The power of generative AI for government and public sector

In its current state, GAI can produce impressive content, conversations, images, and more. But those results are only as relevant as the data the tool has been trained on. When training data sets — which provide the appearance of knowledge inside large language models (LLMs) — are based on publicly available data on the internet, the answers they generate have limited scope. GAI based on public data is often prone to hallucinations — incorrect information presented as if it were accurate.

On the other hand, when GAI is used with an agency's internal data, it can significantly accelerate mission outcomes, improve citizen services, and better connect government knowledge workers like analysts and cybersecurity professionals to the right data at the right time. Why? Because that institutional data adds essential context.

The combination of GAI and private, institutional data has a force-multiplier effect. The naive solution would be to bake private data into the models themselves; however, the complexity and costs of training or fine tuning AI models — multiplied by the number of domains and points of interaction in government — becomes untenable. Instead, the same questions asked to the LLM can be first taken to Elastic's AI-backed search capabilities, where the most relevant fact-based answer based on your internal data can be found.

This domain-specific context that your data brings to GAI can make the output more accurate, relevant, and actionable for your mission. A prerequisite for “bringing your own data” is that your data is stored in a unified data platform where it's accessible and findable in one place.

For the public sector especially, you don't want your highly sensitive data mixed with publicly accessible GAI — or any system where you don't retain control of your own data. Any search query sent to a publicly available GAI product (such as ChatGPT) is consumed by the model, meaning that your internal data is no longer internal. Even if your organization doesn't formally use GAI as part of your tech stack, it's a fairly sure bet that your employees will use it anyway.

Help make sure your internal data stays in the right hands by strategically integrating GAI with your proprietary data in a way that your IT team can control and have insight into. Otherwise, you could have employees unintentionally putting your sensitive data into a public GAI service like ChatGPT, where you can't ensure its security. Ideally, you would integrate your proprietary data into a platform designed to handle sensitive information, where you retain full control of your own data and enable role-based access control (RBAC). More on this below.

Data is one of the most strategic assets that public sector organizations own today. When your data is unified and stored in one platform — where it can leverage GAI and search technology — the real-world implications can be far reaching, providing benefits such as:

Imagine that a citizen is looking to apply for public housing services. The application process involves several steps and forms, which differ based on needs and location. Simply listing generic information on a webpage would be complex and likely would not address a citizen’s unique situation. On the other hand, when agencies bring their own data to GAI, a citizen could find information and instructions tailored to their individual circumstances. This highly relevant information has the potential to reduce the complexity that often prevents people from accessing essential services in the first place.

Or, for another example: you’ve been summoned for jury duty and need to know what happens next. Where do you need to go? How long will it take? Were you selected as a juror? Does your judge allow cell phones in the courtroom? Leveraging your data, GAI can streamline and personalize this complex information, potentially improving the citizen experience and building trust with government services and leaders.

For law enforcement and the intelligence community, democratized access to the right data in real time is critical. This is especially true when you have multiple organizations collaborating on a project — with disparate databases of information in different formats. Having the ability to find answers across data types and sources via a single GAI query has the potential to increase the speed and accuracy of the results, reduce manual and time-consuming work, and ensure that everyone who needs to can work off the same accurate data set.

When you integrate GAI with domain-specific context, you help enable your internal teams to quickly find the information they need that helps them do their jobs. A quick query across multiple data sets and formats can deliver hyper-relevant information in real time — preventing the need to painstakingly (and mind-numbingly) comb through documents or siloed databases. And in most cases, the information your teams are looking for won’t be found on the public internet or in AI model training sets, so it’s important to provide a GAI-powered tool for finding proprietary information quickly so your employees don’t turn to a public tool that could compromise your data security.

When employees spend less time on fruitless searching and manual data correlations, you take out one more source of friction in their day, paving the way for better job satisfaction and engagement, especially if you’re strapped for resources to begin with.

As you consider how to integrate your agency's data with GAI, the Elasticsearch platform can be a powerful tool. It allows you to ingest all types of data, store it economically, access it wherever it resides, and integrate it with GAI transformer models.
Elastic has worked to democratize search for over a decade, and we've been investing in AI and machine learning (ML) for a significant portion of that time. As a result, we've just launched the Elasticsearch Relevance Engine (ESRE) to help our customers find relevant answers to their questions through AI and ML on the Elasticsearch platform.

ESRE combines the best of AI with Elastic's text search, providing the ability to integrate with large language models (LLMs). It's accessible via a simple, unified API that Elastic's community already trusts, so developers can start using it immediately to elevate search relevance.

In other words, you can now connect your own GAI model or a third-party GAI model directly to your data you're storing in the Elasticsearch platform. This allows you to leverage the power of GAI with domain-specific data to produce answers that are accurate, relevant, actionable, and secure.

To learn more about ESRE, read the launch blog.

1. Unified data storage and democratized access. You can affordably store all your data in the Elasticsearch platform for democratized access, findability, and insights. Once your data is in the platform, you can use it for additional use cases, such as threat hunting and infrastructure monitoring.

2. The ability to find mission-critical answers that are:

• Accurate: The answers you'll get from GAI and your own data are based on mission-relevant facts — not hallucinations.
• Relevant: By using proprietary data in Elasticsearch, you avoid having to repeatedly retrain LLMs on your internal data, saving you time and training costs, and making sure your information is always up to date.
• Actionable: The Elasticsearch platform democratizes access to data and insights, allowing your teams to collaborate and make decisions in real time, from anywhere.
• Secure: Not every employee should be able to access every document, and certain data needs to reside in specific locations for data sovereignty purposes. Elasticsearch allows you to limit data access to certain roles within your organization, while still retaining the ability to search across your entire data store.

3. Cost-effective implementation. Because of the decades of optimizations in information retrieval, Elasticsearch presents knowledge to GAI interactions in a way that is orders of magnitude more CPU efficient than extracting the same knowledge from trained or fine tuned large language models. Some estimates put semantic retrieval at five times more efficient than using just ChatGPT 3.5 or 250 times more efficient than the CPU costs of GPT-4.

How much value GAI can create for your organization depends on your data and whether it's unified and accessible. If your data is spread across multiple tools and teams, you may be lacking the context and content you need to make GAI hyper-relevant to your mission goals. The Elasticsearch platform serves as a single data store for all your agency's data and a centralized jumping-off point for collaboration, AI insights, and automation.

• Join our virtual fireside chat with IDC: The future of generative AI in public sector
• Get a technical perspective on how to implement Elasticsearch and AI for privacy-first use cases.
• Contact an Elasticsearch public sector expert to talk about how AI can bring value to your agency's mission.

In this blog post, we may have used or referred to third party generative AI tools, which are owned and operated by their respective owners. Elastic does not have any control over the third party tools and we have no responsibility or liability for their content, operation or use, nor for any loss or damage that may arise from your use of such tools. Please exercise caution when using AI tools with personal, sensitive or confidential information. Any data you submit may be used for AI training or other purposes. There is no guarantee that information you provide will be kept secure or confidential. You should familiarize yourself with the privacy practices and terms of use of any generative AI tools prior to use.

Elastic, Elasticsearch, ESRE, Elasticsearch Relevance Engine and associated marks are trademarks, logos or registered trademarks of Elasticsearch N.V. in the United States and other countries. All other company and product names are trademarks, logos or registered trademarks of their respective owners.