Insight By Presidio

How realistic expectations lead to greater success when getting started with AI

This content is provided by Presidio.

Executive Office of the President signed an executive order in February expanding his administration’s efforts to foster the research and development of artificial intelligence tools in government, and agencies have wasted no time getting on board. They’re all eager to reap the benefits of this new technology, and for good reason.

“Much time and money can be saved, and business insights gained by taking advantage of the data analysis Artificial Intelligence is making possible,” said Michael Law, Solution Architect at Presidio.

With the massive explosion of data being collected, stored, analyzed and put to use in the federal community, the capacity for humans to operate at such a scale is beginning to fall behind. So agencies are looking to technology to pick up the slack.

“Government is starting to realize that it needs data touched less by humans,” Law said. “That’s what is needed because that’s what slows things down.  Enter AI.  With the right data and the right approach to the analysis of that data, that’s where business will realize new information and intelligence that was difficult for human processes and project timelines to yield.  Humans can then focus on other mission essential tasks.”

One area that agencies are exploring applying AI to supplement human interaction is in customer service. Law compared the process to self-checkout lanes at grocery stores. The stores need to hire fewer cashiers, and can put its employees to work on other tasks by automating the checkout process, while analyzing customers product purchase preferences.

In the Federal space, AI is helping to enhance the customer service processes.  AI “chatbots” are being used to answer customer queries.  This saves time and money, and customers are happy with the assistance received.

“Maybe 10 to 15 years ago, there were many regional offices offering walk-in service centers.  Customers would walk in, and they could talk to a human and get the help they needed.” Law said. “And now, if you look, there are fewer processing centers, and it’s all because the work is done smarter with AI. It’s not as manual, and it’s more automated, and AI is what is used to automate these processes.”

It’s also supplementing human interactions that do still occur by giving feds the ability to access, analyze and put to use more data quicker.

“Why would you want to tell a customer about using a specific form or process if you already knew from past interactions and related data, that’s not going to apply?” Law said. “So it’s really the use of AI and Machine Learning where a larger pool of data that has been accumulated over time, and prepared for that use.”

But as excited as federal agencies and managers are to get their feet wet with AI, the expectations don’t always match the reality. It’s easy to find big flashy examples of AI being put to use in the FBI’s forensic ballistic analysis, or genetic research at the Department of Veterans Affairs, or facial recognition by Customs and Border Protection at ports of entry.

But these kinds of applications are the exception, not the norm. Most federal agencies are finding that AI makes the most sense for them when applied to business processes and workflows, automating repetitive tasks to free workers up to do jobs that require more critical thinking. So a good AI consultant will work with agencies to temper their expectations and find realistic applications that can scale, instead of trying to jump straight to the moonshot.

“We go with a couple of words of encouragement. First is start small, you don’t want to try and boil the ocean with that first attempt,” Law said. “And then second of all, we always tell them let’s make sure that what you’re doing is realistic with the data sets available. What we try and do during our consultative approach is eliminate failed attempts, wasted time and wasted money, which could result in a stalled or failed project”

The smaller successes tend to have a way of building on themselves. A few minutes saved here and there with robotic process automation adds up to hundreds of man-hours when applied across an entire workforce. Customers who spend less time filling out forms and get served quicker raises satisfaction rates with agency services. And successes like this, proofs-of-concept, can be used to justify more spending on similar projects, or scale existing ones.

That’s why it’s important to tailor AI capabilities to the specific needs of the agency, and take a realistic approach to how the technologies can be applied. From there, it’s just a matter of working with consultants to fine tune the technology in order to get the optimal results for the mission.

“What we always try to do is meet or exceed our customer’s expectations?” Law said. “And if not, let’s talk about why that was and how we can maybe change the way we looked at your data or how we sampled your data to be able to get closer to what your objective was.”