YC Root AccessBuilding AI Agents for Everyone
CHAPTERS
- 0:05 – 0:52
Gumloop’s promise: an agent builder every team can use (with IT control)
Gustaf introduces Gumloop and frames it as the “holy grail” many teams want: enterprise-grade agents that non-technical employees can build and share. Max and Rahul position Gumloop as a collaborative agent hub that still lets IT govern data, permissions, and risk.
- •Agent builder used by companies like Shopify, Gusto, and Instacart
- •Designed for broad employee adoption, not just engineers
- •Agents can be shared across the business, not siloed per team
- •IT maintains control over data, access, and governance
- 0:52 – 1:33
From workflow automation to true agents: how the product evolved
The founders explain that Gumloop started as a visual workflow automation product because early models weren’t reliable enough. As models improved, adding agents created a “hockey stick” moment in usage and shifted the product toward end-to-end automation.
- •Originally launched as “Agent Hub,” but was a workflow automation builder at first
- •Early approach: step-by-step workflows with minimal AI for reliability and cost
- •Model improvements enabled end-to-end agent automation
- •Same mission throughout: automate tedious work for real teams
- 1:33 – 2:03
How Gumloop works day-to-day: building, connecting, and deploying agents
Rahul describes the product UX: users create agents in a chat-like interface and add connectors to the apps they already use. Gumloop brings agents into workplaces like Slack and Microsoft Teams, and also supports deeper programmatic use via SDK/API and many integrations.
- •Chat-style agent creation plus app connectors
- •Agents can be used in-product or embedded in Slack/Teams
- •Robust SDK/API mirrors UI capabilities
- •Hundreds of integrations exposed via MCP servers for internal tooling
- 2:03 – 4:18
Real-world agent examples: sales copilots and “the gum between your tools”
Max shares common patterns, especially team-specific agents such as sales copilots that connect CRM, call recordings, enrichment, and web search. Gumloop often acts as the connective layer across tools—sometimes replacing vertical SaaS, sometimes orchestrating them.
- •Sales agents handle call analysis, next steps, CRM hygiene, deduping, enrichment
- •Used continuously in Slack/email as a “coworker”
- •Positions Gumloop as a connector/orchestrator across existing tools
- •Can replace some point solutions when an agent becomes the workflow
- 4:18 – 8:09
Early traction and the enterprise pivot: from indie use cases to Instacart
They recount early PLG usage (including quirky workflows like transcribing D&D sessions) before recognizing enterprise value. The pivot was driven by seeing deeper ROI in serious business workflows and by learning enterprise sales through an internal champion at Instacart.
- •Initial wave: individuals/indie hackers automating tedious tasks
- •Key signal: enterprise customers pay more and have higher-value problems
- •Instacart entry via a power user champion who evangelized internally
- •Hiring the champion taught them procurement, pricing, and champion dynamics
- 8:09 – 9:59
Winning enterprise customers: requirements, security checklists, and shipping fast
Gustaf probes what enterprises demand, and the founders list the “boring but necessary” capabilities: SSO, SCIM, RBAC, audit logs, hosting, and data-lake export. They emphasize learning requirements through many customer conversations and building trust through responsiveness and partnership.
- •Enterprise needs: RBAC, SCIM/SAML, audit logging, data controls, customer-managed keys/cloud
- •Complex permissioning across employee types and outsourced teams
- •Feature requests never end; governance is prerequisite to broad rollout
- •Differentiator: fast iteration and acting like a true partner
- 9:59 – 15:00
Driving adoption: templates, workshops, and meeting users where they work
They discuss the adoption challenge of a flexible platform: users don’t know where to start. Gumloop addressed this with a large template gallery, use-case pages, workshops, and embedding agents into existing work surfaces (especially Slack), plus collaborative “multiplayer” sharing of agent usage.
- •Template gallery and use-case pages help users get started quickly
- •Hands-on workshops: identify tasks to automate and build them live
- •Biggest unlock: put agents where employees already spend time
- •Collaboration: shared agents and public examples accelerate adoption
- 15:00 – 16:19
How Gumloop runs on Gumloop: Heimdall, Sales Loop, and automated CS/education
Max and Rahul describe internal “breakthrough agents,” starting with Heimdall, which connects to all company data sources and answers complex queries in Slack. They explain how internal agents expanded into sales automation, customer success health monitoring, and automating education cohorts end-to-end.
- •Heimdall: company-wide data agent accessible across Slack channels
- •Sales Loop: automates sales operations so reps focus on customers
- •Success Loop: tracks customer health scores and alerts on usage changes
- •Education cohorts automated from signup to follow-ups to video editing/clipping
- 16:19 – 20:53
Shopify rollout, pricing rethink, and what PMF felt like
They describe PMF signals: users tolerating an “awful” early product, power users spending 8–10 hours/day building automations, and the day Shopify rolled Gumloop to thousands of employees. They also explain evolving pricing from low monthly plans to transparent usage-based pricing aligned with “build” infrastructure tooling.
- •PMF moments: users investing hours despite rough UX; extreme power usage
- •Shopify launch: rapid adoption with far less handholding at large scale
- •Pricing evolution: $20/$40 → multiple doublings → credits → transparent cost-plus orchestration fee
- •No per-seat pricing to avoid discouraging broad experimentation
- 20:53 – 34:15
Fundraising, team size realities, and the future: employee-led automation with IT approval
Max outlines fundraising lessons: demos win, traction drives investor interest, and later rounds were preempted without a formal raise. They discuss the limits of staying ultra-lean in enterprise, then expand into the broader vision: the people who know the work should automate it—enabled by IT governance, model/provider neutrality, and controllable long-running agents.
- •Investor strategy: show the product; later rounds were preempted without decks
- •Enterprise forces hiring: sales/support/security demands require more people than expected
- •Vision: task owners (not consultants/eng) should build automations themselves
- •“Switzerland” positioning: model/provider agnostic; customers own agents, traces, and deployment
- •IT is the key decision-maker for secure pilots, data access, and rollout controls
