Competitive Intelligence - Athena

Competitive Intelligence & Research Agent that Delivers High Value Insights Through Voice

Comment

Inspiration

I wanted to put together a proper agent that could output readable and useful reports. Reports executives could make decisions from. I found the majority of deep research platforms available have mediocre outputs, and I think that 1/8th of them are rarely applicable. I'm highly interested in research and market intelligence.

I had the idea to turn a data processing pipeline into structured dossiers and artifacts as the output, but I wasn't sure how to actually put it together. Then I found this hackathon. I have a bit of experience with ElevenLabs and have been spending a ton of time with Google's Gemini 3 recently, and figured I should try this out.

What it does

ElevenLabs (Gemini Flash 3 LLM) -> Gemini Flash 3 -> Gemini Pro Thinking (1024 Thinking Budget)

Athena the Competitive Intelligence Agent runs a data processing pipeline. Voice search -> Initializes Gemini Flash 3 for the competitive brief and interface content -> Gemini 3 Pro Thinking then generates a detailed dossier.

How we built it

Athena was initially prototyped in AiStudio.Google.com - I put together the initial concept there, and then I took it to AntiGravity on my desktop and worked through all of the kinks and kept improving the interface until I was pleased with the final output and process.

For the deep dossier, I enabled Gemini Pro's thinking mode with a 1024 token thinking budget. This gives the model room to reason through the analysis before generating output. That's why the dossiers aren't just search summaries. They're actual strategic synthesis.

Challenges we ran into

The biggest challenge was getting the briefing page populating correctly inside the desktop viewport, and pulling accurate information, and getting the dossier into a downloadable .pdf - that was REALLY hard. I'm glad I pivoted to an html document, it was the easiest solution to maintain a high fidelity design.

I kept running into outdated context windows, where it didn't believe me that Gemini 3.0 Flash or Pro existed, and also picking up deprecated SDK's for ElevenLabs (11labs vs elevenlabs).

When initially prototyping inside AiStudio I couldn't seem to get GitHub to connect properly to the studio, so I had to keep duplicating versions every major refactor I made to save progress.

Accomplishments that we're proud of

The system runs three AI processes in parallel, Voice Search then Flash gives you an instant brief so you're not waiting around, while Pro Thinking quietly generates the detailed dossier in the background.

Seeing a data processing pipeline trigger two different versions of Gemini's LLM was incredible to experience, I did not know I could do that, let alone initiate it with just voice.

What we learned

This really showed me the power of Flash models, and how you can QUICKLY process real time intelligent interfaces. I have been in the design industry for YEARS, and knowing that you could essentially populate a hyper personalized interface, along with a more detailed brief inside a fairly complex pipeline was awesome.

What's next for Competitive Intelligence - Athena

I would like Athena or a version of this to be able to store and compare companies, for example...

"Enter your domain" and it would then pull up your core competitors, and give you insight into who those competitors are via listening to review websites, and understanding a deeper sense of the market based on Google footprint and various other sources.

In addition, I'd like to work ElevenLabs into more of an immersive experience, so it actively reads off the dossier in a seamless way.

Share this project:

Updates