Dealflow

Inspiration

Someone I know negotiated a job offer last year. The recruiter knew the salary bands, what other candidates accepted, and how much flexibility they had. My friend had Google searches and Reddit threads. They accepted the first number. Later they found out they were about $20K under market. That gap is the real problem. In most negotiations, one side has the data and the other is guessing.

What it does

DealFlow is an AI negotiation tool for salary, rent, and SaaS contracts. You enter the details of your situation. It pulls real market data, shows where you land, generates a clear counter offer, and gives you a ready-to-send email and call script. You can also practice the conversation with voice roleplay before the real call.

How we built it

We built DealFlow as a simple agent pipeline that goes from research to strategy to action. A research layer pulls market data for the exact situation, a strategy layer turns that data into a clear counter offer and leverage points, and an action layer generates a ready-to-send email, call script, and pushback responses. The backend is Python with FastAPI, the frontend is Next.js, and we added voice roleplay so users can practice the conversation before the real call.

Challenges we ran into

The hardest problem was getting reliable market data. Many sources are inconsistent or fragile, so we had to add caching and fallbacks to keep results usable. We also had to handle occasional malformed model outputs and spend time tuning voice roleplay so it felt realistic instead of robotic.

Accomplishments that we're proud of

We built an end-to-end negotiation experience that goes from raw data to a usable counter offer in under a minute. Users do not just get advice, they leave with concrete emails and scripts they can immediately use, and the voice practice noticeably increases confidence going into real negotiations.

What we learned

The AI is not the bottleneck. Models are already good at negotiation strategy. The real constraint is data quality, because a strong counter offer depends entirely on accurate market context. We also learned that clarity beats complexity, and users want one strong recommendation, not a long explanation.

What's next for Dealflow

We are expanding into more negotiation types like car purchases, contractor quotes, and medical bills, each powered by new data sources. We are also adding outcome tracking so users can see how much value they save over time, and building integrations that surface DealFlow directly at the moment someone is negotiating.

Built With

• anthropic-claude
• elevenlabs
• fastapi
• next.js
• pdd
• python
• react
• rtrvr.ai
• sendgrid
• tailwind-css
• toolhouse
• typescript