The AI Endgame Is Decision Automation

Almost every organization has begun an artificial intelligence (AI) proof of concept (POC) in the past 12 months. With just 13 percent of AI POCs converting into projects, organizations enamored with AI must keep their eyes on the prize. Projects that deliver a tangible return on transformation investment (RTI) focus on decision automation—not just the technology in AI, but how these new systems move from agents to advisers.

Decision automation applies business rules, data analysis, workflows, and AI to automate the decision-making process in both operations and strategy. For CX leaders this could be knowing when to make ad buys for a campaign, change pricing for dynamic discounting, send follow-up texts for a future upselling/cross-selling, or check in on customer satisfaction after a new purchase. The goal is to take every end-to-end CX process and reimagine the five steps toward cognition:

1. Learn. Replicating the five senses, AI systems will collect contextually relevant information around them, including time, location, process, weather, business process context, heart rate, and eye tracking.
2. Understand. Applying some level of reasoning, systems will take into account the current environment and compare past interactions with future predicted interactions to find some level of reason and understanding in the business graph.
3. Recommend. Studying past behavior and adjusting for current conditions, systems will make a series of recommendations that will create dynamic signals to be used in future learnings.
4. Act. Acting on decisions takes the decision automation life cycle into reality and allows systems to determine the consequences of an action.
5. Refine. When systems understand the consequences of an action, they take the last step of decision automation: seeking to mitigate the false positives and false negatives of a decision outcome.

The path to decision automation requires a holistic approach and begins with creating an abstraction layer on transactional systems, including data, customer journeys, and user experiences. Most organizations have worked hard to relegate these transactional systems to a maintenance mode while adding context, identity, security, and intelligence to create the foundation blocks for intelligent orchestration.

The result has been a flurry of activity to build customer data platforms (CDPs) and tie them to intelligent processes and experience hubs to form intelligent orchestration services. These intelligent orchestration services enable organizations to form the business graph and multimodal models that will power decision automation.

By achieving a state of decision automation, organizations can deliver on personalization, AI, decision engines, and situational awareness. The traditional rally cry for 360-degree customer views can be achieved through decision automation, as can tangible effects on your bottom line.

In the most recent Constellation “State of AI in the Enterprise Survey,” 76 percent of respondents reported operational efficiency as their highest RTI. Basic CX automation is the lowest-hanging fruit for AI use among companies because of cost-cutting and increased revenue. As such, 73 percent of survey respondents also reported RTI in terms of revenue and growth. Cost reduction had the third-highest RTI (52 percent) among respondents.

Five Questions to Test Decision Automation Readiness

Early adopters from Constellation’s AI 150 executive network often ask these five questions during the design process for CX AI use cases:

1. Where and when do you insert a human?Most design aesthetics focus on when and where to automate. Determining when human judgment is required will provide a more effective and efficient design point.
2. Can you operate at machine scale with humans? Machines are making thousands of decisions per second. Humans might not be able to catch up, so how do you harmonize human scale with machine scale?
3. Do you have enough data to get to precision decisions? Achieving precision decisions requires internal and external data sources. For example, 85 percent accuracy in CX may be OK, but 85 percent accuracy in finance means someone goes to jail.
4. Who do you partner with to create the last mile or last inch of data? Organizations will have to partner for more and more data across value chains in order to achieve a high level of comfort and trust.
5. Who do you sue when something goes wrong?Does blame lie with the system, the operator, the partner, or another third party? 

R “Ray” Wang is the author of the new book Everybody Wants to Rule the World: Surviving and Thriving in a World of Digital Giants (HarperCollins Leadership) and founder of Constellation Research.