Zero to $6M ARR with AI SDRs
with Gaurav Bhattacharya, GVA.ai
Zero to $6M ARR with AI SDRs
Show Notes
Gaurav Bhattacharya is a three-time founder and the CEO of GVA.ai, an AI-powered sales development platform that went from zero to $6M ARR in under nine months. The company's origin story involves a Series A startup, a near-total layoff, an internal tool that outperformed every vendor on the market, a cease-and-desist from Disney, and a pivot to a new name - all before reaching product-market fit. In this conversation Gaurav breaks down the exact mechanics behind AI SDR deliverability, his waterfall enrichment stack, outcome-based pricing, and why the founders who win the next decade will be the ones who use AI to find their own customers.
What Is an AI SDR - and Why Does Deliverability Change Everything?
A traditional Sales Development Representative (SDR) spends most of their day finding leads, researching prospects, writing personalized outreach, and booking meetings. An AI SDR automates the entire top-of-funnel: it finds leads from dozens of sources, enriches their contact data through a waterfall of providers, writes hyper-personalized emails and LinkedIn messages, triages replies in the inbox, and syncs everything to the CRM. The remaining job for the human rep is to review drafts and hit send - roughly five to ten minutes per day. The critical insight Gaurav discovered is that email deliverability collapses the moment emails are sent via API, because Google and Microsoft tag machine-generated sends and route them to spam. GVA solves this by creating drafts in the user's actual inbox and letting the human send them. That single workflow change produced a 90% improvement in deliverability without any domain rotation, warm-up sequences, or infrastructure tricks.
5 Frameworks for AI-Driven Sales & Startup Pivots
1. The Draft-Send Deliverability Method
- API-sent emails are tagged by Google and Microsoft as machine-generated - this kills deliverability.
- Instead, AI writes the email and places it as a draft in the user's real inbox; the human reviews and clicks send.
- This bypasses the API tag entirely because the send originates from the user's authenticated session.
- Result: 90% better open and reply rates, zero domain rotation, no warm-up infrastructure needed.
- Human time investment: five to ten minutes per day - just enough to maintain authenticity at scale.
2. The Waterfall Enrichment Stack
- No single data provider has complete or accurate contact information - coverage gaps are the rule, not the exception.
- Waterfall enrichment queries providers sequentially: Hunter → Apollo → Lusha → Crunchbase → and others.
- Each provider fills in what the previous one missed, producing a composite record with far higher completeness.
- The stack is query-cost-optimized: cheaper or higher-coverage providers run first to minimize spend on fallbacks.
- This is now table stakes for any serious outbound motion - single-provider enrichment leaves 30-50% of leads under-researched.
3. Outcome-Based Pricing for AI Products
- GVA charges credits only when a sequence runs fully or a positive reply is received - not for activity.
- This aligns incentives completely: if the AI does not produce a result, the customer does not pay for that attempt.
- Individual tiers at $20/$50/$100/month serve SMB and solopreneurs; enterprise contracts run $6K-$60K/year.
- The PLG free tier creates top-of-funnel volume (150-250 organic signups per day) that feeds the enterprise motion.
- Outcome pricing also forces the product team to obsess over what constitutes a "positive response" - a forcing function for quality.
4. The Internal-Tool-to-Product Pivot
- Involve AI (Gaurav's prior company) raised a Series A, grew to 50 people, and never found PMF - leading to a 90% layoff.
- The remaining small team built an internal AI outbound tool called R2D2 to survive and source their own pipeline.
- R2D2 worked so well that prospects began asking to buy it - the product found them before they went looking for a market.
- Disney sent a C&D over the R2D2 name within 10 days of going public - the company became GVA.ai (echoing how Slack was renamed from Glitch).
- The lesson: the best B2B products are often tools the team built to solve their own most painful problem at their lowest moment.
5. The PLG → Enterprise Sales Flywheel
- A free tier with no credit card required generates organic signups at scale - GVA sees 150-250 new users per day.
- Free users become power users, generate case studies, and seed word-of-mouth within their networks and companies.
- The most engaged free users upgrade to paid individual plans; the highest-value accounts trigger enterprise conversations.
- GVA uses GVA to run this entire motion - 100% of the company's own pipeline is sourced through the product itself.
- Using your own product to grow your product is the strongest possible proof point in a sales conversation with any prospect.
Founder Experiment: Launch Your AI SDR in 5 Days
Step 1 - Define your ICP with brutal specificity. Write down the exact title, company size, industry, and trigger event (funding round, new hire, product launch) that makes someone a "ready-to-buy" prospect. Vague ICPs produce vague outreach - specificity is what allows AI to write personalized messages instead of generic blasts.
Step 2 - Build your waterfall enrichment stack. Sign up for a free tier on Hunter.io and Apollo.io. Export your first 50 target accounts from LinkedIn Sales Navigator. Run each account through Hunter first, then Apollo for any gaps. Note which provider filled which fields - this is your live data on coverage quality for your specific ICP.
Step 3 - Write three personalized email templates using the draft-send model. For each template, identify one specific trigger (a recent blog post, a LinkedIn update, a funding announcement) that a prospect in your ICP commonly produces. Write the email as if you read that trigger. Then configure your AI tool to pull that trigger field into the draft - you review, you send.
Step 4 - Launch a 20-contact test sequence and measure deliverability. Send your first 20 emails as human-triggered drafts (not via API auto-send). Track open rate and reply rate. If open rate is below 40%, your subject line or sender domain has a deliverability problem - fix it before scaling. If open rate is high but reply rate is low, the message body needs work.
Step 5 - Use your own product to sell your product. Whatever you sell, find a way to use it to generate at least one inbound lead before you invest in any other channel. Gaurav's rule: if you would not trust the tool enough to stake your own pipeline on it, your prospects should not trust it either. Dogfooding is both a product discipline and a sales proof point.
Glossary
Tools & Resources Mentioned
Q&A
How did Gaurav go from zero to $6M ARR in under nine months?
GVA launched in January or February and crossed $6M ARR by the time of this recording. The growth came from three compounding factors: a PLG free tier that generates 150-250 organic signups per day, an enterprise motion funded by Alt Capital (Jack Altman), and the fact that GVA uses GVA to source 100% of its own pipeline - making every sales conversation a live product demo. The product had also been validated internally at Gaurav's prior company before it was ever sold to a customer, which meant it shipped with real-world proof rather than a pitch deck hypothesis.
What is the core deliverability problem with AI email outreach, and how does GVA solve it?
When emails are sent programmatically via API - the default for most AI outreach tools - Google and Microsoft add metadata tags that classify the message as machine-generated. Those tags dramatically increase spam routing and suppress open rates. GVA's solution is architecturally simple: the AI writes the email and saves it as a draft in the user's real inbox. The user reviews the draft and hits send. Because the send originates from an authenticated human session rather than an API call, there is no tag and no deliverability penalty. This produced a 90% improvement in deliverability for GVA users with no additional infrastructure.
What is waterfall enrichment and why does it matter for AI outbound?
Waterfall enrichment is the practice of querying multiple contact data providers in sequence rather than relying on any single source. No provider has complete coverage - Hunter may have an email where Apollo does not, Lusha may have a direct dial where neither does. GVA's waterfall runs providers in order of coverage and cost efficiency, filling gaps at each step to produce a composite lead record that is materially more complete than any single provider could deliver. For AI outreach to be genuinely personalized, it needs accurate and complete contact data as an input - garbage in, generic output. Waterfall enrichment is the data infrastructure that makes hyper-personalization possible at scale.
How did GVA.ai get its name, and what is the origin story of the product?
Gaurav's prior company, Involve AI, raised a Series A and grew to 50 employees but could not find product-market fit. He made the painful decision to lay off 90% of the team. The remaining small group built an internal AI outbound tool they called R2D2 to survive - to find their own customers without a large sales team. R2D2 worked so well that prospects started asking to buy it. When the company went public with the R2D2 name, Disney sent a cease-and-desist within 10 days. They renamed the product GVA.ai. Gaurav draws a parallel to Slack, which was originally named Glitch - sometimes a forced rename produces a better identity than the original.
What did Jack Altman mean when he said 'be as close to AI as possible or as far away as possible'?
Jack Altman's framing is that the AI disruption is bifurcating markets. Products and businesses that sit adjacent to AI - neither fully AI-native nor safely differentiated from it - face the highest replacement risk. Companies that are fully AI-native (like GVA) are building the picks and shovels of the new workflow stack. Companies that are deeply human, relationship-driven, or embodied (like certain forms of therapy, physical services, or creative work) are insulated because AI cannot replicate them. The dangerous middle is any business that partially relies on cognitive tasks AI can now perform but is not built around AI as a core competency - those businesses are most exposed.
How does GVA handle inbox triage and meeting prep, and why does Gaurav call this the most valuable part of the product?
Beyond outbound, GVA reads the user's inbox and triages replies - flagging positive responses, objections, and out-of-office messages, and drafting response options for the rep to review and send. For meetings, GVA delivers a morning briefing that reviews the day's calendar and prepares the user on each attendee: recent LinkedIn activity, news mentions, shared connections, and relevant talking points. Gaurav calls this the most valuable feature for enterprise users because it makes every rep look like they spent an hour preparing for each call - without the hour. The cognitive overhead of context-switching between prospects is one of the most underrated productivity drains in sales.
What is Sam Altman's quote about sales, and how does Gaurav apply it to founders?
Sam Altman said: 'At some point, every job becomes a sales job.' Gaurav interprets this as meaning that the ability to articulate value, generate belief, and move people toward a decision is foundational to every function in a company - not just the sales team. For founders specifically, this means that even technical or product-focused founders must develop sales fluency to hire great people (selling the mission), raise capital (selling the vision), land early customers (selling the product), and build partnerships (selling mutual benefit). Gaurav's view is that the founders who use AI to eliminate the mechanical parts of selling - lead research, email drafting, meeting prep - free themselves to focus on the irreplaceable human part: building trust and conviction in a live conversation.
What pricing model does GVA use and why is outcome-based pricing the right structure for AI products?
GVA offers a free PLG tier, individual paid plans at $20/$50/$100 per month, and enterprise contracts ranging from $6K to $60K per year. Credits are consumed only on positive replies or fully completed sequences - not on activity. Gaurav argues that outcome-based pricing is the correct model for AI because it forces the vendor to care about real results rather than activity metrics, and it builds trust with buyers who are skeptical of AI ROI claims. If your AI produces a meeting, you get paid. If it sends 500 emails to the wrong people, you do not. That alignment makes the product roadmap self-correcting: the team is incentivized to optimize for outcomes, not outputs.
What advice does Gaurav give to founders thinking about building with AI today?
Gaurav's core advice is to use AI on your own business first - specifically on whatever problem is most painful and most manual in your current workflow. For GVA, that was outbound sales. The act of building for yourself produces a level of product depth and authentic empathy that cannot be faked in a customer conversation. He also emphasizes that the best founders are not the ones who integrate every AI tool, but the ones who identify the single workflow where AI creates asymmetric leverage for their specific business and go all-in there first. Finally, he echoes Jack Altman's positioning: if your business is in the middle - partially automated, partially human - do the work to understand which side of the line you should be on and move decisively toward it.