Labelbox Raised $100M+ at a $1.25B Valuation: AI Agents That Improve on Their Own Are Now Business | inicIA
Direct answer
On October 14, 2026, Labelbox announced it raised more than $100 million in a Series E (equity plus a credit facility) at a $1.25 billion pre-money valuation. The round was led by Infinitum Partners, with participation from Wipro Ventures, Andreessen Horowitz, B Capital, Tau Ventures, and Snowpoint Ventures. The funding accelerates two products: Horizon, reinforcement-learning infrastructure for AI labs, and Recursion Managed Agents, managed agents for enterprises that improve with every run — opened to enterprise customers in early October 2026. The company reports 400+ customers (including 90% of the leading U.S. AI labs), reaching profitability in 2026 and growing 5x year over year.
What makes Recursion different
- It takes a goal, a trigger, and a rubric: the agent works from an objective, an event (schedule, alert, Slack message, webhook, API call), and a written definition of what good work looks like.
- A grader agent scores it line by line: every run is scored against the rubric with evidence; when something misses, the agent revises and retries.
- Three operating modes: multi-agent (a coordinator chains specialists), team (copies of the agent split work and check each other), and auto (the agent decides on its own when it needs specialists).
- Works across 100+ business applications and any MCP server; every tool is scoped (read-only, changes data, or destructive) and credentials live in a vault the model never sees.
- Pricing per agent session hour, with compute and model spend combined into one estimate.
The internal proof that convinced investors
Labelbox reported its own deployment: 51 agents running across 11 teams — engineering, reliability, security, research, evals, product, finance, sales, strategy, and procurement. Per the company, the surprise was not that the agents could do the work, but how quickly they got better at it: they cite 32% cost savings in internal code reviews after agent improvements. The lesson for your business: the value is not in the smartest model, but in the system that measures, grades, and improves every run.
What to demand from your agent provider (the checklist this round leaves behind)
- Rubric-based evaluation, not promises: ask that every run be scored against written criteria, with evidence.
- Per-tool permissions: the agent must operate with exact scopes (read, write, destructive), never with full access.
- Isolated sessions and zero data retention: serious infrastructure already offers this as standard, with certifications (SOC 2 Type II, GDPR, CCPA).
- Price per outcome, not per token: per-session-hour pricing with everything included is easier to budget than per-token billing.
Related links
- Sierra and Meta Launched the Personal Agent Protocol: Your Business Needs a Front Door for AI Agents
- eXp Realty Launched "MCP First, AI of Choice": Your Favorite AI, Connected to Your Business Data
- What to Do When Your AI Agent Makes a Mistake
Frequently asked questions
What did Labelbox announce on October 14, 2026?
A Series E of more than $100 million (equity plus credit facility) at a $1.25 billion pre-money valuation, led by Infinitum Partners, to accelerate Horizon (reinforcement learning) and Recursion Managed Agents (enterprise agents that improve with every run).
What are Recursion Managed Agents?
Agents that take a goal, a trigger, and a quality rubric; a grader agent scores every run against the rubric and the agent corrects what fails. They run across 100+ business applications and MCP servers, with per-tool permissions and isolated sessions.
Why does this matter to a small or mid-sized business?
Because it sets the standard you should demand: measurable evaluation, exact permissions, and per-session pricing. Production reliability — not the newest model — is what the market is rewarding with capital.
Is Labelbox profitable?
Per its own announcement, the company reached profitability in 2026, is growing 5x year over year, and serves 400+ customers, including 90% of the leading U.S. AI labs.
Take the next step
If you want AI agents that work with clear objectives, exact permissions, and real measurement — not promises — talk to inicIA on WhatsApp: https://wa.me/17866120930
Chat on WhatsAppBy inicIA
