Correspondence AI Agent

AgentHack submission type

Agentic UI Automation

Name

Gokul Jayakumar

Team name

Team Agentic AI Mavericks

Team members

@chilla.shashreek, @sujith_naik, @Gokul_Jayakumar

How many agents do you use

One agent

Industry category in which use case would best fit in (Select up to 2 industries)

Healthcare Pharma
Insurance

Complexity level

Intermediate

Summary (abstract)

“From Document Chaos to Instant Action – Meet Your AI-Powered Correspondence Agent!”
We’ve built an Intelligent UiPath Agent that reads healthcare documents, understands what type they are, and instantly triggers the right workflow — no human triage, no delays, no errors.

Detailed problem statement

In healthcare, we get flooded with all kinds of documents — EOBs, denial letters, appeal responses, you name it. The good thing is, we already know what type each document is. But here’s the catch: depending on the type and subtype, we’re supposed to take different actions — like following up on a claim, triggering an appeal, or routing it to a specialist team.
Right now, a lot of this is done manually — someone has to open the document, read through it, figure out what needs to be done, and then actually do it. It’s slow, error-prone, and honestly, not scalable.
We’re wasting valuable time and resources just moving paperwork around when we could automate it.

Detailed solution

We built a smart UiPath agent that doesn’t just stare at healthcare documents — it reads them, understands them, and knows exactly what to do next.
Each document already comes with a type and subtype label. Based on that, the bot refers to a predefined SOP (like a rulebook) and automatically kicks off the right UiPath process — whether it’s a claim follow-up, denial resolution, routing to the right team, or something else.
No more opening every PDF, scratching your head, and forwarding it manually.
The agent does the thinking and doing — instantly.

Demo Video

Expected impact of this automation

:white_check_mark: No more manual triaging – the bot reads and reacts faster than any human could.
:three_o_clock: Huge time savings – what used to take hours now takes seconds.
:briefcase: Ops teams can finally breathe – they’re not drowning in documents anymore.
:prohibited: Fewer errors – decisions are SOP-driven, not guesswork.
:chart_increasing: Easily scalable – whether it’s 10 docs or 10,000, the bot doesn’t break a sweat.

UiPath products used (select up to 4 items)

UiPath Agent Builder
UiPath Orchestrator
UiPath Robots
UiPath Studio

Integration with external technologies

OpenAI

Agentic solution architecture (file size up to 4 MB)

Sample inputs and outputs for solution execution

Input is the document type and Document subtype like
The Document Type is EOB. The Sub Type is Payment EOBs. The account number is 4965. Route it to the queue ‘payment Team’

The Document Type is Claim Response Letter. The Sub Type is Claim Acknowledgement letters. The account number is 85643. The patient name is John Doe. Post Notes in billing system

The Document Type is W9 Form Request. The Sub Type is Requests for W-9. The account number is 344512. Send the W9Form via fax application

Other resources

PPT: Agentic AI Maverics.pptx - Google Slides

Hello,

I can’t find your idea in the submission form in the accepted ideas sheet. Can you give me the link of the org where you built it? Did you build it through Labs, as per the rules?

Hello,

Thanks for reaching out!

I had submitted the idea via the official Hackathon submission form, but I didn’t receive any follow-up link or instructions to post it in UiPath Labs.
Could you please confirm if my idea was not approved, or if I missed a step?

Happy to share any details or re-submit if needed.

Thanks again for your support!