Artificial intelligence (AI) is increasingly becoming a standard technology. From industrial manufacturing to office applications and complex CRM solutions, AI enables the automation of processes across the board. However, most companies still have little experience working with AI. Pegasystems Inc., a leading provider of strategic software solutions for sales, marketing, service, and operations, highlights the common mistakes made when implementing AI.

Seven mistakes when introducing AI

No real added value: Many companies are planning to implement AI simply because it is a buzzword, not because they can realize clear added value—for example, through higher customer satisfaction, cross-selling or upselling opportunities, process optimization, or improved quality.Lack of communication: If users are not informed that AI is being used, they may not even realize that they are interacting with an AI-powered system. This can lead to a loss of trust and rejection later on.Insufficient data: As a rule, AI systems identify recurring patterns in large amounts of data; if the dataset is insufficient, the results are irrelevant.Blind Trust: Even though IT systems can be astonishingly effective in some use cases, "Intelligence" However, not all use cases are suitable for AI; for example, when sufficient operationally usable data cannot be provided.Lack of transparency: Given the large number of influencing factors and methods used to determine AI results, it can be difficult or even impossible to trace, after the fact, how a particular result was arrived at. Depending on the use case, however, it may be necessary to ensure appropriate transparency—for example, when making treatment recommendations.Lack of flexibility: AI is designed in such a way that business units do not have sufficient options for customization; however, employees in those business units must be able to configure AI solutions on their own so they can experiment with different scenarios in an agile manner.Lack of practical applicability: Companies often implement AI solutions that work well in pilot projects but do not scale well in practice; the AI methods used must provide responses quickly in large-scale applications with numerous users—such as web infrastructures; as a guideline, response times of less than 50 milliseconds should be achievable.

A Cautious Approach to New Technology

"For many companies, AI is a new field, so we can’t expect the processes to work the way we’re used to with other technologies. Rather, we have to anticipate higher error rates here", explains Carsten Rust, Director of Solution Consulting for Germany, Austria, and Switzerland at Pegasystems in Munich. "This is by no means an argument against AI, but rather a case for using it with caution. It is important that users are aware of the pitfalls when implementing AI systems."

Artificial intelligence (AI) is increasingly becoming a standard technology. From industrial manufacturing to office applications and complex CRM solutions, AI enables the automation of processes everywhere. However, most companies still have little experience working with AI. Pegasystems, a provider of strategic software solutions for sales, marketing, service, and operations, highlights the common mistakes made when implementing AI.

No real added value: Many companies are planning to implement AI simply because it is a buzzword, not because they can realize clear added value—for example, through higher customer satisfaction, cross-selling or upselling opportunities, process optimization, or improved quality.

Lack of communication: If users are not informed that AI is being used, they may not even realize that they are interacting with an AI-powered system. This can lead to a loss of trust and rejection later on.

Insufficient data: As a rule, AI systems identify recurring patterns in large amounts of data; if the data set is insufficient, the results are irrelevant.

Blind Trust: Even though IT systems demonstrate remarkable «intelligence» in some use cases, not all use cases are suitable for AI—for example, when sufficient operationally usable data cannot be provided.

Lack of transparency: Given the large number of influencing factors and methods used to determine AI results, it can be difficult or even impossible to understand, after the fact, how a particular result was arrived at. Depending on the use case, however, it may be necessary to ensure appropriate transparency—for example, when making treatment recommendations.

Lack of flexibility: AI is designed in such a way that business units do not have sufficient options for customization; however, employees in those business units must be able to configure AI solutions on their own so they can experiment with different scenarios in an agile manner.

Lack of practical applicability: Companies often implement AI solutions that work well in pilot projects but do not scale well in practice; the AI methods used must provide responses quickly in large-scale applications with numerous users—such as web infrastructures; as a guideline, response times of less than 50 milliseconds should be achievable.

«For many companies, AI is a new field, so we can’t expect the processes to work the way we’re used to with other technologies. Instead, we have to anticipate higher error rates,» comments Carsten Rust, Director of Solution Consulting for DACH at Pegasystems in Munich. «This is by no means an argument against AI, but rather a case for approaching it with caution. It’s important that users are aware of the pitfalls when implementing AI systems.»

Pegasystems offers software applications for sales, marketing, service, and operations. Pega's solutions incorporate artificial intelligence (AI) technology.