AI for the public good: The future is bright

With public sector organisations typically more strained for resources and personnel, AI can offer significant opportunities to automate tedious tasks and free up time for staff to focus on more strategic, impactful initiatives. Regional research is backing up this idea with the following estimations:

• In the UK, it’s estimated that AI can easily automate 84% of citizen-facing transactions and has the potential to save £200 billion each year.

• In the US, it’s predicted that the public sector will see $519 billion in productivity gains from generative AI by 2033.  
• In Australia, generative AI has the potential to add between $45-$115 billion each year to the Australian economy, primarily through enhanced productivity.

A recent Elastic study on global generative AI implementation found that 99% of respondents recognize the positive impacts generative AI can have on their organization. At the same time, 89% of respondents report that their use of generative AI is being slowed. In the UK, only 37% of government bodies surveyed by the National Audit Office (NAO) were actively using AI. 

Factors slowing generative AI implementation globally

To speed up AI adoption and help citizens see the positive effects of AI sooner, there are a few tangible steps agencies can take.

While government agencies are understandably daunted at the prospect of reconfiguring their underlying data strategy, there are technology platforms that enable organisations to get more utility and insights out of the data they already have — no matter the location or format — without moving it or duplicating it. This “data mesh” approach can then serve as a unifying data layer that helps organisations most effectively integrate their own data with generative AI.

Further, many public sector organisations have been finding it beneficial to use retrieval augmented generation (RAG) workflows to connect their proprietary data with generative AI applications. This allows them to securely reap the benefits of generative AI technology without compromising the privacy of sensitive data. It essentially serves as a “context layer” that sends only the most relevant data to a generative AI application.

When implementing this context layer between private data and public generative AI, a vector database is a critical determinant of success. A vector database can normalise and streamline data so that you can search it quickly, no matter what format it’s in. Because a vector database stores information as vectors, or numerical representations of data, it’s much easier to find and correlate data across traditionally incompatible formats, such as text, images, sensor data, and more. Having the ability to quickly find what you need from all data then enables you to send the right, context-rich information to generative AI applications, thereby increasing the accuracy and relevance of AI outputs.

As a longtime partner with the public sector, Elastic is working with many government, healthcare, and education organizations around the world to securely implement generative AI into their technology environments. The generative AI use cases for public good are myriad, but here’s just one example of what it looks like in action.

Imagine you’re a high school student applying to college. Your family is supportive but doesn’t have the financial means to pay for your tuition. You’ve heard about a few classmates who have gotten financial aid packages and scholarships, but you don’t know where to start. What opportunities are available to you? How do you apply? What schools can funding be used for?

One university set out to help potential students answer these questions, through the power of generative AI. Georgia State is a public university that serves 50,000+ students across the US state of Georgia and has been named the second most innovative university in the US.

The team set up a proof-of-concept for generative AI at the university, using the financial aid use case as its pilot. It’s an ideal use case for GenAI because there are so many disparate, publicly available resources on financial aid from multiple sources and formats. All this data is ingested into Elastic and turned into vectors (numerical values), which allows it to be searched semantically (or in a user’s natural language instead of through a specific keyword).

1. Take a privacy-first approach. It goes without saying that the data that public sector organizations are responsible for is often highly sensitive and subject to regulatory scrutiny. Using a platform like Elastic — with its role-based access controls — can provide a secure bridge between your organization’s own data and publicly available LLMs. Elastic’s power as a vector database is critical in the process, as it allows data to be searched and analyzed as numerical values with private information removed.

2. Make sure your data foundation is healthy. The quality of generative AI outputs is only as good as the data you give it. Make sure you have a way to see across your entire data landscape regardless of data format and location. That way, when a RAG workflow searches your organisational data for relevant context, it will have an accurate, holistic set of data to draw from.  

3. Modernise legacy IT for unified data and visibility. With limited budgets and disconnected tech tools, many government agencies find themselves piecing together legacy tools and systems that don’t speak to each other or offer a unified view of your environment. This is problematic when agencies begin using generative AI since the AI will be producing outputs based on incomplete, siloed, and outdated data.
4. Save costs by avoiding fine-tuning. Although AI is poised to save public sector time and budget, behind the scenes, it can take a lot of effort to fine-tune a large language model (LLM) yourself because the job will never be done. There will always be more data to add. When you use RAG, however, the generative AI model will always pull from your most up-to-date, relevant data.

Learn more about Elastic’s approach to AI for the public sector.