Presented by LightningIQ & Carahsoft
Giving an AI system access to more information does not automatically make it more useful. Agencies first need to understand which records are relevant, which are outdated, and which contain information that requires special handling. Dan Kinney, EVP, Field CTO at Lightning IQ, said those decisions can determine whether an AI initiative produces useful results or costly confusion.
In an interview with Fed Gov Today at the Google Public Sector Summit, Kinney connected information governance with AI readiness. Organizations may hold enormous volumes of unstructured information across cloud and on premises environments. The challenge is identifying the portion that supports a specific task before introducing it into an AI workflow.
Kinney illustrated the problem with a vehicle maintenance example. An organization seeking better service insights might need five years of repair history for a particular model. Providing 25 years of broader information could introduce irrelevant records and conflicting details. The system would be processing more information without necessarily improving its understanding of the intended problem.
Duplicate records create another concern. So does sensitive information that an organization may need to flag, exclude, or handle differently. Kinney said agencies should make deliberate choices about those categories based on the purpose of the initiative and the requirements governing the data.
Preparation also has a financial dimension. Kinney noted that AI processing costs can depend on the volume of information entering and leaving a system. Selecting relevant data can therefore support both better results and a more disciplined investment. The planning question becomes which information will contribute to the intended outcome.
At the summit, Lightning IQ introduced Lightning Recall, a product Kinney described as addressing the cost of retaining archived Google Workspace users. Organizations with lengthy or permanent retention requirements can accumulate substantial archived-user license expenses, even after those employees have left.
Kinney said the product allows organizations to export archived-user data to Google Cloud storage and retain search capabilities for legal and compliance requests. He described the approach as an alternative for organizations seeking to manage retention costs while maintaining access to the information they must preserve.
The archived-user example and the AI readiness discussion share a common requirement: organizations need to know what they have and be able to find the relevant records. Storage decisions alone do not resolve questions about discoverability, sensitivity, or the business purpose of retained information.
For agencies planning a larger AI effort, Kinney’s advice was to examine the data estate early. Identify useful time periods, subjects, and sensitivity categories. Apply those choices to the actual records, rather than assuming an entire repository belongs in the initiative. That preparation gives agencies a clearer basis for deciding what an AI system should use and how the investment can support the mission.