Looking beyond the hype on AI

ChatGPT systems and other generative AI tools have the potential to remake the federal government, allowing for improved citizen experience and employee experience. But the federal government needs to understand that this technology is far from a silver bullet. As with any new technology, there are issues, complexities and yes, lots of hype. So while it’s important for the federal government to pursue generative AI and experiment with its capabilities, this must be done in a thoughtful manner and with a clear-cut strategy that accounts for the limitations of generative AI.

Last year, McKinsey & Co. announced a wide-ranging survey about the satisfaction levels of government services — for both the federal and state levels. The research included responses from over 78,000 Americans. On average, the performance was at the lowest levels among various other private industries like airlines, cable, car insurance and mobile phone services. 

Yet using generative AI can certainly be critical in boosting the citizen experience. This technology is the ultimate in self-service. By using a natural language prompt, a citizen can write or say a detailed question, such as about retirement or veterans’ benefits. The AI system will then process relevant information like the person’s age, years of service, salary history, residence and family members. It will even understand the implications of taking actions. For example, how might applying for a federal retirement benefit might impact Social Security, disability or Medicare? 

The AI will then come up with a response – which can be in various languages -- that is easy to understand. 

All this sounds great, right? Definitely. But there should be a dose of caution. The technology is still fairly new and is quickly evolving. There are also notable issues with generative AI. For example, it is susceptible to “hallucinations.” This is when the system makes a statement that is convincing but still false or misleading. When it comes to essential government services and complicated issues of benefits and taxes, this can certainly be a serious problem. 

For the federal government, this means there will need to be a commitment to long-term investments as well as the help from trusted technology partners that have proven systems to deal with the challenges of generative AI.

What can AI do right now?

For generative AI, EX is where the federal government can get more near-term results. For example, an LLM can create interesting copy for social media, blog posts and press releases. Let’s face it, writing this type of content can be difficult but an LLM can write with a certain tone and style. It can also spark ideas. This can not only improve the quality of the content but also speed up the process. 

Another area where generative AI can help is with analyzing policy statements, laws and regulations. The data can be in any format, say PDFs with handwritten content. An employee can use the summaries to help create useful reports, white papers or presentations.

A more ambitious use case would be to leverage generative AI for service desk operations. For a human support representative, answering a question could easily take over 30 minutes. This would involve accessing disparate legacy IT systems, cutting-and-pasting information, writing emails and even putting together knowledge base articles.

But generative AI can handle all this in seconds. In fact, it can essentially process many of the tasks for tier-1 and tier-2 support. This means that human support representatives will have much more time to focus on complex issues, not mundane tasks.

Besides the service desk, another opportunity for EX is to improve workflows. By using natural language, an employee can create a bot that automates a repetitive and mundane process. There will not be a need to learn a complex language like JavaScript. In light of the tech talent shortage, this could be a huge benefit.

A big advantage with generative AI is that it allows for unsupervised learning. This means there is no need to label data or create workflows. The LLM will understand the data on its own.

This can certainly be a critical advantage for the federal government, which has massive repositories of data — usually in PDFs. But generative AI can unlock the value of the data.

However, this process still requires expertise. Part of this is due to understanding prompt engineering, which can often be more of an art than a science. 

In the meantime, the federal government will certainly need to remain vigilant with the privacy and security of its massive date repositories. When it comes to CX and EX applications, the best outcomes are usually from using personally identifiable information. 

While hype is common in the tech industry, the emergence of generative AI appears to be very real. The impact is likely to be ubiquitous and long-term. 

For the federal government, it does represent an opportunity to realize significant gains for CX and EX applications. So it’s definitely time to explore and experiment with generative AI. The technology is too important to ignore. But along the way, there needs to be a recognition of the risks and challenges.