Product 🚀
Floworks.ai
YC W23 · Early engineering
By Arush Bansal
Shipping the GTM MVP
I joined Floworks as one of four engineers on the early team, right as the company was turning YC W23 momentum into a product customers could actually buy. The mandate was not a slide deck — it was a go-to-market MVP: agentic automation for B2B sales that could run prospecting, outreach, and follow-up with minimal human babysitting.
Pipeline is the north star — everything else is overhead.
On floworks.ai today, the product is positioned around autonomous AI agents backed by ThorV2, their proprietary model, with claims of 43% higher accuracy versus other sales agents and plug-and-play onboarding. Customers report 6× ROI, 48%+ email open rates, and pipeline generated at scale — the kind of outcomes we were designing toward from day one.
What early engineering looked like
Three shifts that defined how we built:
- 01
Agents, not chatbots
Sales workflows are multi-step — research, personalize, send, follow up, log to CRM. We treated each lane as an agent with a job description, not a single prompt wrapper.
- 02
Trust over theatre
Hallucination kills outbound. Grounding agents in real contact data, intent signals, and verifiable research was non-negotiable for anything customer-facing.
- 03
Speed with compliance
Enterprise buyers care about SOC 2, GDPR, and deliverability. We shipped fast, but security and reputation had to keep pace with feature velocity.
The best sales tool is the one your team actually runs — without a human in the loop for every mundane step.
Suggested next read
Phoenix IITD
IIT Delhi · Product community