It’s called Dialog and promises to turn “fragmented interactions into continuous customer relationships.”
Cisco is updating its Webex cloud collaboration <a href="https://myappsplus.com/google-experiments-with-an-ai-powered-gaming-platform/” title=”Google experiments with an AI-powered gaming platform”>platform to better accommodate AI agents. Perhaps the biggest bit of news here is a new toolset called Dialog, which is an “agentic harness” that directs AI to handle customer interactions.
The company says that Dialog tasks various agents with keeping the consumer happy, as they’ll “keep working on the customer’s behalf after a conversation ends.” This means that agents under Dialog’s umbrella have the ability to coordinate “across people, agents and backend systems.”
The platform also allegedly speeds up the onboarding process, treating “AI agents like new teammates.” This means they’ll be given the opportunity to peruse existing employee handbooks, knowledge bases and operating procedures.
Cisco promises a safety-forward experience, but this is a new feature and you never know what issues are gonna pop up. Agentic AIis gonnaagentic AI. The company does suggest that the Dialog platform “continuously refines agent performance with every interaction.” However, that’s kind of the promise with all AI agents. We’ll have to wait and see
As for employees, the updated Webex will let anyone in a workspace directly invite AI agents into “spaces, meetings and calls to execute complex, multi-step work” across the platform and in third-party applications. Cisco gives an example of an agent quickly pulling up statistics during a presentation and updating a PowerPoint on the fly. That could be extremely useful, assuming it works.
“The numbers remain dismal. Researchers at Carnegie Mellon University found earlier this year that even the best-performing AI agent, which was Google’s Gemini 2.5 Pro at the time, failed to complete real-world office tasks 70 percent of the time.”
To that end, multiple studieshave indicated that AI agents are still prone to mistakes which lead to cascading failures. The failure rate sits somewhere between 60 and 70 percent with regard to multi-step tasks. This is because all it takes is one single mistake during one step of the process to throw the whole thing off. Even an agent with a 95 percent per-step reliability rate will only succeed around 36 percent of the time when dealing with a 20-step workflow.
This will likely improve over time, but we aren’t quite there yet. Even a 99 percent per-step reliability rate seems incredibly risky when money and a company’s reputation is on the line.
Update October 7 2026, 10:55AM ET: This article has been updated to clarify that Cisco owns Splunk.
