The agent conversation has moved fast. We went from “can it talk?” to “can it do it?” and now to “can they work together?” Everyone wants the seamless version where multiple agents collaborate like a well-run team.
But the moment you put more than one agent in the room, you ask the system to behave like a team under pressure. Agents step on each other, lose context, and take the wrong action with the right confidence. When you add more agents, you multiply the chances of that happening.
You do not solve this by polishing prompts forever. You solve it by making decisions up front about how the system runs: how work is directed, how context is carried, what actions are permitted, what is verified, what gets logged, and what triggers a repair or a handoff.
In other words, coordination is an architectural problem.
That’s why we are excited to see Rasa featured in Gartner’s reference architecture for conversational AI. Not because a Gartner diagram is “the answer.” but because it reflects something teams learn the hard way: building customer-facing agents is an architecture problem before it’s an interface problem.
This recognition also caps off an exceptional year of milestones across customers, product, and the next chapter of AI agents:
- Over the past year, we have been lucky to earn a lot of trust from teams running customer experience at serious scale to partners who care about getting the architecture right. A meaningful signal of that trust is Rasa named by Accel as one of the Top 100 AI and Cloud Companies to Watch.
- Customers like Autodesk, nib Group, N26, Groupe IMA, Swisscom, and Providence Health run production agents on Rasa, handling millions of interactions across industries where reliability and scale aren't optional.
- Across our product, we shipped a big step forward in enterprise voice and the broader platform innovation. At Customer Contact Week in Las Vegas, we launched Multimodal Voice AI so agents can handle speech more directly, without the old tradeoffs that come with traditional speech-to-text pipelines. In parallel, we rolled out additional building blocks for the next chapter of agent systems, including deeper orchestration patterns across MCP, A2A, and autonomous agent skills, etc., so teams can connect tools, skills, and multiple agents without losing control of the end-to-end system.
- As more companies put agents (and usually multiple ones) directly in front of customers, trust becomes the product and the brand. We’ve taken that seriously from day one, with clear designs on what an agent is allowed to do and strong handling for sensitive data. Working with partners like Lakera adds another layer of confidence for teams deploying agents in high-stakes environments.
We’ve spent years building Rasa for the moment we’re in now. The future is not one perfect agent. It’s portfolios of agents, coordinated across channels, connected to enterprise systems, and governed like any other production software. That’s the work we’re here for, and it’s what we’ll keep pushing forward with our customers.
If you want to see where we’re taking agent building next, try Hello Rasa. It’s the fastest way to stand up a customer-facing agent in minutes, iterate in plain language, and shape behavior with real boundaries from the start. No spinning up a heavy environment first, no maze of nodes. Just build, test, and feel what “modern agents with real control” is supposed to be.
Talk to us about your agent architecture
As your agents move into higher-stakes customer journeys, now is the time to get the system right. Reach out, and we’ll show how leading teams design for consistent behavior, clean handoffs, and predictable outcomes across channels.
About Gartner's Reference Architecture Brief: Conversational AI Chatbots and Assistants
Gartner does not endorse any company, vendor, product or service depicted in its publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner publications consist of the opinions of Gartner’s business and technology insights organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this publication, including any warranties of merchantability or fitness for a particular purpose.
