Business It News / Business Intelligence
Splunk declares transformational shift as AI agents reshape enterprise IT
Splunk has made a number of announcements this week that has been described by executives as transformational rather than incremental.
At its .conf26 event in Denver the company announced a broad set of platform, security and observability enhancements as it prepares enterprises for an era dominated by AI agents.
Delivering the opening keynote, Cisco President and Chief Product Officer Jeetu Patel outlined a number of upgrades including a multi-year agreement with long-term partner AWS to co-develop security solutions and increased collaboration with NVIDIA to bring Splunk AI to on-premises customers.
-
A new tokenomics solution, Splunk Agent Observability, which tracks AI spending and coding agent usage in real time, tying adoption to business value.
-
Advancing agentic SOC (Security Operations Centre) by combining enterprise-wide telemetry with specialised AI agents powered by leading frontier and domain-specific models through the Splunk Agentic SOC Workforce
Patel set up the announcements by looking at what was happening in the industry and what companies needed to do to get the fullest potential out of AI while keeping in mind the downside risks that might be there.
He talked about the move from chatbots to agents and its impact on token consumption as well as the infrastructure needed to cope with the additional demand. Patel said that in February token consumption by agents exceeded the token consumption by humans for the first time.
Since then, the rate has increased 14-fold and agents are consuming about five times more tokens than humans, despite only an estimated 2% of humans currently being power users. “Imagine what happens when it is 10% of humans, 15%, 20%, 40%, 50%? The infrastructure that is going to be needed is going to be very different.”
Patel said the implications for infrastructure are significant. He said agents can operate 24/7 rather than waiting for humans to initiate interactions, while Cisco studies indicate an agent consumes 450% more network bandwidth than a human performing the same task.
The scale of inference is also changing. Patel cited an AMD projection that around 60% of global AI compute capacity will be consumed by inference rather than training this year. He said this was evidence that AI models are moving from experimental technology into workloads generating real-world returns.
Patel expects this demand to spread across cloud, on-premises data centres and a new category of “desk-side” computing, with organisations potentially running tens of thousands of desk-side computers and eventually trillions of agents. Individuals, he suggested, could ultimately have hundreds or even thousands of agents working on their behalf.
That creates three major challenges: optimising AI at scale, defending at machine speed and turning machine data into agentic action.
In return Splunk will provide three critical capabilities: AI observability, an agentic SOC and a unified data fabric.
AI observability will enable organisations to monitor infrastructure utilisation, model performance, application runtimes and agent behaviour, while new tokenomics capabilities will expose AI spending and enable organisations to intervene when an agent begins consuming excessive resources.
The agentic SOC is designed to move security response from human speed to machine speed, while Cisco’s Data Fabric will correlate network, application, infrastructure and security telemetry.
In fact, the Cisco Data Fabric is the centrepiece as it can analyse information where it resides instead of requiring customers to copy everything into Splunk.
As part of the keynote, several Splunk clients and partners were brought on stage including the billionaire CEO of Intel Lip-Bu Tan.
Patel asked the 66-year-old what advice he would give to maximise the potential for humans in the current era while minimising the risks that AI might pose.
“I highly recommend you should embrace AI. And security is critical. It is very important to protect your data. Let me model come to you and then make sure you have all the security to protect it. Don’t feel like everything is so negative. There is actually so much that is positive and so much potential that we can enjoy.”
The two had some banter back and forth with Tan revealing he has 22 meetings a day including three meetings for dinner where at one he eats a salad, one meeting for main course and one meeting where he eats a desert. He also swims every day.
