Nezerac

Deals sealed, stress repealed.

Comment

Inspiration

Small restaurant owners wear many hats—chef, manager, marketer, and more. One of the most time-consuming jobs they face is sourcing quality ingredients and supplies. We wanted to free business owners like them from these tedious tasks. That’s why we built an AI agent that does the heavy lifting, so restaurant owners can focus on what they love most—running their business and delighting their customers.

How it works:

  • Tailored Prompts for Your Needs: The AI agent begins with a specific task, like "Find suppliers who sell quality cuts of steak. We need 200 lbs per month with a budget of $2,000."
  • Intelligent Lead Generation: Using tools like Oxylabs, the agent scrapes web data based on location, company information, and the prompt. It creates a targeted list of suppliers or warehouses likely to meet the restaurant's needs. Most importantly, it scrapes their emails which we will use to prospect and negotiate potential deals with them.
  • Smart Outreach: The agent emails each supplier on the list, asking questions to determine their suitability.
  • Negotiation and Quotes: It negotiates pricing and terms, retrieves quotes, and schedules proposals on behalf of the restaurant.
  • Centralized Decision-Making: All potential deals are displayed on a user-friendly workbench where restaurant owners can review quotes and schedules, then simply click "Accept" or "Decline" to finalize deals.

Tech Stack

  • The brain of the application (the bot replying/negotiating/prospecting with suppliers) was built using python and Anthropic. We created multi-classification stages to handle approvals, rejections, and email follow ups
  • The email sending and replying are done using AWS SES
  • oxylabs was used to scrape data
  • most of the backend, from UI update/retrieval apis to the auto-reply bot/agentic workflow was built with Lambda functions
  • s3 was used to store our filtered lists (after web scraping)
  • dynamo stores all of the restaurant and agent data, as well as the email history and relevant metdata
  • The workbench for the restaurants was built in nextjs14, hosted on AWS Amplify

Challenges

  • Fetching leads was difficult because small businesses often don't have emails listed. We explored various options such as Apollo
  • Setting up email threading was challenging

With our tool, restaurant owners can focus less on the back-and-forth of supplier management and more on creating memorable dining experiences for their customers. All the restaurant has to do is give natural language instructions to the agent, start the agent, sit back while the agent finds deals, and when the deals populate in their workbench, complete the final step of closing the deal with a click of a button.

Built With

Share this project:

Updates