We've raised $150M in Series C funding at a $1.2B valuation, led by Prysm Capital and co-led by @eurazeo.
We're expanding across supply chain, insurance, energy, telecom, and airlines - building toward Enterprise Superintelligence.
Thank you to our team, customers, and investors. We’re just getting started.
Full story in the thread below 👇
As more teams within an enterprise build on HappyRobot, it’s important that access matches the job.
A claims ops team should be able to edit workflows and update table rows, but not publish to production or change schemas. An analytics team should be able to see every workflow and run, without editing any of them. A platform engineering team should own production: publishing, schemas and who gets access to what.
Role-based access control splits this into two decisions: a role sets what someone can do, and a scope sets where they can do it.
Roles map to your existing SSO groups, so permissions follow your org structure. Create custom roles for each team, scope them to the workflows and data they own, and keep production changes with the people accountable for them.
Access control is one of many ways we keep the HappyRobot platform secure and reliable. Learn more in the comments.
Some of @Habitissimo's customers have asked, in writing, to be served by José. José is an AI agent.
Habitissimo’s team started with one agent checking leads. Then, working side by side with our engineers, they built a whole team. José now answers every customer service call and resolves about 90% of requests on its own.
Read the full case study in the thread 👇
HappyRobot is now officially in Singapore! 🇸🇬
After a year of deploying AI agents with customers here, our team now has a home base to grow across Asia-Pacific. Want to join us? Careers link in the thread. 👇
Adding more tools to an agent works until the instructions for different situations start competing. Every new use case makes the prompt longer, and every change means retesting the whole thing.
With agent orchestration on HappyRobot, an orchestrator agent can triage each conversation and route it to a specialized sub-agent, each with its own prompt, tools and knowledge. When the conversation changes, it takes control back and re-triages.
A customer calls to move a delivery, then mentions their last invoice looks wrong. The orchestrator hands the reschedule to a scheduling agent, takes control back when the second request comes up, and routes the invoice question to a billing agent connected over MCP.
To the customer, it’s a single seamless conversation with an agent; behind the scenes, it’s a synchronized team of specialist agents.
That’s how the operation scales. A new use case becomes a new sub-agent, not a longer prompt. Each agent stays narrow and is tested on its own, so changing one doesn’t mean retesting the rest. Handoffs can cross channels, from voice to text, and external agents plug in over MCP.
More on agent orchestration in comments.
¡Hola, Buenos Aires! Our Argentina office is officially open. 🇦🇷
After Mexico City and São Paulo, we keep growing across Latin America, and we couldn’t be more excited to build a team here.
We’re hiring for our deployments team. Careers link in the thread. 👇🏼
An agent on a live call only knows what it knew when the call started. But operations don’t pause for the conversation to finish.
Signals close that gap. A weather service, your scheduling system, or any internal tool publishes an event once, and it reaches every live agent session that needs it, across voice and text.
A storm closes a highway, and the agent tells drivers on that route mid-call. An appointment is cancelled, and the agent offers a new slot instead of confirming the old one. A rate changes mid-negotiation, and the agent works from the new number.
Each update reaches only the conversations it affects. The agent either adjusts the answer it was about to give or holds the update until the next turn, re-running tool calls on the new information when it needs to.
Agents that work from what’s true right now, not what was true when the call started.
See how Signals work - link in comments.
Olá, São Paulo! 🇧🇷 HappyRobot’s newest office is officially open.
Want to help build HappyRobot in Brazil? We’re hiring our first local team members. Careers link in the thread. 👇
Grupo Flecha Amarilla has been moving people across central-western Mexico since 1932. It runs five intercity bus brands, @Primera_Plus among them, and the buses run around the clock. The contact center used to close at midnight, and about 200 calls a day went unanswered.
Now MarIA answers it — an AI agent their own team built with ours, named for María with IA, the Spanish initials for artificial intelligence. She sits inside the existing phone menu rather than in front of it, so callers press 1 to reach her. They have to choose her.
Around 20% do, roughly 4,000 calls a week, and she closes 60% of them herself. Unanswered calls are down to fewer than five a day.
In late February an incident disrupted road travel across several Mexican states and passengers called all at once. The human contact center saturated within hours. MarIA didn’t.
Read the full case study in the thread 👇
While agents scale the work that can be done within an enterprise, every new agent comes with a workflow to build, tests to write, data to wire up, and another custom interface someone has to log into to monitor operations.
How much of your business runs on agents depends on how fast your team can build, fix, and run everything around them.
Today we’re launching Prisma: an intelligent copilot that acts as your chief of staff, using the platform the way your team would and proactively raising what needs your attention. Use it to plan and build agents and workflows, the tests that gate them, the data they produce, and the apps you’ll monitor and manage them from.
Prisma keeps a live read of your whole operation, and when something slips, it traces the cause from the metric, to the graded sessions, to the node that broke. Then it writes the fix, re-runs the suites, and carries on monitoring.
Build, run, and improve your agents. All in one place.
More on how our team built Prisma, in the comments.
That’s a wrap! Our #SFTechWeek, by land, air, and court: a blimp over the Bay, HappyRobot cars between venues, and our very own padel tournament.
It was a pleasure meeting so many of you. Until next time, San Francisco @Techweek_ 🤖
In my @Forbes_es interview, Marina Specht asked me what our Series C meant beyond the number. My answer was trust. We got here because our customers keep choosing to work with us.
They show that trust in many ways: they let our engineers work side by side with their teams, show us how their operations really run, let us come back with ideas for what to build next - and some have even invested in our rounds.
That trust is earned slowly, and it comes from being customer-obsessed. When you put customers first in everything you do, revenue, fundraising and growth inevitably follow.
Read the full interview (in Spanish) 👇