How We Made That App: Igor Jablokov
Explore the future of AI with Pryon Founder Igor Jablokov on this episode of How We Made That App, hosted by SingleStore CMO, Madhukar Kumar.
AI for Defense: Key Considerations to Maximize Security, Impact, and Efficiency
Introduction: Government organizations have a knowledge friction problem
Defense and intelligence agencies have a wide range of projects to manage, from allocating funding and training service members to analyzing supply chain risks and summarizing mission-critical data. No matter the mission, though, none of these initiatives can get off the ground if the right people don’t have access to the right information in a timely manner.
Unfortunately, federal government organizations face an inordinate amount of information proliferation. Crucial data is sprawled across dozens of SharePoint and ServiceNow instances, and much of this information is duplicative or outdated. This means employees and service members waste hours every single week simply trying to find the accurate, relevant information they need to do their jobs, leaving less time to act on this information.
In fact, over half of policies, procedures, and knowledge articles are not readily available to those who need access. Even worse, 70% of knowledge workers have reported spending over an hour searching for a single piece of information.
Another way to describe this gap between people and the information they need is knowledge friction. The negative consequences of knowledge friction can go far beyond wasted time — stalled workflows often give way to misinformed employees and delayed or poor decision-making in high-stakes scenarios.
Eliminating knowledge friction isn’t just a nice-to-have. It’s critical to ensuring our national defense. Consider, for instance, a commander who needs to decide how many aircraft to deploy for an important mission. He needs to quickly know how many aircraft the enemy has deployed, and in what formations; how many pilots and how much equipment he has on hand; how other agencies are responding; and so much more. In short, the commander needs all the relevant answers at hand, fast, to achieve a decision advantage.
Generative AI, done right, can help eliminate knowledge friction
Extracting relevant information from systems such as SharePoint and ServiceNow and providing it to decision-makers on demand can increase federal organizations’ operational effectiveness and productivity.
Generative AI, a technology that’s seen enormous uptake over the last couple of years, can help eliminate knowledge friction. AI for defense solutions built on a retrieval-augmented generation (RAG) framework can quickly read and understand a complex ecosystem of content, including policies, procedures, manuals, lessons learned, and other documents spread across different content siloes, and deliver accurate answers based on this content to users.
RAG combines retrieval and generative capabilities to deliver accurate and trustworthy responses. This approach mitigates common issues with GenAI tools, such as hallucinations, and ensures that responses are based on authoritative content with clear source attribution.
RAG works by understanding the semantic meaning of a user’s query, searching an organization’s vast content library for answers, retrieving the correct ones, and generating a smooth, coherent, trustworthy response — all in milliseconds.
RAG-based GenAI solutions can help:
- Agencies rapidly source and implement lessons learned from previous missions
- Procurement and contracting teams expedite the RFI/RFP process by quickly generating template documents, asking questions of long submissions, and receiving summaries of the most pertinent data
- Officers and service members in the field get quick access to policy and procedure guidelines, just by asking a question through a mobile app
Top 4 considerations to maximize security, impact, and efficiency
Implementing GenAI in a defense and intelligence context — where security and reliability are paramount — requires important consideration. In particular, federal organizations must consider four key considerations when implementing AI for defense to ensure their RAG solution keeps sensitive data secure while still driving high mission impact:
- Accuracy: the foundation of trust and reliability
- Security: safeguarding sensitive information
- Speed: rapid time-to-value
- Scale: managing vast amounts of varied data
The following table dives deeper into what each of these considerations means and the attendant challenges federal organizations face.
| What this looks like | Challenge #1 | Challenge #2 | Challenge #3 | |
|---|---|---|---|---|
| Accuracy | The GenAI solution can precisely ingest, interpret, and act on a vast amount of unstructured content stored across multiple systems | Highly complex and duplicative content | Understanding natural language user queries and delivering relevant, correct, and complete responses | Mitigating AI hallucination |
| Security | The GenAI solution is impervious to data breaches, unauthorized access, and cyberattacks, and can be deployed in on-prem or air-gapped environments to ensure strict data security | Protecting against data leakage and cyber threats | Deployment in virtual private cloud (VPC), on-prem, or air-gapped environments while preserving accuracy and speed | Controlling access to sensitive information through data governance |
| Speed | The GenAI solution is quick to implement, integrates seamlessly with existing systems, and delivers answers in milliseconds | Overcoming typically lengthy deployment timeframes typical with many GenAI tools | Integration complexity due to outdated legacy systems and the need to comply with stringent regulatory requirements | Stalled content updates |
| Scale | The GenAI solution can continuously ingest millions of pages of multimodal content stored in a host of repositories, without compromising on accuracy or speed | High volume of content in different formats and places | Frequent content updates | Strict latency thresholds |
How to implement AI for defense — the right way
In addition to ensuring a GenAI solution adheres to the four tenets outlined above, federal defense agencies must also understand what knowledge friction looks like for them and see the results other federal organizations have achieved from an AI for defense deployment.
To dive deeper into knowledge friction, the top 4 considerations for maximizing impact and security, and how leading organizations have implemented GenAI themselves, check out Pryon’s guide, Generative AI for Federal Government.
