Principles, foundations and methods play a key role in the successful implementation of artificial intelligence (AI), enabling organizations to get the most value from evolving and ever-progressing AI technology.
Speaking at All Access: AI in PEX, PEX Network’s latest installment of its free webinar series, Lee Bogner, chief enterprise eCommerce architect and principal generative AI engineer at Mars, discussed the confectionery giant’s approach to implementing AI technology. Bogner leads and advises on strategy, governance, policy and delivery. His subject matter expertise includes digital innovation and AI-enabled business models, omni-channel commerce, GPT/LLM-based and ethical AI, technology policy and governance processes.
Mars’ principles and measurement model
Integrating AI at Mars
“Enter AI; we’re all dealing with that more and more today,” said Bogner. How Mars establishes its foundations is a key element of its AI strategy, he added, “integrating AI into our business needs, operations, workflows and cultures” in line with the organization’s principled aspirations. “The trust of the brand is so important. I like to think of the POST method – people, objectives, strategies and (last) technology. If you have the first three lined up, the technology becomes more relevant.” This is Mars’ overall AI vision as it actively evolves from siloed proof of concepts in the pursuit of reusable foundations, methods, resources and skills, Bogner said.
Mars’ AI use cases
Business AI use cases are increasing, particularly in the area of generative AI, Bogner said. Generative AI is playing a significant role in Mars’ corporate and marketing communications – both internal (including geographic translation) and external (local and regional marketing promotions). “It’s helping us with some of the fun stuff,” Bogner said. “Our M&Ms characters have been a staple of movie theatres and television commercials for years. Now, with generative AI, M&Ms are having chats with each consumers.”
Other use case areas include manufacturing automation, product and software development and natural language-enhanced accessibility of technology reference manuals using DIY search and discovery options.
READ: The guide to AI in operational excellence
AI foundations and principles
So far, AI enhancement has been characterized by experimentation over standardization, Bogner said. “That’s an important distinction. We are now moving towards rules around principles and we’re beginning to craft that in terms of AI technology foundations as well as ethics and responsible AI,” Bogner said. A lot of aspects of AI that have been increasing have, to date, been siloed, Bogner added. “We’re in the process of evolving because one of our principles is freedom – we allow our businesses to independently run and drive revenue.”
The key impact of AI foundation is providing clarity versus vagueness, he added. Also highly important is the human-in-the-loop (HITL) concept – the true collaboration of humans with AI and machine intelligence. “Other realities of AI include risks such as accuracy and bias constraints, and HITL becomes important in those aspects too.”
Around 18 months ago, Mars established a corporate AI working group, integrated with enterprise architecture and security governance for checks and balances. “We’ve evolved our responsible AI principles which we collaborated on with our business units, executives and legal/compliance leaders.”
Mars has created 10 principles that cover a range of AI factors such as bias, privacy, accuracy, transparency and do no harm, Bogner said. A key focus is tying these responsible AI principles to Mars’ corporate principles. “They fit very well, and they also fit well with our broader ecosystems, Bogner added. “We’ve been very big in the world of a ‘marketing code’ so our marketing has to go through very specific reviews. For instance, we do not market directly to children and we don’t allow children under 16 to sign up [without parental consent].”
READ: 7 free AI training courses for business leaders
The future of AI
Mars is laying the foundation and methods for the implementation of AI of the future, said Bogner. “AI is going to get better and better, and it seems to be getting better not just every year but every three/six months. It’s very likely that prompt engineering will recede and the AI will get better at understanding what you’re saying like we’re talking here. That’s what the expectation is. In my opinion, we need to focus on human to AI collaboration as it evolves further.” The more prudent way to look at it is really understanding the problem you are trying to solve, Bogner added. If you understand that, you can better communicate and participate with the AI.
More personalized experiences, advanced natural language processing (NLP), multi-modal AI (text, image, video, audio) and decentralized AI will also be on the agenda, Bogner concluded.
[inlinead-1]