Artificial intelligence has a mixed record of success and failure in business processes. What do you need to do for a successful AI implementation?
On the very first business process design project, my team was tasked with redesigning an internal invoice process in accounts receivable. Issues related to duplicate systems and duplicate BPs that extended invoice processing for days. Hand in hand with finances, we looked at every step of the process, identifying and eliminating duplicate data and processes. The result was a more streamlined billing process that went from three to one day.
I often think about that first BP redesign assignment and how easy it was compared to redesigning BPs with artificial intelligence.
SEE: Ethical Policy for Artificial Intelligence (Tech Republic Premium)
The difference between BP design and AI is that you do more than streamline business processes. You radically change how employees work. This makes the BP change project both a human problem and a process and system change.
IT and business leaders need to be sensitive to this, or a new and redesigned business process could fail if employees resist.
Four best practices for designing business processes with AI
Never start without a strong business use case
This use case must be defined and purchased by the end-user department using the business process. Is there a user pain point or productivity improvement goal that a BP redesign with AI can solve? And if AI and automation are brought into the business process, will the end users support it?
Give the AI enough information so it can do its job
The rules and data provided to an AI decision-making engine come from subject matter experts and the end-user employees who understand the intricacies of an existing BP and the decisions and operations that BP makes and executes. You can’t inject AI decision-making and execution into a business process if you don’t have the expertise required to define AI’s operating and decision-making rules.
Determine the impact of the AI on the system
Any system that interacts with the AI in a revised business process needs to be modified to receive and send data to the AI. One of the most common mistakes in AI BP design and implementation is forgetting about a downstream system of the AI being adversely affected because someone forgot to build the necessary interface between the AI and that system.
Determine the human impact of AI
There is probably no step more important than this one.
Users may initially be excited about AI and automation taking their routine tasks off their desks, but they won’t be thrilled for long if they think the new AI is going to eliminate their jobs.
If job losses are expected, it is best to consult with employees beforehand. Try reassigning them to other jobs in the company.
If there is no impact on job losses, but instead on a new AI-infused BP that requires workers to be retrained, the retraining should be undertaken aggressively and enthusiastically supported.
Employee wellbeing is a vital outcome of any AI infusion into a business process.
