If AI models aren’t relevant to business outcomes or easy to operate, few will use them. Success stories from D&A leaders at BDO UK and AptarGroup demonstrate three practices for achieving widespread adoption: curated workflows, smart business partner communication and convenient model interfaces.
Even the most brilliant AI model runs the risk of lying idle unless employees are excited about the way it will help them save time — especially if it’s a lot of time — or boost the impact of their work. They’ll shy away if it’s not relevant enough to business objectives or if it’s too complicated for nonspecialists to use.
Addressing these issues will help data and analytics (D&A) leaders with a problem that’s been vexing them for the past few years: scaling their AI initiatives.
BDO UK and AptarGroup, for instance, each applied a user experience (UX) lens to derive high-value insights for their business colleagues and customers.
Taken together, their practices illustrate three lessons for persuading employees to put AI models to work:
Target the parts of the workflow that slow down experts
Teach business partners about the benefits of AI models in a language they can understand
Make the model interface easy to operate
Observing how tax experts do their jobs, the D&A team at accounting firm BDO UK discovered a mind-numbing time trap: These employees spent nearly five hours seeking information from and making decisions for clients.
The main question that BDO’s tax experts answer is: “What is the optimal R&D tax credit a company can claim?” Most of the inputs that go into reaching a conclusion between the client and the expert are implicit and highly subjective (see Figure 1). BDO UK understood that its tax experts needed a way to help them make these decisions faster.
The organization’s D&A team worked closely with tax experts to curate a knowledge graph for this common question that could make judgments by inferring knowledge from an information base — and in just a few seconds. Then, clients could directly consult the graph for their information needs.
Because of these graphs, tax experts found more productive ways to spend their time, such as onboarding new clients or solving more complicated client questions. Also, since the clients were now interacting with the graph knowledge interface rather than a human expert, BDO UK could offer automated, customized and precise reports, as an additional product, attracting 30 new clients and creating more revenue for the firm.
Fabio DiMemmo, Aptar’s chief D&A officer, pivoted away from talking about technical complexities or incremental gains of the team’s AI model and spoke instead to a topic near and dear to the heart of the business — earning more money.
Aptar is a global leader in drug delivery, consumer product dispensing, and active material science solutions and services. Initially, clients would dispatch their experts (e.g., physicists, chemists)to a testing lab to evaluate the compatibility of their samples (e.g., cosmetics or medicine) with Aptar’s dispensers. The D&A team streamlined the process by creating an AI lab subscription, offering testing capabilities in a digital interface rather than a physical setting. This service made a huge difference for customers. Testing that took up to seven months in the physical lab took zero months online — the process was nearly instantaneous.
When D&A executives sought support for their pilot from business partners, they highlighted the significant revenue that Aptar would gain from faster client go-to-market speeds (see Figure 2).
This D&A service offering would allow Aptar’s clients to simultaneously test and develop their samples, an effort that would thoroughly align the streamlined process to the needs of its clients’ experts.
The client would make more money from getting the product into the marketplace rapidly. And because the client would be paying for the AI lab subscription, Aptar, too, would bring in revenue faster than before.
Aptar also kept the interface for its AI lab simple enough to be used by people who are not data scientists. Client physicists and chemists who worked with the model encountered three variables they could adjust to obtain precise predictions of the best container for the sample they were testing (see Figure 3).
The D&A team integrated basic UI/UX principles into the model interface, including:
A few inputs with a few outputs
Rapid changes — no lag time — when recalibrating inputs and outputs
Optimal packaging options suggested based on experience
All of this effort paid off. Once the business and experts embraced the AI lab subscription service, Aptar reduced lags associated with client product launches by 30% to 60%. Previously, clients had needed up to 24 months to get their new products on the market; the digital service reduced the speed to market by up to seven months.
Contact with questions or comments.
This article is from the .
Recommended by the Authors