Skip to main content

Play it Smart with AI

Heavily promoted as an effective way to improve business efficiency, generative AI strategies can aid your company 

The Bottom Line: We’ve all heard the claims in the news about how generative AI will change business. What is the best strategy to handle this emerging technology?

AI-generated image of a handshake between human and machine
Editor's note: This is an Adobe AI-generated image created using InDesign with its text to image beta feature. Prompt used: Non-gendered human shakes hands with a machine hand in a helpful but careful way. I added "make it a close-up of the hands" to the prompt when it returned three initial choices with full bodies of people and robots. According to Adobe, their AI is  "designed to be safe for commercial use." Creating the image took less than a minute. Researching Adobe's practices for AI-generated images and doing a Google image search to assess the result's sources, clearly Adobe's stock images, took 30 minutes. 

The past year has been marked by contradictory claims about the benefits and risks of using generative artificial intelligence at work. These claims get daily news coverage that often includes exaggerated headlines. Understandably, business owners are uncertain about handling this new technology.

Let’s first put these headlines in the proper context. Many inflated claims come from company founders who have much to gain if they are right. I consider those claims to be biased. Expert forecasts should also be suspect. Historically, they are seldom accurate about the speed, magnitude or direction of new technologies. Why should this time be any different?

AI challenges

Despite their impressive performances and hyped-up claims, generative AI systems face serious challenges. Let’s focus on two of these challenges that could affect business users.

1. Errors and fabrications

Answers from generative AI systems can contain factual errors and outright fabrications. The fabrications are called “hallucinations” and are estimated to happen 5% to 27% of the time. The problem is that AI can present inaccuracies and hallucinations confidently and articulately.

Consequently, a human reviewer may not always catch them.

Consider the following scenario. An executive asks an AI chatbot to summarize a complex technical report. How confident can the executive be that this report is accurate? Would you feel comfortable relying on it?

Let’s take this example a step further. Consider a support chatbot that interacts with customers. How confident can you be that its information will be presented correctly?

2. Legal issues

Some generative AI systems can create output that potentially infringes on copyrighted materials. Consequently, a company could inadvertently deploy copyrighted materials and face infringement claims.

Consider the following scenario. Your company uses a generative AI system to create marketing copy for your website, images and customer documents. The AI system, trained on data available on the internet, inadvertently includes copyrighted material it found online. Your company could face legal consequences.

AI hallucinations could also create legal problems. Consider a chatbot interacting with a potential customer and providing fabricated information. Do you have to honor the commitments? I found at least one case of a judge ruling in favor of a customer.

Is generative AI useful?

Harvard University and the Boston Consulting Group did a study about the usefulness of AI, which found that AI improved the quality and productivity of most workers. Most of the gains, however, favored employees with the least experience.

On the other hand, the reviews of Microsoft’s AI Copilot tool were lukewarm. The tool often got things wrong and needed substantial editing. This situation will surely improve as the technology develops.

Undoubtedly, the capabilities of generative AI are impressive. But are they valuable to most small and midsize companies? The results appear to be a mixed bag with no convincing case.

A careful strategy

Many companies have publicly deployed generative AI in customer support and other areas. 
An approach should be cautious, evaluating the technology internally and taking deliberate steps to become proficient.

  1. Learning basic prompt engineering. The quality of the prompts you provide an AI system affects the output quality. “Prompt engineering” is not difficult and you can find tutorials on YouTube.
  2. Improving writing skills. AI can be useful when trying to simplify complex paragraphs. We often ask the system to simplify paragraphs and generate several alternatives. These alternatives help us rewrite the paragraphs more clearly and concisely. 
  3. Transcribing online meetings. Generative AI systems can be useful tools for summarizing and transcribing online meetings. This use is valuable if a key member misses a meeting and needs to catch up.
  4. Summarizing documents. We are evaluating AI’s capabilities to summarize internal documents. Since we know the information well, we can catch hallucinations and inaccuracies more effectively.
  5. Avoiding direct client-facing uses. We do not use AI-generated content in any client-facing capacity. This includes client-support chatbots, written materials or images. The legal risks remain unknown, and we prefer to avoid the exposure.
  6. Taking a wait-and-see approach. Perhaps the most important question is whether your company needs to use generative AI to remain competitive. Will your competitors have an insurmountable advantage if they use AI and you don’t? If you run a small to midsize glass company, you can probably afford to wait and see how things play out before jumping in.

Author

Marco Terry

Marco Terry

Marco Terry is managing director of Commercial Capital LLC, a factoring company and provider of invoice financing to companies in the glass industry. He can be reached at 877/300-3258. Opinions expressed are the author's own and do not necessarily reflect the position of the National Glass Association or Glass Magazine.