Output Is Not an Outcome: Resolving a Misperception About AI in CX
There is no question that artificial intelligence is a game changer: It is plain to see the impacts AI has already had across business and personal life. And from time and financial savings to leaps in knowledge, analytics, and insights, people are just seeing the tip of the proverbial AI iceberg. But unlike the captain of the Titanic, business leaders can deftly navigate this iceberg to avoid a catastrophic crash. Part of that steering is understanding and accepting that not all output from AI is an outcome. To conflate the two is the first point of impact.
AI outputs are nothing to be scoffed at. Even the narrowest application of AI set to analyze a customer’s buying patterns over time to recommend the next best product is valuable. But that recommendation—the product that best aligns with a customer’s behavior, previous transactions and current buying posture—is an output. Its value is incremental. The outcome is achieved if the customer is compelled to take up the recommendation and make the purchase. The recommendation itself is not the end goal.
Calculating the value of the recommendation can only be made by understanding the totality of outputs required to achieve the optimal outcome. There are any number of AI-driven outputs that will be required to reach an outcome that has real value to the business. And there are any number of connected costs that could turn that simple recommendation into diminished margin.
Output, by definition, is an action or product that is produced, or power or information that leaves a system. It’s the shirt made by the seamstress. It’s the email sent by the marketing automation tool. It’s the answer generated by the AI
assistant.
OUTPUTS LEAD TO OUTCOMES
On the other hand, outcome is the result or consequence of an action. Connecting the thoughts, information, or action that leaves a system, when directed and orchestrated intelligently, will lead to a positive consequence. The difference between an output and an outcome seem, on paper, to be clear—and for those reading this a bit elementary. So why have we landed in an AI messaging landscape where the output of AI is so often being mistaken for the outcome of AI?
As generative AI solutions have been deployed across customer experience functions, the mere result of conversational output can feel like an AI outcome. Conversations can be ingested and interpreted by systems almost as quickly as
they can be spit back out, making conversational self-service for the customer a reality. Agents that work as assistants to already stretched-thin teams highlight where and how time can be manufactured as the mundane and repeatable are lifted from employees’ plates. But even with this shift in work, AI output metrics continue to dominate conversations, as calculating the amount of offset work, deflected human conversations, tasks completed, or tokens consumed have begun to define the value of AI-driven transformation.
CX LEADERS NEED AN AI STRATEGY
CX executives are undoubtedly ready to embrace AI, especially agentic AI with its promise of automation with context and reasoning. In a recent poll conducted by Constellation Research, 56 percent of respondents said that implementation of more generative and agentic AI is in planning stages or is already deployed. But 38 percent of respondents are still on the fence, noting that their organizations could very well be ready to advance an agentic agenda, but questions must be addressed. Some of those questions include if they will ever see a return from their AI investments. So far, of executives who have overseen AI shifting from experimentation to production, only 20 percent feel their returns have met or
exceeded expectations.
But 41 percent of executives also admit that they lack an organization-wide AI strategy, leaving their agentic visions to flounder. More than a lack of vision, organizations have put the AI cart before the outcome horse, as 43 percent of respondents agree that they have no clear vision of what the business outcomes of AI should be. Far too many leaders are
reverse-engineering results and impact once the outputs of AI have been delivered. So how do we stop this cycle of misinterpretation?
Start from the top and work down. When outlining AI outcomes, start at the top of the hierarchy of business need, typically the overarching revenue, margin, and profit goals of the organization. What specific workflows, automations and actions will most directly impact these objectives. Work down from there to identify strategic outcomes for the function. If the goal of service is to lower operational costs, increase customer lifetime value, and impact return rates to improve margins, identify specific workflows and automations that, through AI, deliver predictable and quantifiable outcomes.
Finally, dive headfirst into the functional goals of the department. This is where the outcome identification will likely be easiest, as these functional goals, in the contact center, have long included improved call handling, the number of service
requests managed per human agent or AI agent.
Far too often, leaders start to apply and, in turn, quantify AI impact from the bottom up, tallying the functional savings and the human headcount that can be reduced, redeployed, or redirected to new assignments once AI-driven self-service automation can take over simple tasks. The real beauty of AI is that thanks to its capacity to ingest, analyze, and normalize
complex data sources and types, it can extrapolate far beyond human capacity, so why limit AI or humans to churning outcomes faster? Why not unleash both critical resources to collaborate so that this new work can be a force to be reckoned
with?
Become the center point for market and customer knowledge. Shift left in thinking about the output of AI by thinking of that output as a renewable resource constantly feeding enterprise knowledge. Knowing more about the customer and their definition of value shouldn’t be locked away in a functional CX solution. Thanks to AI’s capacity to ingest, curate, and contextualize customer conversations to better understand the meaning, intention and reality of a customer’s connection, the supply chain of insight and knowledge now spans across the entire front line of experience, synthesizing the customer voice into a real-time intelligence asset.
Automating time-consuming tasks like call summaries and follow up emails is the performance-driving starting output. Where this leads is a company-wide strategy where AI analyzes and surface unified intelligence across the entire organization: Service, for instance, has access to shipping and supply chain information that impacts the customer; supply chain and shipping has access to information around potential points of customer friction; and marketing shares intelligence on market shifts and product behaviors. Such unity results in a fully connected business that can collaborate in real time with a customer who is likely asking their own agent to do some of the tedious shopping on their behalf.
If organizations stay rooted in outputs and lose their vision of how to impact business outcomes, the customer, already innovating and ideating ways to leverage AI to simplify and free up their own time for more creative and rewarding tasks,
will leave the experience behind.
Outputs—no matter how hard we try—can’t live up to being business outcomes. The hard truth for many will be that the reason AI is failing is a lack of clearly defined business outcomes required across functional, strategic, and corporate needs before a single output has arrived. We need to know where this bullet train of AI is heading before we can claim
we have navigated there successfully.
Liz Miller is vice president and principal analyst at Constellation Research, covering the broad landscape of customer experience strategy and technologies.