You probably have already heard of Artificial Intelligence (AI), and most of you are likely already involved in or driving at least one program trying to bring the value of AI to your enterprise. The purpose of this post is not to explain the benefits or the differences between various acronyms that you come across these days. You are going to be influenced by at least one initiative that is centered around AI, and this post is geared toward providing a framework that will help you measure and compute the value of these initiatives over time.
Like every other initiative within the enterprise, AI initiatives should be measured by their impact to the two most important and tangible aspects of enterprise reporting:
- TOP LINE: Improving the revenue of the enterprise
- BOTTOM LINE: Improving the earnings of the enterprise
While these two are the most obvious measurements, it is important to understand that these are not the only measures of determining value of an AI investment. To explain better, let me illustrate the lifecycle of a typical AI program.
Identifying the Input for the AI Program
At its core, AI enables decision making to be performed by machines with the use of data. A good AI program should start with identifying the input.
This means, identify the decision that would be augmented, helped by the machine, and the data that is required for enabling and training the machine to make the decisions.
Once the inputs are identified and analyzed, the output provides for a hypothesis of how the decision making by the AI models can be accurately developed.
The Missing Link to Business
Unfortunately, many programs start with the technology and the hypothesis in decision making, when they should be looking at the business benefit of the decision being made. Most decisions can be aligned to the following:
- A machine augment improvement in decisions effecting a customer journey eventually impacting the top line
- An improvement in efficiency of employees and vendors within an Enterprise through automated decisions and processes eventually impacting the bottom line
It is important to prioritize the impact to business before Identifying the input to the AI initiative is resolved.
Impact of Augmented / Automated Decisions on Business
Enterprises create the ability to provide digital scale only when machines allow for or assist in decision making. Companies like Amazon, Uber and Netflix use this intelligence in enabling systems to make real-time decisions often in only seconds. Without these systems, it is practically impossible to scale a business for the digital economy.
These decisions made by machines impact the nature of business and provide a real-time view into the operational performance of the Enterprise. Unfortunately, the effectiveness of these decisions are measured and end up adding to the measured business benefits of top line and bottom line improvements.
From my experience delivering AI solutions to many clients, it is critical to create a framework for measuring the impact of these decisions.
We look at the impact of decisions as leading indicators and the business measurements as lagging indicators. Since each Enterprise is unique and operates with a unique value proposition, the impact of the leading indicators against the lagging indicators can vary.
AI programs within Enterprises need to establish a co-relation between the leading indicators and the lagging indicators.
Reality of AI – Time Consuming Process
The effectiveness of Artificial Intelligence within an enterprise is, unlike many other digital technologies, a slow process. It requires coordination and change across the enterprise, with an understanding of the purpose and the objectives of the program. The process typically involves:
- Identifying data for training the models
- Implementing and tuning the models
- Integrating the models into the business process
- Manual review of the decisions (starting with a full comprehensive review, and gradually reducing to ~25% over a 3-9 month period based on variety and effectiveness)
The effectiveness of AI solutions can take between 6-12 months to present, based on the complexity of the solution. Throughout the process, measurement is important to ensure that the program is meeting its objective and is being tracked continuously. Effective measurement ensures effective use of AI within the Enterprise.
The table below captures the metrics for implementing AI for operational efficiency using automation:
Decision Confidence Scores
% of exceptions
Average Cost per Process / Transaction
Processing Times per transaction
Quality of Data
Improvement in Decisions
Reduction in Time
With the tangible benefits discussed in this post, it is important to also understand the intangible benefits, such as reduced training times, from Artificial Intelligence that enable the Enterprise to scale over time.