The phrase “artificial intelligence” was first coined in late 1950s. Over the years, advancement in AI research has introduced many enhancements to our daily lives. The fun body movement game such as Microsoft’s Xbox Kinect and the virtual assistant, Apple’s Siri are two creations that have millions of us welcome AI technology.
Today, AI is embedded in software applications, robots, and more. Businesses are exploring how to leverage and incorporate AI into their business model. Benefits such as productivity gain and innovative ideas for product development are difficult to ignore.
There are several considerations as you explore AI opportunities that would bring meaningful business outcomes.
Business problem and value
The value of AI comes from solving a sticky problem that has costly impact on the business.
Lyft uses an algorithm to maximize revenue by matching ride requests with driver’s locations based on a number of factors. With the help of AI, Lyft uncovered that conversion rate is an important determinant of ride order frequency. As a result, Lyft refined its algorithm to include the optimization of conversion rate when assigning drivers to ride requests.
Labour cost makes up a high percentage of the total operating expense for many businesses. A well trained chatbot would help to minimize the need for live agents. The cost savings could be significant.
However, most chatbots can only answer basic questions. They stumble with complex questions. Subsequently, customers get cycled through the same responses over and over again. This leads to frustration especially where there isn’t a contact number customers can call.
In this case, the intent is good but the AI solution is poorly designed so it doesn’t generate the intended value.
Reliable data
Training an AI model requires reliable data for generating trustworthy output. The more complete and accurate the data is, the AI model would produce output with a higher accuracy.
Using a generative business intelligence platform to perform analytics enables the user simply type in a question in plain English and get a response within a minute.
For example, a sales manager wants to know the close rate for each sales rep for the quarter. Shortly after the prompt is entered, the AI model returns a graph showing the performance of each sales rep. This removes the need for the manager to learn how to set up the data elements for the analysis and design the graphical presentation for the results.
As many businesses struggle with fragmented data, they need to have a data strategy for handling historical data and managing the lifecycle of data better going forward.
It is important to note that more data is not better but noise to the AI model you want to train. Reliability of the output is a key metric businesses need to look at when consuming AI generated output.
Caveat on hallucination
ChatGPT has demonstrated that AI can generate meaningful content related to a topic using the massive volume of data it accesses through the internet. The application has proven widely to save time needed to write a product description, compose an email, and generate text content in general.
Nonetheless, hallucination could create misinformation. For example, Microsoft had to remove an AI-generated blog on its travel site. The post suggested visitors to Ottawa go to the Food Bank with an empty stomach. Microsoft claimed the post was an unsupervised AI-generated creation.
In the medical world, AI is helping to discover new drugs and detect tumors. However, doctors are cautious about deploying AI for patient care. Without full transparency on how the AI tool is built and how big and reliable the training data base is, it is concerning to put full trust in AI programs.
As a result, it is worthy to exercise prudence in evaluating the effects of hallucinated output. In the event that the AI model is incorrect, a business needs to be prepared to correct course quickly. Undesirable effects include personal harm, societal impact, and breach of privacy.
Despite current efforts by governing bodies to develop legislations to manage AI risks and potential harms, businesses need to be cognizant of them and practise responsible AI.
While it is important to explore the possibilities and potential benefits, let’s not forget that AI is not necessary the best approach to tackle every business problem. It is worthy to consider other options and compare their pros and cons. It is an investment to develop an AI model, feed the proper data to train it, test and verify output. Business acumen is still needed to make the final decision. It would be best to treat AI investment just like any other investment options, determine the criteria for evaluation of benefits and risks, and do an objective assessment.
Leave a Reply