The second era of artificial intelligence (AI) is upon us, in which we move from chatbots that intelligently answer our questions to agents that accomplish tasks autonomously, or nearly so, Cisco President and Chief Product Officer Jeetu Patel noted at the recent Splunk .conf25 in Boston.
That change demands a new data architecture, Patel said, casting Splunk as “the machine data fabric for the AI era,” and tying the new Cisco Data Fabric and the forthcoming time-series foundation model and AI Canvas to faster, cross-domain incident response and more trustworthy AI operations.
“This is going to be much more about automation of workflows, rather than just individual productivity,” Patel said. “There’s going to be exponential compounding of throughput capacity per capita that you’re going to see over the course of the next few years. … [which] is going to have massive implications on technology architectures.”
The three hurdles to AI at scale
Scale matters. As Splunk Senior Vice President and General Manager Kamal Hathi noted, the Splunk community already drives 2.6 exabytes ingested and 300 billion searches per year – a reminder that any AI approach must operate at machine scale.
Three factors could constrain AI adoption and scalability, Patel warned: infrastructure limits, a trust deficit, and a machine-data gap because models have been trained mainly on human-generated data, not logs, metrics, traces, and events. Splunk and Cisco address all three, he noted, with networking and data-center plumbing, AI observability and defense, and a machine-data platform.
What’s new and next under the Cisco-Splunk hood
Cisco is addressing the second era of AI with new solutions. The new Cisco Data Fabric, announced at .conf25, is designed to harness the value of machine data with AI at “ludicrous scale,” unlock proprietary data for AI, and unify experiences for humans and agents, Patel said.
To operate on machine data at extreme scale, Splunk is extending its “bring search to the data” federation model beyond Amazon S3 to Snowflake, with an alpha release in February 2026 so users can query business data in Snowflake from Splunk and enrich it with Splunk data without centralizing the data first.
To unlock the value of proprietary data by closing the machine-data gap, Cisco aims to “teach AI to speak machine data,” backed by a time-series foundation model and a machine data lake to tune models on enterprise telemetry, enabling multivariate anomaly detection and forecasting across logs, metrics, and traces. The model is slated to be listed on Hugging Face this month and the machine data lake available in 2026.
This means that “agents are going to be multilingual,” Patel said. “They will be able to speak the human language, which is natural language, and then they’ll be able to also speak machine language, which is time series data, and they’ll be able to … correlate that data together [to help us] see and respond to events that you could never have seen before.”
To unify experiences between humans and agents, Cisco is expanding its agentic ops capabilities with AI Canvas, which provides an AI agent to orchestrate the analysis workflow across ITOps, SecOps, and NetOps, as well as a workspace for team collaboration. In a live demo, an analyst investigated a suspected insider incident using natural language; AI Canvas tapped into the Cisco Data Fabric, built charts and network maps, suggested next steps, enabled multiplayer troubleshooting, and generated an incident report.
Why the second era of AI matters for government IT
For Federal IT shops juggling sprawling telemetry, tight budgets, and staffing constraints, new capabilities in the second era of AI can enable agencies to leverage data where it lives, add business context without moving data, and let agents compress the time from detection to resolution. The aim is to “detect the undetectable, react autonomously, and anticipate the future,” Patel said, shrinking investigation time and improving resilience across the stack.