How do you balance winning support across the business for implementation of radically new technology with the risk of unrealistic expectations being set by the community you are trying to influence? This is the dilemma facing many digital leaders and CIOs leading breakthrough implementations of AI.

Artificial intelligence and its role within business operations is still an evolving topic. While enthusiasm for adoption is high, it’s critical that the first experience of AI is positive. Rushing into an implementation without defined goals from the start, with a profoundly transformative technology like artificial intelligence, only sets up a deployment for failure. Here are some questions potential AI pioneers should be asking themselves before mapping out unrealistic objectives:

“What are the short term and long term gains we desire?”

While this an obvious question to ask with any endeavor, often times the ultimate goals can be lost as people start envisioning the “art of the possible” with AI. The management of expectations is hugely driven by the sort of adoption environment an AI-powered solution is placed into. Similar to what’s mentioned in Ovum’s latest report “IPsoft’s Amelia: More than a chatbot”, are executives at a company more enthused over cost-savings, or are their objectives much more ambitious than simply reducing operational costs?

Of course, AI can deliver immediate monetary impact to an organization, but the real ROI comes in down the road—cost savings is only scratching the surface of AI’s range of possibilities. As a technology capable of completely altering the way customers interact and how business processes are handled, artificial intelligence possesses the potential to not only refine one aspect of an operation, but rather transform and create a new type of business ecosystem all together.

“Now, do I have an ideal use case in mind?”

Once an enterprise is ready to take on an AI implementation, the next step in the process involves choosing what will be the best possible use case to start off the pilot program phase. The key is to begin small and create a roadmap for the future where more business cases can stem from the initial one. The emphasis on starting with a small use case comes from a place of caution because for most companies the deployment of artificial intelligence is a first and there is no precise way to determine how the presence of AI will affect adjacent business areas. It’s a learning experiment so to limit the scope of any unwanted side effects, use cases should be kept small and controlled.

It is also important to make sure an impactful business case is chosen. Select a use case that will turn heads upon its successful completion and generate more momentum behind the AI initiative. For example, Tom De Carlo, Managing Director, Head of Client Services at UBS, explained during our Digital Workforce Summit that instead of going for a simple use case with Amelia, he opted for a more challenging task that showcased her cognitive capabilities. Amelia was trained in a pilot to assist customer-facing assistants in working through the information needed to determine which documents need to be filled out by a policyholder’s family members in order to disburse a deceased customer’s funds. For UBS it was important to show that Amelia could add value even in demanding and complex roles. This has paved the way for teams to assess Amelia for additional roles within UBS.

As this prime example depicts, by building internal confidence in the application of AI, executives can expand their program into other business areas and encourage collaboration on how AI can transform processes.

Introducing AI into a business operation is a delicate process that needs to occur incrementally and, more importantly, the vetting process for electing the ideal use case should be as thorough and well-substantiated as possible.

“Do I have the right team and do they have the bandwidth?”

AI’s long term effects are most prevalent in the customer experience space where future competitiveness is set to be decided. To formulate a CX approach that will differentiate itself from competitors and exceed customer expectations, the foundational work needs to be done now. To fine tune an AI-powered cognitive agent to interface with customers on a variety of customer service topics, subject matter experts will need to be involved directly in training the new virtual employee on how to resolve different business scenarios.

These subject matter experts need to be complemented with technical support from the technology provider so at least in the early going, the deployment of a digital employee can proceed smoothly. An initial deployment team calls for implementation architects who will prepare a virtual agent for live interactions by refining its conversational readiness and project leads to ensure the right benchmarks are being reached as the go-live date approaches. Without a well-built and orchestrated team, a virtual agent could take several months to deploy or worse yet, could be sent into live production prematurely.

Augmenting your own implementation team with an AI provider’s skills and expertise is something every first-time adopter should strongly consider. By focusing on impactful outcomes and leveraging an internal and external set of expertise, the benefits of artificial intelligence can be had in your enterprise.