As artificial intelligence (AI) continues to disrupt the digital landscape, counties are looking to leverage its potential to improve public services, enhance constituent engagement, and optimize operational efficiency. And while AI Is capable of achieving these goals, it isn't always an easy journey. County governments can encounter barriers to adoption, difficulty integrating AI into existing operations, ethical questions, and more. With the right strategies in place, counties can overcome these challenges and embrace a future with AI at the forefront.
This article summarizes a conversation between Matt Peters, Chief Technology Officer at CAI, and Menaka Indrani, Technical Services Director, Alameda County, California. Their discussion covers strategies to overcome barriers to AI adoption, real county use cases, and the challenges of integrating AI. They’ll also take a look at emerging technologies and explore how county governments can build a strong foundational understanding of AI before planning future investments.
This article has been adapted from a National Association of Counties (NACo) webinar of the same title. Read on for a closer look into the transformative role of AI in county government services – both now and tomorrow.
Overcoming barriers to entry for AI in government
Matt Peters (MP): I'm excited to discuss this because one of the things that I find really fascinating about this topic with respect to local government in particular, is how fast the government has actually been moving on understanding, adopting, and being responsible with AI solutions. But despite all that enthusiasm and everybody's willingness to try to lean in on AI, everywhere that I go and every time I end up talking about it there's a mix of both enthusiasm and caution.
There are a lot of barriers to entry in getting AI off the ground and getting people to trust it, understand it, and use it effectively and responsibly. So, my first question is: what have been the key challenges in investing and/or integrating AI in your county’s services? How did you address them?
Menaka Indrani (MI): AI has been moving very fast, and I would say the first challenge was data privacy and security concerns. When you integrate AI, you have to protect sensitive data, and that's been top priority. I work with a healthcare agency often, and with that we need to ensure compliance with data protection laws and state level privacy regulations. To solve this, we adapted our data governance policies. For example, we tried to anonymize datasets for AI projects and worked very closely with the legal and compliance teams to ensure we're handling the data properly. We also implemented robust cybersecurity measures to protect the AI systems and the data even more.
Another challenge we faced was the ethical side of AI and the bias related to it. A lot of data is used to train AI, and sometimes this data is not diverse or representative of the whole, and this can lead to inaccurate results. In government services, it is very important that we evaluate these AI tools before implementation. To combat this, we implemented a process called the technology acquisition request. When agencies want to procure or implement any software, they need to get it reviewed by our IT department where we have added a lot of reviews about AI.
And another thing is the skills, the workforce readiness for the AI systems. Implementing these AI solutions requires specific skillsets that initially we were lacking, so as AI technology came in, we needed to upskill the staff. For that we leveraged a lot of comprehensive training programs and partnered with AI vendors. We partnered specifically with Microsoft and Salesforce who have been helping us with the development of some of our systems.
And lastly, it was cost, because AI is not cheap. So we have to invest in our infrastructure, data processing, and skilled personnel.
MP: You mentioned earlier in this conversation being attentive to what I would call the AI Ops of deploying some of these solutions. There are always risks associated with new technologies, and that can bring ethical concerns. Have you deployed any specific strategies to try to, on an ongoing basis, quality check any AI solutions that have been deployed so that you can ensure their longer-term use hasn't impacted their performance?
MI: Like I'd mentioned before, we implemented a review process for any new software that needs AI, but we also need to review our policies and everything we've already implemented. And we continue to test as we move along to make sure our data is clean, because when we train AI on the data, the data itself may contain some private information or internal biases. And once AI gets trained on data that has some fault, it amplifies those biases. So continuously cleaning up that data set on which we are training the AI is the new norm.
"And once AI gets trained on data that has some fault, it amplifies those biases. So continuously cleaning up that data set on which we are training the AI is the new norm."
AI use cases in government
MP: Can you share any planned, ongoing, or completed AI projects in your county? And what are your expected project outcomes?
MI: As with any technology initiatives, AI projects all depend on the need. Modernization is my top priority, and we've created a converter that can migrate legacy applications to more modern platforms. It significantly reduces the time it takes to modernize an application. We're also able to automatically convert our mainframe applications.
Another thing we've implemented is what we call an invoice monitor. It reads invoices automatically and parses the data, putting it into workflows. It can parse invoices in any language and handwritten invoices too, from any of the agencies we work with.
We're also looking to use AI to enhance our ability to provide comprehensive and responsive cybersecurity solutions. We've leveraged AI for threat hunting and to identify any bad actors in our ecosystem, which helps us stay proactive.
MP: I know there are a lot of challenges getting an AI solution deployed, but I'd be interested in the biggest challenge or two that you've experienced so far in getting an AI project successfully out the door.
MI: The key challenge is the buy-in from our customers, and most importantly the trust level that we have to set with them, making sure we are fully transparent on the work that we’re doing. We need the AI models to be clearly explained, what they do and what data they receive, so there is no concept of a black box.
And another big challenge, I know we are working on one project which is based on predictive AI technology, and we do a lot of procurement for all our county agencies. And the need was, can you let us know ahead of time what will be needed in one year? How many devices, what type of devices, and what will be the projected cost for each agency? For that we needed to work on the traditional AI technology where we take in the past 5 or 10 years of data and see what we can predict going forward. But we didn't have enough data to be able to do the kinds of predictions that were expected of us. So again, data cleanup is something we have to do, and we have to go back and add more relevant data to make a better prediction. We've started to implement that change in our existing applications to collect that improved data, so our prediction algorithms work better in the future.
MP: It's interesting that you're not necessarily pointing to resistance inside of your organization as a barrier to entry. When we talk about AI, we tend to talk about the important difference between augmentation and automation. You know, AI is here to augment what people are doing, which is why you can get away with a common phrase like, well, yeah, AI isn't going to take your job. But if you won't use AI, someone who is may take your job. And I think we see a lot of great examples of that here.
"AI is here to augment what people are doing, which is why you can get away with a common phrase like, well, yeah, AI isn't going to take your job. But if you won't use AI, someone who is may take your job. And I think we see a lot of great examples of that here."
The future of AI in government and county services
MP: Over the next 10 years, what emerging AI technologies hold the most promise for county governments? Either based on the trajectory that AI seems to be going on right now, or in a different direction based on your experience.
MI: There are few technologies we definitely need in the government, definitely cybersecurity and monitoring. That is the area which is key for us in terms of protecting our applications, systems, data. So that is an area where we will be putting a lot of our efforts.
And another promising technology is predictive analytics, which can be used for resource allocations, public safety, and emergency response planning. By analyzing all these trends and patterns in the data, our county or other counties can proactively address issues like health crises, infrastructure needs, and crime prevention, and definitely help our residents.
Another AI technology which is used right now, and the future will enhance it, is the natural language processing through our advanced conversational chat bots that can enable residents to access information and services.
I know we've talked a lot about modernization. I know other counties may have legacy data and legacy applications and need to quickly modernize them, and I know we do as well. Modernization is the key to protect our data, and AI can do it easily.
MP: As we've talked about here, this is a very exciting new space and with the right planning it can make a big impact. Finding a safe place to start and having some guardrails and policy can certainly make it easier for organizations to get AI initiatives off the ground. Thanks for a really eye-opening session.
"Finding a safe place to start and having some guardrails and policy can certainly make it easier for organizations to get AI initiatives off the ground."
Getting started with AI for county services
Accelerating AI adoption within an organization involves integrating into digital workforce solutions and platform services. CAI provides AI solutions that automate routine tasks and provide the necessary infrastructure to support rapid AI integration and deployment. With AI accelerators, we help organizations quickly and efficiently embrace AI technologies, driving productivity and innovation.
To learn more about how CAI can help your organization successfully implement AI, fill out the form below.