Most early-stage founders think they need a marketing team, a sales team, and a growth hire before they can scale. Bhanu Teja Pachipulusu and Sai Dheeraj Pachipulusu built SiteGPT to $500K+ revenue with zero of those.
Two brothers. Both coders. No marketing background. They handle demos, support tickets, blog posts, and product development themselves. SiteGPT now serves 1.8 million websites.
Founded by Bhanu Teja Pachipulusu (CEO) and Sai Dheeraj Pachipulusu (Co-founder).
This Week's Breakdown: SiteGPT
What they do: AI-powered customer support chatbot platform. You paste your website URL, and SiteGPT trains a chatbot on all your content: docs, blogs, product pages, FAQs. The bot answers visitor questions 24/7 in 95+ languages.
The numbers: 1.8M+ websites. $500K+ total revenue. $190K ARR. $15.8K MRR. 130+ paying businesses at ~$100/month average. Customer lifetime value of $1,700-$1,800. Fully bootstrapped with zero outside funding.
How They Built It
| Layer | Choice | Why |
|---|---|---|
| Frontend | Next.js (React) | SSR for SEO on public pages and free tools, SPA for dashboard |
| Backend | Node.js | Unified JS stack, fast iteration for a 2-person team |
| LLM Infrastructure | Portkey | 13+ LLM models in production, automated failover, rate limit handling |
| Vector Database | Pinecone Serverless | Migrated Nov 2024 for scalable semantic search across customer data |
| ORM / Database | Prisma (migrated May 2025) | Type-safe queries, simplified migrations for fast-moving codebase |
| Workflow Automation | n8n | Handles internal automation, prompt orchestration via Portkey |
| Auth | Email + OAuth (Google, GitHub) | Standard social login for SaaS onboarding |
| Payments | Stripe (inferred) | Subscription billing across 4 tiers ($39-$259/mo + Enterprise) |
| Hosting | Cloud (AWS/GCP inferred) | Elastic compute for LLM inference at scale |
| Integrations | Slack, Crisp, Intercom, Zendesk, Messenger, Freshdesk, Zapier | Multi-channel deployment across 7+ chat platforms |
| Data Sources | Google Drive, Dropbox, OneDrive, SharePoint, Notion, Confluence | 10+ connectors for training data ingestion |
Stack inferred from Portkey case study, changelog, product docs, and integrations page. SiteGPT hasn't published full technical details.
The Real Story
Bhanu Teja Pachipulusu was a software engineer for eight months before deciding it wasn't for him. He moved back to his parents' house to cut expenses and started building products. His first bet, Feather (a Notion-to-blog tool), grew to $6,000 MRR before he sold it for $250,000 in 2022. That exit gave him both capital and confidence.
In early 2023, AI tools flooded his Twitter feed. Bhanu felt like he was missing out. His approach: spend a weekend learning by building something useful. Feather had 100 customers who all ran blogs. What if he built a chatbot that could answer questions about their content? As he built it, a bigger idea clicked: why limit this to 100 people?
SiteGPT went from idea to live product in two weeks. Bhanu posted it on Hacker News' Show HN and hit the front page. Two days later, it hit #1 Product of the Day on Product Hunt. First month: $10,000 MRR.
Then reality hit. Almost 50% of customers churned after month one. Revenue dropped to $5K.
Bhanu didn't panic. He saw the churn as a filter. The launch hype brought curiosity seekers. The people who stayed had real problems. He and his brother Dheeraj focused on building for them. No investors. No advisory board. Two coders fixing customer issues within minutes, shipping features daily, and figuring out growth by doing.
Today, SiteGPT processes millions of requests and billions of tokens across 13+ LLM providers. They've built a second product, SourceSync.ai, a RAG-as-a-Service API, spinning out the core retrieval engine into infrastructure that other developers can use. The Portkey case study reveals an operation running at enterprise scale: versioned prompt management (100+ prompts), automated failover across providers, and end-to-end observability on every LLM call.
The Marketing Playbook: Engineering as Distribution
This is where SiteGPT's story gets tactical.
Bhanu built 50+ free tools: PDF-to-Markdown converters, AI Chatbot Name Generators, Sitemap Validators, FAQ Generators, Blog Title Generators. Each tool targets a specific keyword. Each has a CTA that leads back to SiteGPT.
The system works like this:
- Use Ahrefs Keywords Explorer to find keywords with 1,000+ monthly search volume
- Filter for keyword difficulty under 10 (almost no competition)
- Build each tool in under 5 minutes using Cursor AI with existing tools as templates
- Deploy with a clear call-to-action connecting back to SiteGPT
The results: 50,000 monthly visitors. 90% from organic Google traffic. 200 leads per month. 60 trials. 25-40% trial-to-paid conversion. Total cost: $0 in marketing spend.
SiteGPT spends nothing on paid ads, nothing on content agencies, nothing on SDRs. The free tools are the marketing team. When Bhanu wants to target a new keyword, he opens Cursor, points it at an existing tool as a template, and ships a new one in five minutes. This is engineering as marketing at its most literal.
How SiteGPT Makes Money
| Plan | Price/mo | Messages | Web Pages | Key Features |
|---|---|---|---|---|
| Starter | $39 | 4,000 | 200 | 1 chatbot, 10 file uploads |
| Growth | $79 | 10,000 | 2,000 | 2 chatbots, integrations, API access |
| Scale | $259 | 40,000 | 10,000 | 5 chatbots, webhooks, 10 team members |
| Enterprise | Custom | Custom | 100,000 | 1,000 chatbots, priority support |
The white-label offering is the growth vector. Agencies and industry partners can resell SiteGPT under their own brand. Bhanu's stated goal: $1M revenue through white-label partnerships rather than expanding vertically into any single industry.
The Global Landscape: AI Customer Support Chatbots
SiteGPT competes in a market valued at $12.58 billion in 2024, projected to hit $47.82 billion by 2030 (25.8% CAGR). The space spans from billion-dollar incumbents to bootstrapped indie products. The competitive dynamics look radically different depending on what tier you're in.
The Big Players (Enterprise Scale)
| Company | Where | Funding / Scale | Revenue | What Makes Them Different |
|---|---|---|---|---|
| Zendesk | US | Acquired for $10.2B (PE) | $1.93B (2024) | Full customer service suite, AI add-on, enterprise standard |
| Intercom | US (Ireland-founded) | $241M raised, ~$1.3B valuation | ~$323M | Fin AI agent, per-resolution pricing, 1,600 employees |
| Ada | Canada | $202M raised, $1.2B valuation | $70.6M (2024) | Automation-first, 5,000 customers, outcome-based pricing |
These companies have hundreds of employees, dedicated AI research teams, and enterprise sales motions. SiteGPT is not competing with them head-on. The gap is structural: Zendesk's annual revenue is 10,000x SiteGPT's.
The Mid Players (Funded, Growing)
| Company | Where | Funding | Revenue / Scale | Target |
|---|---|---|---|---|
| Tidio | Poland | $27M (Series B) | ~$30.8M ARR, 193 employees | E-commerce SMBs, Shopify/WooCommerce |
| Drift | US | $100M+ raised | Enterprise ($2,500+/mo) | Conversational marketing, B2B |
| Chatbase | US | $0 (bootstrapped) | $8M ARR, 18 people | AI-only chatbot, developer-focused |
| CustomGPT | US | Undisclosed | Serves MIT, Adobe, Dropbox | Anti-hallucination, 1,400+ file formats |
Chatbase is the closest comp. Also bootstrapped. Also started as a side project. But Chatbase is at $8M ARR with 10,000+ paid customers. That's 42x SiteGPT's ARR. Both prove the model works. One is further along.
The Small Players (Indie / Early)
| Company | Where | Funding | Revenue / Scale | Target |
|---|---|---|---|---|
| SiteGPT | India | $0 (bootstrapped) | $190K ARR, 2 people | SMBs, agencies (white-label) |
| Botsonic | US | Part of Writesonic | Undisclosed | ChatGPT-4 powered, easy setup |
| FastBots.ai | US | Undisclosed | Undisclosed | SMBs, simple deployment |
| Crisp | France | Bootstrapped | Undisclosed | All-in-one inbox for SMBs |
The Pattern: Two-Person Teams vs. 1,600-Employee Companies
The gap between SiteGPT (2 people, $190K ARR) and Intercom (1,600 people, $323M revenue) is 1,700x in revenue and 800x in headcount. What's changed is that the floor for building a competitive product has dropped to near zero. LLM APIs, serverless vector databases, and AI coding tools mean two developers can ship features that took teams of twenty three years ago.
The question isn't whether SiteGPT can replace Intercom. It can't. The question is whether there's a large enough market of businesses that need 80% of Intercom's functionality at 5% of the price.
Built for the AI Age: What SiteGPT Got Right
SiteGPT is an AI-age company in every dimension, not just in what it sells, but in how it operates.
Product: Trained on customer content using RAG (Retrieval Augmented Generation) across 13+ LLM providers via Portkey. Pinecone Serverless for vector search. SourceSync.ai as a separate RAG-as-a-Service API product. The entire product is LLM-native.
Development: Bhanu builds new free tools in under 5 minutes using Cursor AI. Existing tools serve as templates. The codebase is a Cursor-first workflow where AI writes first drafts and humans review. Prisma migration, n8n automation, Portkey prompt versioning: every layer is optimized for speed with a skeleton crew.
Marketing/SEO: 50+ free tools targeting low-competition keywords. No content agency. No paid ads. The tools themselves are the acquisition engine, and they're built with AI in minutes.
CRM / Support: Bhanu and Dheeraj handle all support themselves. They fix issues in minutes. This direct feedback loop shapes the product faster than any support team reporting to a product team through a ticketing system.
Distribution: White-label API lets agencies resell SiteGPT. SourceSync.ai turns the core tech into developer infrastructure. Both are leverage plays for a 2-person team.
The OpenClaw Effect: AI Agents as a Force Multiplier
OpenClaw, the open-source AI agent framework that became the fastest-growing GitHub repo in history (157K+ stars in 60 days, 200K+ total), represents a shift that directly benefits companies built like SiteGPT.
The core idea: always-on AI agents that can browse the web, manage email, automate workflows, run code, and integrate with tools like Zendesk, Intercom, Slack, and CRMs. OpenClaw's ClawHub marketplace has 5,700+ community-built skills covering SEO automation, customer support triage, competitor monitoring, content marketing, and analytics.
Why this matters for SiteGPT's model:
A 2-person team running marketing, sales, support, development, and product simultaneously is already running at the edge of what's humanly possible. OpenClaw-style agent frameworks represent the next lever. Founders using similar setups report saving 36 hours per week and ~$93K annually by deploying AI agents for:
- SEO Content Engine skills that publish at agency scale without hiring
- Customer support agent skills that auto-resolve tickets using knowledge bases (exactly what SiteGPT sells, now applied internally)
- Browser automation skills for competitor monitoring and market research
- Google Workspace skills that unify Gmail, Calendar, Drive, and Docs into one AI-managed interface
SiteGPT and OpenClaw also represent two sides of the same trend. SiteGPT is a chatbot you deploy on your website. OpenClaw is an agent you run on your machine. Both use RAG, LLMs, and tool integrations. Both are replacing functions that used to require full-time hires. As these frameworks mature, the minimum viable team for a competitive SaaS product keeps shrinking.
The bootstrapped founders of 2023 used AI coding assistants to ship faster. The bootstrapped founders of 2026 use AI agents to operate faster: writing content, monitoring competitors, resolving support tickets, and managing workflows, all autonomously.
Quick Hits
Chatbase hit $8M ARR with zero funding and an 18-person team. Proof that bootstrapped AI chatbot products can scale. Their 92% retention rate suggests sticky product-market fit. Read more →
Intercom's Fin AI now resolves 50%+ of customer queries automatically. Per-resolution pricing aligns their revenue with AI performance. When the bot gets better, Intercom makes more money. Read more →
Ada crossed $70.6M in revenue powering support for Verizon, Shopify, and Canva. Outcome-based pricing ($0.99-$1.50 per resolved interaction) is becoming the enterprise standard. Read more →
OpenClaw hit 200K+ GitHub stars and 5,700+ community skills in under 60 days, then faced a security crisis: 341 malicious skills (11.3% of its marketplace) were caught stealing credentials. The viral growth came with a trust tax. Read more →
How We Got Here: AI Customer Support Timeline
| Year | What Happened | Key Event |
|---|---|---|
| 2011 | Intercom launches | Conversational customer relationship platform starts the modern category |
| 2013 | Tidio founded | Early chatbot play for e-commerce, long before AI hype |
| 2016 | Ada launches | Automation-first customer support, pre-LLM era |
| 2022 | ChatGPT launches (Nov) | Everything changes. LLMs make conversational AI accessible to anyone |
| 2023 | SiteGPT, Chatbase launch | Weekend projects become real businesses. SiteGPT hits #1 Product Hunt |
| 2024 | AI chatbot market hits $12.58B | Pinecone, Portkey, and RAG infrastructure matures |
| 2025 | Chatbase at $8M ARR, SiteGPT at $190K ARR | Bootstrapped AI chatbots prove the model at different scales |
| 2026 | OpenClaw + AI agents explode | Agent frameworks let 2-person teams operate like 20-person companies |
$12.58B market in 2024, projected $47.82B by 2030. The customer support AI space is growing at 25.8% CAGR. Every company with a website is a potential customer.
The Pattern You Can Steal
SiteGPT's playbook has three parts, and all three are replicable:
1. Engineering as marketing. Build 50+ free tools targeting low-competition keywords (KD < 10, volume > 1,000). Use AI to build each one in minutes. Each tool has a CTA back to your product. This replaces a marketing team for $0.
2. White-label as distribution. Instead of scaling your own sales team, let agencies and partners resell your product under their brand. You build once, they distribute to their networks. Leverage without headcount.
3. Spin out your infrastructure. SiteGPT built SourceSync.ai by packaging their RAG pipeline as a developer API. If your core tech is good enough for your product, it's good enough for other developers. This is the same B2B infrastructure play Beatoven.ai made with maestro on fal.ai (Issue #002).
The meta-pattern: AI-age companies don't just use AI in their product. They use AI to build faster (Cursor), market cheaper (free tools), and operate leaner (agent frameworks). The stack is the strategy.
What We're Watching
The Chatbase gap is the benchmark. Chatbase is at $8M ARR. SiteGPT is at $190K. Both bootstrapped, similar products, similar customer profiles. The 42x gap is the distance SiteGPT needs to close. White-label partnerships and SourceSync.ai are the two bets to close it.
India's SMB market is underserved. 63 million SMBs in India. Most have websites but no AI-powered support. A Hyderabad-based team with local context and pricing flexibility has a distribution advantage that US-based competitors don't.
AI agents will reshape operations. When agent frameworks like OpenClaw mature past the security issues, a 2-person team with 10 AI agents running SEO, support, monitoring, and content will operate at the capacity of a 15-person company. SiteGPT is already halfway there.
Ship it.
— The FounderSpec Team