AI implementation veterans have plenty to say about their experiences with AI projects. They recommend determining clear goals, setting proper expectations for ROI and staying focused on user experiences, among other best practices.
Any business interested in transforming its operations with Artificial Intelligence should be prepared to answer several key questions: What are your implementation goals? How should the user experience impact functionality? How will you measure initial success in order to convince company leaders to move a project from pilot to production? And finally, if you could guarantee return on investment (ROI), what is your dream AI use case?
A trio of AI project veterans provided insights and best practices for these questions and others during a panel discussion at this year’s Digital Workforce Summit (DWS). Moderated by IPsoft Chief Commercial Officer Jonathan Crane, the audience of 600 executives heard from representatives of Alvarez & Marsal, as well as a wealth management expert who worked previously at a major global bank.
Focus on User Experience and Ease of Use
Jean Hill, Managing Director at Alvarez & Marsal, emphasized the productivity gains that come with implementing digital labor. Alvarez & Marsal monitored employee productivity and how IPsoft’s Amelia and IPcenter were used. The company did some draft calculations to determine how much faster work could be completed. “Ten minutes a day times 30,000 call center agents, times two to 250 days a year, and it was in the hundreds of millions of dollars,” said Hill. “We also look at showing people the true productivity of the tool. So, not just on what it costs to make, but really thinking on the value that it is giving.”
Improving the user experience was the primary concern for Tom DeCarlo, a longtime wealth management professional with substantial experience in AI projects. Whether dealing with internal users or customers, AI should first make tasks easier, particularly in financial services firms. “Keeping [users] happy with tools, with products and services that are best for them and their clients…Having a tool like AI is the differentiator at some points,” DeCarlo said.
Win Stakeholder Approval
A successful AI pilot project means very little if it doesn’t get approval from the C-Suite. Learning the best strategies for convincing stakeholders that an AI project is a worthwhile investment is crucial, as DeCarlo explained while reviewing one of his AI projects.
“I think when we initially introduced Amelia into the organization, we did so with a complex proof-of-concept pilot around retirement, death distribution, and document requirement for beneficiaries,” said DeCarlo. “This is a task that would take a 15-to-20-year employee's knowledge and experience to answer, [especially] when you have 25 to 30 different documents that a sales assistant would need to have a beneficiary sign in order to take control of the assets in the account.”
DeCarlo said the use case was composed of a set of tasks that human workers struggled to manage. Customers would call in and a human worker would provide information that the customer would then use to find additional information via other channels, such as online or on another phone call. “Two, three, four calls at a time CSAs [customer service agents] would have to make because we were not giving them the right information,” he explained.
DeCarlo’s team had a 100% success rate with Amelia during first interactions in the proof-of-concept pilot, and user satisfaction improved by more than 20%. “Amelia was filling out the documents for the CSA and at the end of that conversation [she was] emailing a PDF to the CSA,” he explained. “The CSA just needed to have the client sign it, and it was done. That was it.”
To get the project approved for a company-wide production, DeCarlo worked with the IPsoft team to create a presentation emphasizing the success of the three-month pilot project. DeCarlo and IPsoft also created a video outlining the experience and the interactions between Amelia and sales assistants. “That movie went all the way up to our CEO, and within hours I had conversations across the globe with my peers on how they can help to take part in this journey,” DeCarlo said.
Prove AI’s Value
Hill agreed that proving the value of AI for tough projects helps to convert pessimists to evangelists — even if it takes additional effort compared to a more straightforward implementation.
“We'll tackle the tough things first, not the easy things,” she said. “You have to understand who are the naysayers, and how do you show them that this technology will work in their organization? We look at the organization and then an initial use case. Where is the elapsed time to solve a problem?... Show them an elapsed time, and show them that you took a task from three hours to nine minutes.”