6 frequently asked questions about artificial intelligence CEOs need to answer
In previous reports on artificial intelligence, collected data showed that the use of artificial intelligence (AI) among SMBs focused on business operations. In our most recent Vistage CEO Confidence Index report, analysis revealed that the primary application of AI has shifted to customer engagement.
When asked about their approach to Open and Regenerative AI platforms such as ChatGPT, 43% of CEOs reported they are either currently using or testing it, while 33% said they are not officially using or testing it.
To help CEOs better understand the various applications, strategies and how to mitigate the risks of AI for small and midsize businesses, we invited 3 Vistage speakers who have expertise on the topic — Ross Hartmann, David Karandish and Amy B. Goldsmith — to join us for an expert roundtable on AI (available on demand).
This session provided basics on AI literacy to help CEOs understand the benefits and risks of leveraging this emerging technology.
Key takeaways from experts in this session:
David Karandish is the CEO & co-founder of Capacity, a SaaS company headquartered in St. Louis, Missouri. Capacity is a support automation platform that uses AI to deflect emails, tickets, and phone calls. They work with 1,900+ customers to automate customer and team support across web, SMS, voice, email, social, Slack and MS Teams.
- Be able to differentiate between the types of AI. Automation is software that runs in an uninterrupted fashion. Non-generative AI is software that learns, while generative AI is software that creates new content from existing data.
- Understand AI’s relationship with labor. Think of automation as a replacement for manual labor versus AI, which is more software that learns. I think of that as a replacement for cognitive labor. Ultimately, generative AI is software that creates new content based on existing data that becomes a replacement for creative labor.
- Find the best application for AI. Clients do their best when using generative AI to write new content instead of using it to find new information.
- Double-check AI’s work. Be sure to have a system in place that flags when AI does not know the right answer and bring in someone who can correct it.
Ross Hartmann is the founder of Kiingo AI, which offers AI consulting services and custom AI software solutions. He is an expert at integrating the latest AI technology in businesses and has worked with businesses from $2 million to $500 million to design and implement an AI roadmap and strategy.
- Get comfortable using AI. 85% of U.S. workers have used AI tools for tasks at work, while 69% are afraid to tell their managers about using AI — they’re afraid they’ll be replaced.
- Use AI as a co-pilot before using it as an autonomous agent. Using AI throughout the workday can free up employees to work on more creative tasks and think more strategically, thereby increasing retention and decreasing turnover.
- Companies today are using AI for more than just content creation. Other use cases include data insights, revenue opportunities, sales enablement and workflow automation.
- Use AI to access critical data. The ultimate application of AI is having all of your company’s data accessible in one place through an AI brain, creating one interface that can give you true transparency into how your company works today and how to make it more efficient going forward.
- AI isn’t going to replace you, but a person using AI will. While implementing AI can give us a temporary competitive advantage today, the use of AI will soon become a necessity. Ask yourself, “What can I do to capture this competitive advantage and move one step closer to AI today?”
Amy B. Goldsmith is a partner of Tarter Krinsky & Drogin LLP. As co-chair of the firm’s Cybersecurity, Data Management & Privacy Practice and a partner in its Intellectual Property Group, she ensures that her clients meet and comply with all requirements set forth under global and local privacy and data security laws.
- Be careful what you tell AI. Do not put confidential or personal information in ChatGPT. ChatGPT and many of the other platforms (Midjourney, DALL-E) do not agree to keep the information confidential. Plus, your company may have received this information from others who rely on you to keep the information confidential.
- Fact-check every output — if you can. Currently, there is no failsafe way to confirm that the information used by any AI platform isn’t owned by someone else. This leaves the user of the output at risk for copyright infringement. Some of the platforms are being sued for using others’ copyrighted works without a license to train their AI to mimic the artist’s styles.
- Remember: ChatGPT’s database is not verified. It is everything that was on the internet as of 2021. The terms and conditions of ChatGPT say that they may hallucinate, which means they’re making up information when they can’t determine the correct answer or when there isn’t enough information in the dataset to give a good answer. Therefore you must be sure to fact-check against a reliable source — if you can.
Join us on September 29th at 10 a.m. PT for “Artificial Intelligence Roundtable: Best Practices from Small Businesses,” where a trio of Vistage members who adopted AI early share their applications, results and lessons learned. Register now.
Ross, David and Amy took some time to answer the following questions that came from our audience during this webinar:
1. If you are integrating CRM data into AI tools, are you giving away your customer data?
Ross: Existing CRMs deploy their own version of these AI models, so it depends if you want to enhance the functionality they have. Most CRMs expose APIs that allow you to securely connect them to AI through authentication so it comes down to having to trust your data vendor. You can either use an API to extend your CRM’s functionality through AI or use the functionality that it already has naturally.
2. Are all of the questions we feed into ChatGPT being used to train AI to do it better in the future (when using the free version of ChatGPT)?
Ross: Essentially you agree to allow OpenAI to train AI on your data if you are using the free version of ChatGPT. Even if you are using the paid version of ChatGPT, training is turned on by default. Make sure you tell employees to keep it off if you don’t want your information to be used in future models. The alternative is to use a service that calls OpenAI through the API. Any calls through the OpenAI API are not trained. If you are using a third-party AI service, such as one for meeting notes, presentation creation, etc., your data is going onto their servers, so you have to think about what are they doing with your data.
3. What is the balance between generating content and checking it/making it your own?
Amy: Use the Plagiarism checker via Grammarly to know the source of text and check if there is true plagiarism or just citations. What you may not be able to determine is if the AI tool is using your competitor’s material.
Therefore, there still must be some level of curation, and what you use depends on the level of risk acceptable to your company.
Using AI-generated images carries a very high level of risk since the source of the images may not be known, and given the Copyright Office’s position that only human-created works are copyrightable, your company may not own the resulting image — anyone can copy it.
4. What kind of AI guidelines are you seeing companies put in place?
Amy: You should not input any confidential information, so be sure that employees know to ask whether the information is confidential or not. Questions to ask when creating AI guidelines include:
- How are you using AI?
- What AI applications are you considering allowing employees to use?
- Who is responsible when you have received the output? Who will be looking at it? Where does that responsibility lie?
Post your AI guidelines on the company intranet. The guidelines are not static; AI is continuing to evolve, and this needs to be a continuous conversation. The bottom line is you should have guidelines, know what you are doing, and train your employees.
Ross: Get the paid version of ChatGPT and turn off training. With the free version, training is always on so your data is potentially being used to train the model. Make sure everyone has training turned off and that they don’t accidentally turn it on at some point. My top suggestion is to deploy the AI onto your company’s private network — that removes the security risk. You can also deploy an open-source model on your intranet so you have complete control over the model.
5. What can companies do to mitigate risks from AI?
Amy: Training, training, training! Budget allocation for continued training is key. Check everything you get, and if there is doubt, pick up the phone and talk to your manager before clicking on a link, answering a question from someone you don’t know, or sending money to a new account. Slow down, think and ask, especially when getting an email request you were not expecting. Put on your buyer-beware hat!
Ross: Create a passphrase when things seem fishy.
6. How can a company ensure they will get recognized by AI?
David: Have unique content. AI is an excellent tool to use for a first draft, but a poor final draft in terms of uniqueness and something that will add value to search engines. Write most of your content, don’t let AI write a good chunk of it. However, using AI to create a draft marketing blog post is a great way to get started.
Ross: Just like when you’re optimizing for Google, you want to be popular, you want to have engagement, and you want to have valuable content. The way AI companies work is that they take a snapshot of the internet and feed that data into an AI model to find patterns. The more often you appear in the training data (i.e. the Internet), the more likely you will come up, so you want to be everywhere!
Amy: AI is not replacing people, we need people to curate, which allows people to do the deep thinking companies need to stay competitive.
These interviews have been condensed and edited for clarity.