2024 is just around the corner and Generative AI (Gen AI) is projected to stay and expand its use in various industries all over the globe. As different sectors continue to navigate a world where GenAI is the norm, there are still some limitations and cautions regarding the technology.
Jason Gartner, GM & CTO of Technical Sales & Client Engineering at IBM sheds light on the limitations of GenAI, managing its cost, the necessary skills needed to facilitate its adoption, and what technology leaders need to do to adopt GenAI solutions across their organisations.
Editor’s choice: Podchats for FutureCIO: Generative AI, from hype to intent
Limitations of Generative AI
Though Gartner acknowledges the remarkable advantages of GenAI, he says the main limitation revolves around trust and transparency.
“Can you trust the results that you get from generative AI? And what does that look like? When you begin to automate, do you have the IP rights? Do you have the copyright? Are the responses that you are giving to a customer free of hate speech, racism, and all kinds of other types of interactions associated with it?”
Gartner says that CIOs need to grapple with the issue of transparency and should ask more questions when using GenAI: "What level of transparency do they get? What visibility can they get into the response that they get from it? Do they know where the data comes from? Do they know where the answers come from? Do they know what source the answers came from? What is the data’s source?"
He says this is a big challenge for CIOs. “The last thing they want to do is build dependency on a model or a system but not knowing the copyright, the IP, or any of the other types of response they get from these large language models.”
Gartner adds what differentiates IBM’s solutions is providing their clients with a data pile. “We have published the IBM data pile, and it is something that we own, and that we have full copyright and IP rights of. That is different than some of the other large language models out there. We were the first ones in the market to identify our large language model.”
“There are very few companies willing to not just stand behind, but also put money in the indemnification of the responses that come out of their generative AI models,” he adds.
Managing the cost
Deploying GenAI in business operations becomes costly to some organisations. Gartner says that the return on investment for GenAI solutions is how they accelerate productivity.
Gartner acknowledges It is a challenge to push GenAI technologies to finance teams especially when they are tight on budget.
Implementing something cool, but neither saves money nor makes money does not interest our clients. They are not interested in the bright and shiny. They are interested in productivity. It either generates more money, save money, or both.
Jason Gartner
To him, the returns always prove GenAI solutions to be a good investment. “We are not talking about a 10% return. We are talking about a 1000% return depending upon the use cases that we have in the spaces.”
They have one client, for example, that receives millions of emails a day with only 40 people working full time with a response time of about 55 hours.
“With a generative AI solution which we built for them, we have taken that response time down to two people, 90% automated, and response time in four hours.”
He says that they only spent less than a day training the model for it to be able to respond to emails, providing significant cost-savings and driving customer satisfaction.
“That drives greater customer loyalty, which in the end affects the revenue streams associated with this client,” he quips.
Necessary skills for AI technology
Among all the skills needed to integrate with GenAI, Gartner’s top-of-the-mind is “your imagination. It is.”
Running a watsonX challenge where they gather over 170,000 IBM employees answer a challenge using their GenAI platform, he realised how important imagination is in navigating AI technology.
“It came down to curiosity. It came down to the ability to dig themselves into it and enable them to get to a solution using the technology.”
He concludes that the overall skill barrier is not that high as “someone who can write a good Google query can write a good prompt to AI.”
“There are more skills involved with it: a little bit of instruction, a few examples associated with it, and then a very good prompt can elicit an accurate response. That type of training is something that can be done fairly quickly and fairly easily,” he says.
He says skills in data science, governance, risk, and compliance may come in as handy AI skills. However, it ultimately boils down to imagination.
“Ultimately all of those are limited by the imagination. That is something that we have been working with our clients and our teams.”
Adopting AI as a business solution
When adopting the GenAI solution, Gartner suggests that organisations focus first on the use cases that return the most significant value for the business, and the market, as well as address organisational problems.
“There are so many ideas and bright and shiny objects for us to chase out there that we should be grounded in business value.”
He also suggests business leaders consider transparency before adopting GenAI solutions. “Can you get the level of trust and transparency out of the technologies that you're looking at that you need to be able to sustain you in the future if you're building a code assist or you have some sort of code generator? Are you confident that the code it generates is yours?”
He says that trust and transparency are some of the things technology leaders need to think about. At the end of the day, Gartner says, it is not about the technology but about what organisations can do with it.
“What we're going to be seeing here is the next new business model coming out of GenAI and that there are many chapters to be written in this story.”