Navigating the AI frontier: How agencies can embrace digital transformation with Salesforce

At the moment, all conversations about digital advancement in the public sector seem to be focused on the potential uses of artificial intelligence. Although not a new development — many government organizations were early adopters of the technology — AI implementation has previously been on the more exclusive side. 

But something has changed over the past year: The rapid evolution of widely accessible AI-backed tools, mainly generative AI, that have the capability to solve everyday problems. 

“AI has been around for a while. What's changed from the past is that it's at our fingertips,” says Andy Cather, a distinguished strategic solution engineer for public sector at Salesforce. “The impact on the public sector is, now that it's front-of-mind and intertwined with everyday life, agencies want to jump right in and start using it on a variety of projects.”

The truth is that AI has the potential to provide countless advancements for government employees and constituents alike. Automated and streamlined internal systems, increased productivity, innovative citizen services and improved correspondence seem just within reach. Alongside its promise, however, AI brings considerable uncertainty around ethics and security.

To help government organizations navigate this uncertainty, the Biden Administration recently released an Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, calling for a government-wide effort to assess and regulate AI use. 

“The executive order is telling government agencies that it's okay to get started thinking about policies that you want to put in place, and high-level use cases,” Cather says. “It's okay to take a risk-based approach to things if you want to start off with a small project with generative AI, or if you've been thinking about predictive AI.” 

While deciding on a starting point can seem daunting, agencies don’t have to embark alone. Private sector partnerships can be instrumental in successful AI adoption, especially with companies that have a breadth of experience in the space. 

Fortunately, government organizations need look no further than Salesforce, which has made significant investments in the research and development of trustworthy AI-powered solutions. The company prioritizes the many ethical considerations around AI, as illustrated in its AI Acceptable Use Policy, which outlines parameters for responsible operation and strongly encourages customers to “consider if their use of technologies is safe.” 

“Salesforce can bring that expertise in … and note things to be aware of, and why you just don't want to plug into any open AI model that's out there,” Cather says.  

Salesforce’s trusted tools

Salesforce has been working with AI for a number of years now, initially launching Einstein — the first comprehensive AI technology for CRM — in 2016 at it’s user conference, Dreamforce.

Salesforce Einstein has since evolved with the next generation of Einstein AI: Einstein GPT, a suite of predictive analytics and generative AI capabilities that power Salesforce's Customer 360. The company builds each tool on the premise that anyone should be able to successfully use and deploy AI, not just data scientists. By making this technology natively available across their platform, customers can more seamlessly integrate AI into the daily flow of work. 

Security considerations, like those brought forth by the executive order, are key to Salesforce’s AI development. That’s why Einstein GPT products are supported by the Einstein GPT Trust Layer to deliver secure generative AI experiences across applications and workflows. 

“What we mean by the trust layer is, your data is not our product, your data is your data,” Cather says. “When you submit a prompt to the AI, it doesn't need your data. If your data is part of the prompt, we have relationships with vendors where they will not retain your data, they will answer the prompt and then the AI will forget it.” 

Salesforce's commitment to being an open ecosystem also sets it apart, offering integration with various AI vendors. As part of Salesforce’s current product roadmap with the Einstein Trust Layer, they are working towards evaluating prompt responses from large language models to remove toxicity and bias. 

Critical considerations for AI implementation

Cather cautions agencies not to bite off more than they can chew. He recommends government customers start by addressing the basics and gives three key areas of evaluation to determine how the technology could positively impact organizational fundamentals. 

First, agencies must begin their AI journey by reexamining and shaping policies to better establish comprehensive ground rules and strategies for use. This includes what areas AI will be off limits, and identifying the individual(s), such as a dedicated AI manager, board or council, who will manage AI functions within the organization. 

The second step is for teams to conduct in-depth analysis and discovery to determine how AI can support their specific mission, and what tools are best for their particular journey.  

“Writing email with AI is great. But is that the most powerful part of AI that would be beneficial to your agency or department? Probably not,” states Cather. “For example, agencies are asking us how AI can help reduce fraud. AI in the flow of work could help generate predictions and insights to help end users make better decisions and reduce the distribution of fraudulent benefits based on historical information and similar use cases."

The third and perhaps most important aspect of AI adoption, according to Cather, is recognizing that AI tools may not be useful on their own. For predictive AI or creating a new model the technology must first be trained, which requires supplying the AI with sufficient amounts of relevant data for context so it can develop the best information outputs for uses and fully align to an organization's policy. Large language models in the marketplace are trained on publicly available data and public sector agencies need to consider which data AI will need for their use cases. 

“If you are unsure what data are you going to leverage or look at for utilization, put AI to the side and figure out what data would be best to look at and analyze first,” he says. 

Keeping up with a constantly evolving landscape 

Amid much mystery, one thing is certain about the future of AI: It’s here to stay. And as the proliferation of tools continues to grow, Salesforce solutions will evolve as well, helping its government customers meet their missions. Indeed, the company releases three new tools each year and are currently piloting the next generation of Einstein products. 

“If we look even one year forward, what I tell customers is that they will see a huge jump from Salesforce and what our AI can do,” says Cather, noting that if public sector organizations can consistently evaluate their use cases and needs, focus on simple pillars of work, and partner with seasoned organizations like Salesforce, they will be equipped with the right resources and technology to safely embrace AI. 

“We've seen a lot of agencies that have been successful when they can get down to brass tacks and they focus on very simple pillars of work,” he says. “Government organizations would benefit from those three basic fundamental steps that I mentioned, if they do that they'll be able to enhance their readiness long-term as AI shifts.” 

Learn more about Salesforce’s work to ensure AI readiness in the public sector. 

Any unreleased services or features referenced in this or other press releases or public statements are not currently available and may not be delivered on time or at all. Customers who purchase Salesforce applications should make their purchase decisions based upon features that are currently available. Salesforce has headquarters in San Francisco, with offices in Europe and Asia, and trades on the New York Stock Exchange under the ticker symbol “CRM.” For more information please visit https://www.salesforce.com, or call 1-800-NO-SOFTWARE.