with the patient. As the healthcare facility receives a large number of patients every day, generating reports has become a bottleneck. Therefore, the facility intends to implement an automation solution to streamline the process.
 

Rules

 

Your automation should launch the challenge page, then launch the mailbox from the challenge page, as both are interlinked. Then, send the information in the doctor's notes to a Generative AI provider of your choosing (ChatGPT for example) to extract the required details in a format that can be easily interpreted by the automation.

The mailbox will have 10 emails in total, of which only a few will be unread. Data needs to be extracted only from the unread emails to be entered into the report generation application. Note that the unread emails will change every time you open the mailbox.

You should select the 'Medication Prescribed' checkbox only if the doctor has prescribed any medication.

 

Sharing Your Results

 

Once you’ve successfully completed the challenge, share a screenshot of your results on any social media platform using the hashtag #BotGames and tag 3 other Automation developers who you’d like to challenge to complete this exercise. We want to celebrate your accomplishment as well, so our team will be looking for individuals who share their results on social media (and tag/challenge others) to reward them with limited edition Bot Games swag. To be eligible for swag, you must be registered for Bot Games - sign up here.

You can also download a certificate of completion once you solve this challenge!

Please share your feedback. We want to hear what you enjoyed and what frustrated you about this challenge. Reply & share some tips with your fellow developers - help us build a stronger community!

 

Conclusion

 

One great way to implement Generative AI with Automation is to extract/summarize data into a standard format. In this case, Generative AI is used to extract specific information from long-form unstructured text and your automations can leverage this data to automate complex end-to-end workflows. This can be useful in different domains to extract information from legal documents, research papers, and so much more.

 

Unlock all Bot Games Season 4 challenges and resources - visit our Bot Games Season 4 HQ today! 

Completed the challenge!!

Completed Bot Games Season 4: Challenge 1: "Healthcare AI"

Thank you @arjun.meda for this Challenge

Followed Best Practices:
Stored the API key in Credential

Packages Used:
Steps
Loop
Recorder
If
Browser
Javascript Package

OpenAI Model used:
text-davinci-003


OpenAI: chatGPT action in giving error: "OpenAI failed to process the request. Error: HTTP 404". Have used ‘text-davinci-003’ model, changed account/ API key but to no avail. What could be the issue? @arjun.meda  


OpenAI: chatGPT action in giving error: "OpenAI failed to process the request. Error: HTTP 404". Have used ‘text-davinci-003’ model, changed account/ API key but to no avail. What could be the issue? @arjun.meda  

As per OpenAI documentation, APIKey is incorrect, could you please double check. Here is the link I am referring to: https://help.openai.com/en/articles/6891839-api-error-code-guidance


Yes Arjun I did check it and generated new API keys, cleared browser cache but still error remains the same.

 

 


Yes Arjun I did check it and generated new API keys, cleared browser cache but still error remains the same.

 

 

Share a screenshot of your config for the OpenAI action


 


 

I would swap to the GPT-3 action if you’re going to use text-davinci-003. So your options are to swap to the GPT 3 action, or continue using the ChatGPT action (which works with GPT 3.5 and 4) and use the gpt-3.5-turbo model.

Note: if you’re going to stick to this ChatGPT action (instead of GPT3) - you’ll want to remove the session - as your query for doctors notes would be independent of other requests - not an ongoing conversation


Worked! Thanks a ton for the rescue Micah!


✅Completed Bot Games Season 4: Challenge 1: "Healthcare AI"

It was my first ever bot games participation and I enjoyed it a lot.

It was a great use case to leverage the capabilities of RPA + AI.

✅Packages Used:
🔸Open AI (GPT-3)
🔸JSON Object Manager
🔸Recorder
🔸Loop
🔸Step
🔸If
🔸Error Handler
🔸Browser
🔸Number
🔸String
🔸Comment

Thanks @Micah.Smith, @arjun.meda for bringing such wonderful challenges as part of #BotGames
 

Bot Games 4 - AI Healthcare Challenge

#Community #Pathfinder #GenerativeAI #RPA



Great challenge!😎
Fantastic use case. 

Good luck to everyone!🤖


Completed Bot Games Season 4: Challenge 1: "Healthcare AI"

 


Happy with that.. first use of ChatGPT within RPA.

 


ce12c188-8d7e-4cdd-857b-fb2a03441cec


Completed the first challenge Bot Games Season4 - Healthcare + AI #botgames #season4 #a360

Tips:
✅ Package Used:
✅ Browser, Loop, If. List, Error Handler, Number, Step, String, Comment
✅ OpenAI: GPT-3 (Used Model: text-davinci-003)
✅ Json Parser: to convert data to multiple strings
✅ Recorder > Use Dynamic DOMX path for webmail iteration


Thanks @Micah.Smith@arjun.meda for this great challenge!!!


Reply