Analytical AI & Generative AI: Why, What, How

The hype about Generative AI with ChatGPT & Co. and now the publicity around AI agents lead to confusion what AI really is. This distracts from how Generative AI, and even more Analytical AI, can help to increase profit for any company. It’s then not surprising that executives wonder what business value AI can deliver.

This even gets worse with technology vendors coming with AI from the technology point of view, instead of starting with business needs. But the question for executives is simple: How can data and AI be used in my company to reduce expenses or increase revenues?

With this hype, Analytical AI, which has been delivering benefits for companies for 20+ years, is overlooked, despite delivering 4 times the business value than Generative AI, according to McKinsey. In other words, of all the business value that AI delivers, roughly 80% comes from Analytical AI and only 20% from Generative AI.

So, why should you use AI in your company, which AI for what, what are the limitations, and how are the different forms of AI implemented?

Why should companies use AI?

Artificial intelligence (AI) is a machine’s ability to perform cognitive functions we associate with humans. There is no AI today that can perform all these cognitive functions, while different forms of AI address different sets of these functions. Even if they are still limited, these AI can already add business value today.

Analytical AI is useful in replicating the cognitive functions of problem-solving and visual & audio perception. It can analyze patterns in data and predict what may happen. Generative AI is good in natural language processing and, if you are generous, creativity. It can summarize existing content and generate new one.

What are the business benefits of using AI?

To be more specific, here are a few, non-exhaustive examples of what AI can do for business. Analytical AI is useful in forecasting customer demand, which helps the company to produce sufficient goods to meet that demand. It can create customer segments to identify the most valuable customers. It can optimize marketing spend to achieve better ROI. It can recommend maintenance activities to reduce production downtime. And, it can detect anomalies to minimize losses due to fraud.

Generative AI can summarize text, reducing time to synthesize long documents. It can inexpensively create images for a marketing campaign, reduce time to document software code, and generate new design ideas in R&D. As a chatbot, it can add personalized customer experience in customer service.

What limitations does AI have?

AI is trained on data or is using historic data created by humans. Humans are not perfect, so data is not perfect either. “Garbage in, garbage out” applies to AI too. When using AI, it’s critical to understand limitations of each AI.

Analytical AI could provide inaccurate predictions if the underlying data is incomplete. And, it could provide biased outcomes if the data used is biased.

Generative AI could provide inaccurate outputs if it’s trained on inaccurate and unvetted data. Since the Large Language Models (LLM) underlying Generative AI provide text that is based on the probably of words following other words, these models may hallucinate. LLMs may also violate copyrights if they are trained on copyrighted material.

Who develops AI?

You need several skill sets to develop AI solutions. Usually these involve data scientists working together with data engineers and machine learning engineers. There are a few other functions that are also critical for successful AI development, including data governance, data stewardship, and data visualization.

Analytical AI is often developed in-house leveraging internal technical and business domain expertise. With this approach companies can create explainable AI solutions that are easily expandable and cost less in the long term than hiring external consultants.

Since Generative AI solutions require vast amount of data and massive computing power, these solutions are often developed by data scientists & co. at a handful of large tech companies. These tech vendors offer then the Generative AI solutions to other businesses.

When will AI be ready for business use?

<LOUD BUZZER> This is a wrong question. AI is already being used today to increase profit. Analytical AI has been used for 20+ years to drive business decisions. Amazon, for example, implemented machine learning models to provide first product recommendations already in 1998. Generative AI solutions became useable when OpenAI launched ChatGPT in 2022.

And now what?

Although it sometimes may feel like there is competition between Analytical AI and Generative AI, there is none. Each of these AI forms addresses different business needs.

The only competition are investments required to implement AI, including monetary and time resources. But this competition happens at the business level where executives need to prioritize investments based on their business goals. Technology decisions are only the last step during the decision process to determine how these business goals will be met. Whether it’s Analytical AI, Generative AI, or no AI at all, the technology used to address the business goals should not be the worry of executives.

For companies operating in the EU, by the way, GDPR and now the EU AI Act are not good excuses for not using AI. These regulations do not limit companies to use data or AI, unless you want to exploit people’s vulnerabilities, deploy manipulative techniques, or implement social scoring. If none of that applies to your company, these regulations are just guardrails to ensure safe use of data and AI.

Executives don’t need to wait for the latest AI model or for future data centers since that won’t change what AI does for companies already TODAY:

AI helps executives to increase profit.

--------------------

Hi. I’m Jack. With my 27 years of data & AI expertise, I offer KEYNOTES, strategic ADVICE, and EXECUTIVE SPARRING to help companies increase profit through AI. Visit my website to find out more: www.jacklampka.com