Agencies should reach out to potential industry partners about their artificial intelligence needs now, even if they have no immediate plans for the technology, because preparation will take years of upgrades and training.
Most agencies are aware they must make extensive upgrades to their IT infrastructure before deploying AI, which explains why some are hesitant to act.
Despite the fact that 55% of organizations were piloting AI agents, many acknowledged they weren’t close to ensuring the level of data quality or cybersecurity necessary to achieve their generative AI goals, according to a KPMG survey from September.
Government is typically three to five years behind industry in adopting emerging technologies, but agencies may soon face a mandate to leverage AI, which only makes matters more urgent.
Click the banner below to ready your data foundation for AI.
The Resources That Make Up an AI-Ready Foundation
Civilian agencies have a chance to get out ahead of the Department of Defense on AI because of the military’s intensive vetting processes to identify security vulnerabilities in new technologies. All agencies undoubtedly need to establish data governance guardrails around what information will be accessible and ingestible by AI.
A handful of agencies are already broaching conversations about AI-driven software-defined WAN tools, and CDW Government has teams that can help establish zero-trust security and segment networks to prevent the technology from being misused, based on its data access, in the event of a breach.
Generally, agencies will need to upgrade to servers equipped with graphics processing units, which offer high-performance storage, but AI also demands more memory at a time when AI companies are buying up RAM and causing prices to skyrocket. Smaller organizations are stockpiling DDR5 RAM, as newer memory models are being created specifically to run AI farms.
Agencies will also need to modernize their power and cooling and network connections. Many still have 1-gigabyte connections, but AI requires a minimum of 25 gigabytes, and some switches run up to 400 gigabytes.
Click the banner below for the latest federal IT and cybersecurity insights.
Then there are cloud considerations, as many agencies want to scale AI resources up or down faster than data centers are may be able to. Infrastructure as a Service via Amazon Web Services or Microsoft Azure is always an option.
All these requirements will need to be adjusted based on the type of AI models being used. Generative and agentic AI have different demands.
Pick Your AI Flavor Once You Know Your Objective
CDW has an AI factory team that offers a range of two-day workshops on DOD’s Joint Enterprise Defense Infrastructure, AI development, AI research and AI security. A good place to start is the executive workshop, which can help align an agency’s goals to the AI models that can make those goals a reality.
Even within agencies, components often differ on their desired objectives. They need to identify areas where personnel have been phased out or where time-consuming tasks could be automated before deciding on what type of AI they need to address those challenges.
