AI Agent: SalesOS​

Project Summary​​

Over the course of this summer, I designed and developed a proprietary Python-based AI agent called SalesOS -- created to analyze and score digital marketing performance across websites. This system now functions as a foundational business development tool for our agency, enabling us to deliver fast, insight-rich audits that uncover performance gaps and position our agency as the strategic solution.

​

SalesOS audits both qualitative and quantitative aspects of a company’s digital marketing. It evaluates factors ranging from brand clarity and trust signals to SEO visibility and CRM readiness. The output is an actionable report designed to accelerate new client acquisition and close rates.

​​

​

Framework & Methodology

I began with extensive research into what digital marketing attributes are most correlated with business growth and performance. I studied dozens of factors across UX, content, SEO, analytics, and advertising to determine which areas best predict a company’s ability to attract, convert, and retain customers.

​​

​

AI Platform Evaluation

I conducted deep prompting experiments with multiple LLMs, including ChatGPT, Claude, Gemini, and Genspark to test:

• Which AI models could most consistently interpret and analyze web content

• How to prompt each model for nuanced evaluations

• Workarounds for inconsistent or limited outputs

• Which model would best integrate with our scoring system

​

This led to developing a 20-category scoring framework:

• 10 Qualitative Categories scored by ChatGPT/Claude (e.g., brand clarity, UX, messaging, trust)

• 10 Quantitative Categories scored via tools like SEMrush and native APIs (e.g., SEO visibility, CRM integration, social presence)

• Each category is scored on a 0–5 scale, with custom weighting based on its impact on performance.

​​

​

Technical Implementation

I built the agent in Python, using VS Code as my development environment. The architecture is designed to:

• Accept website URLs as inputs

• Leverage third-party APIs and scraping tools to extract quantitative data

• Use the ChatGPT API to evaluate qualitative elements based on custom scoring rubrics

• Output results in structured reports via terminal, CSV, or presentation-ready formats

​

I used ChatGPT extensively as a pair programming assistant — helping generate code, debug errors, and iterate on structure. Through this process, I developed three major versions:

• V1: A working prototype that could begin to evaluate a single website and generate a qualitative score

• V2: A framework that integrates with API tools and prepares accurate qualitative analysis of a website or multiple websites

• V3: I constructed the framework for full batch-processing capability with CSV input/output, automated scoring, and customizable reporting. For this version, I outsourced a developer to help implement advanced technical components and integrations. I led the outsourcing process, including interviewing, hiring, and managing the developer. I also coordinated the statement of work (SOW) and agreement to ensure alignment with our goals and timelines.

​

This process sharpened my Python development skills, deepened my understanding of AI integrations, and taught me how to turn prompts into production-ready tools.

​​

​

Revenue & Growth Potential

SalesOS directly fuels our revenue engine by generating qualified leads at scale — enabling high-impact client acquisition through partner channels.

• Pilot Program Impact:
  Through 7 beta partners, we’ve unlocked access to ~350 businesses for audit. Based on early analysis:

  • ~70 businesses are high-fit leads

  • With an expected 30% close rate, we project 21 new clients from the pilot alone

  • Average client LTV = $348K

• Revenue Potential from Pilot Alone:
  ~$7.3M in total potential lifetime value

• Long-Term Growth Target:
  Scaling to 25 strategic partners by end of 2026 would yield:

  • $25M–$30M in total lifetime value

  • Equivalent to $8M–$10M in annual recurring revenue (ARR)

​

SalesOS transforms cold outreach into a predictable, scalable sales funnel — positioning our agency to grow quickly, strategically, and with high-quality clients.

​​

​

What I Learned

• Python & AI Integration: Gained foundational proficiency in Python, and learned how to collaboratively develop code using ChatGPT as a pair programming assistant -- from writing functions to debugging and iterating based on test results.

• Prompt Engineering: Developed advanced prompting techniques to get consistent, reliable results from AI models. Learned how to tailor inputs to match the unique capabilities of each LLM and how to frame requests for maximum effectiveness.

• Overcoming AI Limitations: Built practical problem-solving skills by navigating around limitations in ChatGPT and other AI platforms. When the model failed, hallucinated, or misunderstood context, I learned how to diagnose the issue, adjust my approach, and achieve the desired outcome.

• Strategic Thinking & Sales Enablement: Gained hands-on insight into what drives successful digital sales. I learned how to position insights as opportunities, frame marketing gaps as areas for growth, and translate technical findings into persuasive, client-ready language.

• Digital Marketing Strategy: Through deep analysis of client websites, I developed a sharp understanding of the digital marketing landscape, including which web elements drive trust, conversion, visibility, and growth. I also gained exposure to how business models influence marketing decisions and performance.

• Cross-Disciplinary Problem Solving: This project demanded a unique combination of coding, AI interaction, business strategy, and storytelling. I learned how to navigate complexity, take ownership of ambiguity, and deliver a finished product that creates measurable business value.

​​

​

How SalesOS Powers Growth

SalesOS is not just a tool — it’s a growth engine:

• Scalable Audits: Enables us to audit hundreds of businesses per partner with minimal manual effort

• Qualified Leads: Filters low-fit leads and focuses our efforts on high-impact targets

• Pitch Personalization: Each audit becomes a data-driven, personalized proposal

• Faster Sales Cycles: Clarity = confidence = quicker decisions from prospects