When HubSpot Veterans Build an AI Agent for the Thankless Work Nobody Wants to Do
Najmuzzaman Mohammad knows the tedium well. Years as a senior product manager at HubSpot taught him that some tasks in sales and marketing operations resist the promise of artificial intelligence not because they're complex, but because they're repetitive, high-volume, and deeply boring. Auditing thousands of outdated CRM entries. Scoring endless lead lists. Re-engaging deals that died months ago. These workflows break most AI assistants not through difficulty but through sheer scale.
So Mohammad left HubSpot and built Nex, an automation platform designed specifically for that grind. The startup launched publicly in early September 2026 with backing from HubSpot founder Dharmesh Shah and emerged from what appears to be Y Combinator's 2024 cohort (the company's materials reference a "Summer 2026 batch," an apparent typo or placeholder that went unfixed). On September 3, 2026, Nex claimed the top spot on Product Hunt with 365 upvotes, a metric Mohammad later described on LinkedIn as making it "the most viral launch" of his YC class.
Whether viral or merely visible, Nex represents a bet that general-purpose AI agents aren't suited for every job. Mohammad pitches the platform as "Claude Cowork for high-volume workflows," a framing that acknowledges the utility of tools like Anthropic's collaborative AI while arguing they falter when asked to process hundreds or thousands of similar tasks. "General-purpose agents choke on this," he wrote in a Product Hunt comment roughly 10 days post-launch. The reasoning overhead becomes prohibitive. Each decision point slows the system. At scale, the reliability collapses.
Nex sidesteps that by building agents tailored to specific go-to-market motions. Users describe a workflow goal. The platform constructs an agent that runs it end-to-end, claiming over 1,000 integrations according to the YC Launch post, though the homepage claims "100+" as of mid-September. Pre-built templates handle CRM hygiene, deals going silent, closed-lost re-engagement, lead scoring, and meeting prep. The company also released Annex, a browser designed for agent-driven GTM work, currently available for macOS with Windows and Linux versions in development.
The team carries HubSpot DNA throughout. Francisco Dias, Nex's CTO, joined HubSpot as its fourth CRM engineer before moving to the company's AI infrastructure team as its first product-focused engineer, according to the startup's YC profile. Douglas Oliveira, a senior backend engineer at Nex, spent five years at HubSpot before making the jump. Gustaf Alströmer, a veteran YC partner, worked with the team during their batch. YC lists five team members. LinkedIn indicates a range of "2–10 employees," though lists 11 profiles, a discrepancy that may reflect contractors, advisors, or recent hires.
Pricing sits at $49 monthly plus usage, with $100 in bundled AI credits and a seven-day trial. Integrations span HubSpot, Salesforce, Gmail, Outlook, Slack, WhatsApp Business API, and various data warehouses for structured and unstructured sources. The privacy policy, dated July 31, discloses AWS hosting, PostHog for analytics, Kafka for messaging, and optional data sharing with OpenAI, Gemini, and Anthropic when users request automations. Nex claims SOC 2 Type I compliance on its brand-facts page, though the verification link returned a 403 error when accessed during reporting.
Investors include Shah and Girish Mathrubootham (Freshworks founder), plus BoldCap Ventures, Together Fund, Leonis Capital, and Cur8 Capital. The company declined to share funding amounts. Together Fund portfolio lists 2025 as the investment date, potentially outdated or incorrect, and that page may not reflect current information or could contain its own dating error.
Product Hunt users browsing Nex's launch also viewed Clay, Trace, Attio, Copy.ai, and Tadata, a mix that suggests the platform competes across AI-assisted outbound, workflow automation, and CRM categories. Nex's positioning emphasizes real-time organizational context and compliance features it argues workflow generalists overlook.
The company published blog posts in February and March (presumably early 2024 or 2025, given the September launch) covering MCP integrations and context management. An open-source LLM wiki project released in April drew 260 points on Hacker News, a respectable showing in a community that tends toward technical skepticism.
When a Product Hunt user asked CTO Dias about custom research workflows, his response was terse and confident. "Try it! If the information is available it'll drop in a table and then you can action on it." Whether that ease holds at scale is the question Nex will now have to answer beyond the initial launch buzz. The tedious work, after all, never stops coming.
