Project Sarah

Project Sarah automates the full B2B sales workflow: from identifying ideal customers and decision-makers to sending personalized outreach, turning manual sales into a scalable, autonomous system.

Comment

Inspiration

Project Sarah was born out of a very real and recurring pain point: B2B sales is still highly manual. Identifying the right companies, finding
the correct decision-makers, and crafting personalized outreach takes a significant amount of time, time that could be spent closing deals or
building relationships.

At the same time, these processes are difficult to scale. What works for 10 leads quickly breaks when applied to 1,000.

The Problem

Today’s B2B sales stack is fragmented.

Tools like Apollo.io help with parts of the workflow, such as lead generation or contact discovery, but there is no truly end-to-end automation.
Human input is still required at multiple stages:

  • defining target customers
  • identifying the right contacts
  • crafting relevant outreach

This leads to inefficiencies, inconsistent quality, and limited scalability.

Our Insight

Sales should not be a collection of disconnected tools, it should be a continuous, intelligent system.

With recent advances in AI, it is now possible to move beyond assistive tools toward autonomous agents that can understand context, make
decisions, and execute workflows end-to-end.

Project Sarah is built on the belief that B2B sales can be fully automated without sacrificing personalization.

How It Works

Project Sarah transforms a traditionally manual workflow into a seamless, automated pipeline:

  1. The user logs in and provides their website and key company information
  2. The agent analyzes this input to understand the business, positioning, and target audience
  3. It scans the internet to identify potential customer companies
  4. These companies are ranked based on relevance and fit
  5. The system identifies the most relevant decision-makers within those organizations
  6. Sarah generates highly personalized outreach messages tailored to each contact
  7. Optionally, outreach can be extended with AI-generated voice messages using ElevenLabs.

The result is a fully automated outbound sales engine.

What Makes It Different

Unlike traditional tools, Project Sarah does not optimize individual steps, it replaces the entire workflow.

  • End-to-end automation instead of fragmented tooling
  • Context-aware personalization at scale
  • Continuous learning through user interaction and feedback
  • Reduced dependency on manual input

Rather than assisting sales teams, Sarah acts as a fully autonomous sales agent.

How We Built It

The system combines multiple technologies into a unified workflow:

  • LLM-based reasoning and generation (e.g. via Claude)
  • External data enrichment through APIs such as Apollo.io
  • Web scraping and data aggregation for market discovery
  • Optional voice generation through ElevenLabs

The focus was on building a fast, functional MVP that demonstrates the full loop from input to outreach.

Challenges

One of the main challenges was bridging multiple steps into a coherent, autonomous system.

While individual components (data sourcing, contact identification, text generation) are relatively mature, combining them into a reliable
end-to-end flow required careful orchestration.

Another challenge was maintaining relevance and quality in personalization, ensuring that automated outreach does not feel generic.

What We Learned

The biggest insight is that true automation is not just about replacing tasks, it is about connecting them.

Small inefficiencies compound across the sales process. By unifying everything into a single system, the overall impact becomes significantly
larger than optimizing individual steps.

Future Vision

Project Sarah aims to evolve into a fully autonomous sales infrastructure.

The long-term vision is to move beyond outbound automation toward:

  • continuous pipeline generation
  • adaptive messaging based on real-time feedback
  • and ultimately replacing traditional CRM systems with an intelligent, self-operating sales layer

Sales will no longer be managed, it will run itself.

Built With

  • auth
  • base-ui
  • cheerio-(web-scraping)-cloud-&-hosting-vercel-(hosting
  • deployment)-tooling-pnpm
  • eslint
  • functions
  • google-places
  • languages-&-frameworks-typescript
  • lucide-react
  • next.js-16-(app-router)
  • playwright
  • react-19
  • row-level-security)-apis-&-data-apollo.io-(lead-discovery)
  • sonner
  • tailwind-css-4-ui-&-components-shadcn/ui
  • tw-animate-css-ai-anthropic-claude-(claude-sonnet-4-6)
  • vercel-ai-sdk-v6-database-&-auth-supabase-(postgresql
  • vitest
  • zod
Share this project:

Updates

Submission history