Think big for lasting results with AI – MeriTalk
Artificial intelligence (AI) can help make small improvements in low-profile agency operations, or it can be a game-changer by helping agencies fulfill their missions in ways we can’t imagine today. MeriTalk spoke with David Kushner, Executive Vice President of Sales at enterprise IT solutions provider ViON, to discuss how agencies can move beyond narrow use cases to achieve lasting change. and impacting with AI.
MeriTalk: There are many use cases for AI technology, from detecting and assessing cyber threats to building knowledge libraries to improving customer service. Where can AI bring the most benefit to federal agencies in the short and long term?
Kushner: Use cases permeate the federal government – from identifying people, places and objects for defense agencies to tracking COVID-19 cases and forecasting the weather for civilian agencies. Today’s agencies tend to focus on narrow use cases, like predictive aircraft maintenance, and a single bottom line. These can be valuable, but agencies usually don’t invest a lot in them. They are temporary and small.
The real long-term strength of AI is its overall capabilities that can support multiple use cases. We may want to achieve a certain result, but the data can tell us something completely different. That may bring up five other results we never thought of – and those results can help us further the agency’s mission or advance technology. So let’s build real AI practices. Let’s hire and train data scientists. Let’s build systems that have multiple uses. Let’s understand how to build a model, train it, keep it up to date and use it in our everyday life.
From the White House leadership to the bottom, they promote the idea that government needs to be a leader in global capabilities, not just one-off projects that rely on a vendor. ViON wants to encourage this kind of innovation. This is the way forward to where we want to be as a country.
MeriTalk: What do federal agencies need to help meet the goals set out in the national AI strategy?
Kushner: AI is very successful when agency analysts who understand the agency’s data and mission then train in machine learning. The agency’s time and money is well spent training existing staff or re-equipping roles before hiring. If you don’t understand the agency’s data, use cases, and business processes, you can’t move forward with AI. The best AI system in the world wouldn’t help.
MeriTalk: For agencies with little AI experience, what are the first steps they can take to start using the technology?
Kushner: Our first conversation about AI often ends up being about storing, accessing data, and understanding your data, as massive amounts of data make AI models accurate and enable continuous improvement. Without it, you are just spinning your wheels. We always tell organizations to start small and go as slow or as fast as their expertise and infrastructure will allow.
Sometimes we ask an agency, “Where is your data? And they say, “We’re not really sure, and we don’t know if you’re allowed to access it or combine these two data sources.” This can kill an AI project or cause a delay of at least six months.
MeriTalk: What are the main hurdles for federal teams to overcome when implementing AI pilots and then trying to scale those pilots company-wide? How can they overcome these obstacles?
Kushner: Outsourcing is a very tempting path, but this approach can become a sandbox. The contract lasts for a year, and when it’s over, you haven’t built any capacity in-house. Even agencies that are dedicated to AI are caught in this cycle. That’s why some of the bigger agencies appoint AI news executives who understand how AI aligns with the agency’s mission and what success looks like.
We are also seeing more and more companies moving away from the ‘black box’ approach of AI. They are working to make AI frameworks and tooling systems more universal, making it easier for government and industry to machine learning and deploy models. This approach will become more and more popular.
MeriTalk: More and more agencies are moving their operations to the cloud, but you are advocating the production of AI outside of it. Let’s explore this idea.
Kushner: For the lowest cost and greatest benefit, an AI production environment must be on-premises or behind a firewall in a colocation facility due to the scale and volume of data. I always say develop in the cloud and operate at scale on premise.
This is because AI systems are demanding and computationally heavy. They need large amounts of data and need it fast. AI systems tend to be 100% used for a single use case for hours, days, or weeks. They don’t align well with the cloud-as-a-service model of paying for the flexibility of temporary use.
We encourage agencies to take ownership of the systems they will use 90% of the time. This can be in a colo, on site, a direct purchase or a subscription. A recent cost review we did for a customer showed that owning all the equipment and using it for four to eight years would cost less than a quarter of what they planned to spend on the cloud.
MeriTalk: How can ViON help federal agencies adopt and implement AI solutions? What sets ViON apart from other solution providers?
Kushner: We design and implement AI solutions from a wide variety of vendors. We partner with all the major OEMs and many small AI stores. In fact, we have brought several small innovative AI providers into the federal market. We are independent from suppliers. We look at the client’s challenge, infrastructure and expertise as well as the outcome they are trying to accomplish, and then choose the right solution. We have the experience and expertise in AI and machine learning to be able to make that decision. We also have extensive experience in managing data, securing it and making it accessible for complex initiatives.
Our clients appreciate that we generate results more than technologies. We help agencies that want to roll up their sleeves and build their capacity – not just pay a million dollars over five years and have someone else do it. We do pilots and we have internal laboratories. We always have our hands on the keyboard alongside our partner agencies in their data centers. There are many great companies that are doing a good job in the AI arena. But a lot of them are an idea and a webpage, and they provide remote assistance. We are actively involved in the success of the mission. This is why we incorporate the people from the agency who will appropriate the solution, even if they do not feel they have the experience to do so. When we get their membership, the agency is more successful in the future.