Three tips for adopting generative AI in business in a cost-effective and safe manner
Generative AI and its transformative impact
01/27/25
November 30, 2022 will forever remain a day to remember in human history. It was on this date that the ChatGPT generative AI system was publicly launched, triggering a phenomenal media and popular buzz. In the early days, futurists who didn't really understand how this technology worked promised us the moon. AI would automate a multitude of mind-numbing tasks while putting millions of people out of work.
Then we realized that AI was neither a logical nor a mathematical tool but rather a statistical technology and that it could hallucinate if we talked with it for too long. Having regained our senses, we finally realized that generative AI was a powerful tool, but with limitations, like so many others. In particular, generative AI can annotate, segment, correlate, and synthesize information faster than a human.
We are currently at the stage of enterprise adoption. Knowing the strengths and weaknesses of generative AI, what are the most important points that a business leader should consider when establishing an AI implementation strategy? I propose three:
1. Making AI profitable
Evaluating cost vs benefits for AI adoption
Many of the enterprise digital solutions we use every day offer optional modules integrating generative AI. However, these options are often expensive. And as the companies that develop the AI engines at the heart of these modules are not yet generating a profit, it's quite possible that the prices of these modules will increase radically in the future, once the use of the technology is well-established with customers.
Launching pilot projects to identify needs
It is, therefore, important to clearly identify needs and ensure that the technology is actually used so that AI is profitable from day one. In our company, we have chosen to launch pilot projects.
Focusing on high-value use cases
We are helping each department to analyze their needs and identify applications where the use of AI would be profitable, and we lend them a license to try it out. If the trial is conclusive, we ask them to draw up a list of the people who will really need AI, so as to equip only those who will really exploit the technology.
2. Assisting humans with generative AI
Understanding AI’s statistical nature
Although AI stands for Artificial Intelligence, there's really no intelligence in AI. Generative AI is, in fact, quite simply the most gigantic statistical engine ever devised by man. And statistics mean probability and uncertainty. Therefore, an answer provided by an AI system is never certain and is generally provided with a percentage of certainty.
Leveraging AI for time-saving tasks
On the other hand, I see AI as an extraordinary technology for assisting humans. AI can save employees a lot of time, for example by identifying elements in an image, transcribing recorded meetings, summarizing texts, or suggesting answers to an e-mail. This use of AI will succeed as long as a human checks and corrects the work done by the AI. So, in my opinion, the implementation of AI in business requires clear rules of use to ensure that there will always be a human in the loop.
3. Protecting data confidentiality
Using private AI solutions with secure sandboxes
Before using generative AI technology, it's crucial to understand how the data supplied to the AI system will be managed, to ensure that confidential company data remains confidential. Some companies offer private AI solutions into which you can inject corporate data without fear, as it will be isolated in a sandbox that only your company can access.
Educating teams on the risks of public generative AI and developing clear company-wide data usage policies
In other cases, for example using generic generative AI on the Internet, it is highly likely that the AI will digest your data.
There will then be a non-zero probability of your company's confidential information ending up with your competitor. To protect the confidentiality of your information, it is therefore important to use only AI tools that offer a private sandbox and to communicate clearly to the whole company that no generic AI should be used for business purposes if company data is to be provided.
Strategic steps to AI success
Today, we're surrounded by people who are very enthusiastic about AI technology and bombard us with requests to use it professionally. It's important to keep a cool head and make reasoned decisions to seek gains in efficiency while protecting our business from the abusive utilization of AI.
By making sure AI is cost-effective, integrating it as a tool to assist humans, and protecting data confidentiality, you'll have the keys in hand to ensure a successful integration.
ABOUT THE AUTHOR
Marc Soucy, PhD
President and co-founder of InnovMetric
After having invented several innovative point-cloud-processing technologies during his doctoral studies, Marc Soucy cofounded InnovMetric in 1994 to develop and commercialize the PolyWorks® point-cloud-processing software. As the president and global product development owner, he has led InnovMetric to become the leading provider of universal 3D metrology software solutions used by major automotive and aerospace OEMs worldwide, as well as their Tier 1 suppliers, for dimensional inspections related to product engineering, quality control, and manufacturing process optimization. Today, he focuses on expanding InnovMetric’s intelligent metrology solutions to empower manufacturing organizations to deploy digital collaborative processes that improve their 3D measurement efficiency by an order of magnitude.