AI for business users: a glossary

Artificial intelligence, or AI, is still a learning experience for many business users. One way business users can better communicate about AI with IT and data science staff members is to gain familiarity with frequently used AI terms.

When you work with IT staff and data scientists, they’re going to use acronyms that you might not be familiar with. It’s important to know some of the basic terms and acronyms so you can communicate.

Business users should make themselves familiar with these common AI terms to communicate well with the data teams.

AI (artificial intelligence)

Artificial intelligence is a form of intelligence demonstrated by a computer. A computer can be programmed with logic and business rules that will enable it to “reason” through situations and come up with a conclusion. The computer cannot think for itself, rather it understands patterns and rules.

An example of this is a loan decisioning software, which can analyze the financial standing of a loan applicant and determine whether it makes sense to make a loan to that applicant, and at what interest rate. The only way the decisioning software can do this is by using a set of business rules and lending criteria that the bank’s lending staff has provided. An IT person or data scientist will refer to this artificial computer reasoning process as AI.

SEE: Success in implementing AI depends on how well companies define interactions (TechRepublic)

ML (machine learning)

The other thing that a computer using artificial intelligence can do is to recognize patterns in the data that it processes that suggest certain outcomes.

For example the computer might notice that a majority of persons from a certain zip code always buy beef when they go to a grocery store, but in another zip code area, the buyers seem to prefer chicken.

Retailers use this data for purposes of sales promotions and inventory management. When a computer program performing AI notices these patterns and incorporates what it has “learned” into its logic, this is known as ML, or machine learning.

Data aggregation

Most artificial intelligence programs use more than one source of data. This is what makes AI programs different from standard report generators in finance, which only report on the general ledger, or a sales report, which only reports on sales.

Because AI requires information from many different data sources in order to analyze and to make decisions about that data, IT staff and data scientists have to blend all of this data from different systems together into one central database so the AI program can process it. This process of ingesting data from many different sources is known as data aggregation.

ETL (extract-transform-load)

In order to move data into a central database from many different data sources, IT staff or data scientists must first extract (E) the data from each incoming data source, then transfer (T) the data into a format that the target database will be able to accept, and then load (L) the transformed data into the target database that the AI will operate on. This entire process, which is largely automated and based on data transformation rules that IT or data scientists provide, is known as extract-transform-load, or ETL.

Algorithm

When you use a standard report, you submit a query like, “How many widgets did we sell in the first quarter of this year?”

When you use an AI program, you can still ask queries, but the AI is really designed to answer more complex questions, such as: “For a patient with this set of symptoms, what is the most likely diagnosis?”

The AI program uses an algorithm — a sequence of carefully defined instructions that use logic and mathematical tests — to operate against a diverse set of data, then derives a conclusion. Because of the complexity of AI, an algorithm goes far beyond what a simple report query would ask for.

SEE: Momentum for AI is building across organizations (TechRepublic)

Learn the language to deepen the investment

In 2022, nine out of ten businesses are invested in AI technologies, but less than 15% deploy it.

There is clearly more work to be done—and the lion’s share of that work will be done by business users who understand where AI can most productively be used in the business.

To implement use cases that optimize AI, business users can familiarize themselves with common AI jargon. This enables them to more effectively communicate with IT and data scientists, who might understand the technology, but may not know the best business use cases to apply it to.