People make AI go round.
Artificial intelligence, despite its popularity, has a problem: people don’t trust it. It’s a strange, complex undertaking for many, with terms such as “diffusion models” and “emergent behavior” being bandied about.
Add to the mix hallucinations and biased results — and popular fears about AI taking away jobs — and AI certainly won’t be welcome as it expands across organizations. At the opposite extreme, there tends to be too much of a tendency to rely on AI output too much, which could lead to all kinds of complications.
Somewhere, there’s a middle ground — embracing it for all it can deliver, while exercising caution and keeping close to the business.
“As we think about generative AI we have to think about the role it will play,” said Karen Feldman, vice president of marketing and communications for IBM. Feldman led a panel at the recent Adobe Summit that explored the issues with trust in AI and ways to address it. Going forward, generative AI will shape “the trust with our brand, the trust with our customers, and also the trust we need to create with our own employees,” she said. “The only way we can take advantage of this magical technology is if our employees embrace it themselves.”
Plus, she continued, “we hear questions from our clients about AI: ‘Can we trust the output?’ ‘Will it resonate with our customers?’ ‘Is the AI I'm using secure?’ ‘Is it brand compliant?’ ‘Can I trust that the AI will reward the individuals that create?’
The key to a successful AI undertaking is making everyone in the room comfortable with it. For one of the largest services companies, Aramark, that meant establishing a “safe space” for discussion and experimentation, said Laura Powers, vice president of B2B marketing operations and communications at Aramark, a panel participant. “The trust around ai right now is very negative,” she said. “as in, ‘we’re all going to die from AI.’”
Powers seeks to make AI understandable, sustainable, and open to all employees who have ideas on how it could make their jobs better. Her department, which promotes and pilots AI efforts, acts as an internal “agency” for the rest of Aramark, she said.
Aramark’s AI efforts started with the establishment of an AI task force, she recounted. “It sounds very big brotherish, but it actually was reassuring to have AI as something we knew we were publicly starting to talk about,” she said.
A major focus of the task force’s activity was to develop an overarching policy as part of governance. There are two main points in that policy, she related:
- “Don’t ever enter client or proprietary data into the AI tools.”
- “AI is not a substitute for human judgement.”
With those two guidelines, “we really started off on the journey at Aramark with the ability to feel comfortable in an AI space,” she said.
The company’s AI task force consists of two committees — a regulatory committee, and an exploratory committee. “The regulatory committee, consisting of our compliance officer and legal department, oversaw managing the risk, assessing risk, defining it, determining what the risk level is, and setting up governance,” said Powers. “If an idea moves forward, the regulatory committee looks at the tech behind it. Is it a closed environment? Is it populated with Aramark materials? Is it safe for us?”
The AI exploratory committee, “consists of ambassadors from many departments and cross-functions around Aramark,” Powers said. “We are able to talk in a comfortable and public way with our groups about AI, and what they're interested in, what they're fears are, what they’ve heard, maybe what they're researching, what they’re about to explore.”
AI “ambassadors” are sprinkled across the company to provide assistance and listen to employees as AI projects emerge. The company also actively seeks employee ideas. It provides an online form for submitting ideas for AI projects. “You can either submit that form, or you can talk to the AI ambassador that is closest to you in your department, or any ambassador and roll that up.”
Aramark also has regular discussions with its technology partners — such as IBM, Salesforce, and Microsoft — “to talk about what they're doing in their own companies internally with AI – and that’s been eye-opening,” Powers said.
Another channel for employee ideas is a teams channel called AI Buzz, as well as a related channel called AI Pathfinders to provide updates on projects. “It’s a continuous, cyclical process that we have, and were here to encourage this safe space for AI exploration,” Powers said.
An early AI application the task force is moving forward is proposal generation — “getting to a first draft proposal, from an RFP or RFQ,” said Powers. “Since we've identified that kind of discrete area where we think we can improve our efforts, we have started exploring this proof of concept called intelligent responsive assistance. The business objective is to get a draft quicker. We want to give more time back to agency team members, salespeople and subject matter experts across Aramark on draft proposals.”
The team will apply AI to measure “time spent with proposals, money spent with the agency, and quality of proposals.”
Other projects introducing AI include a legal chatbot and HR automated virtual assistant, Powers said.
Powers provides the following words of advice for getting AI efforts off the ground — but keep them well grounded:
Build AI around a business case. “If you are interested in getting started, or don’t know where to start, or are in the middle of your journey, business case is really important,” said Powers. “We did a lot of work thinking about the value drivers, the business objectives, as what we’re going to measure, what was our business case, and how are we going to make this investment really worth it.”
Start small. For starters, “we focused on one line of business only,” she said. “The line of business partnered with us for a proof of concept. They subscribed and participated with us in al the discovery sessions. We had the demo really quickly.”
Think about skills. Ask some bigger questions: “’what kind of skills do we need to build with our team members?’” Powers recommends. “’What kind of organization do we need to align with?’ ‘Do we need to reorganize?’ I think the answer is clearly yes.”
Establish shared governance. “What is really important about establishing governance is that all departments be involved,” Powers said. “We needed a shared voice to get this started and define how these committees would work and set up the channels. A lot of us felt we had a role in governance as well.”
Stay positive. “Talk internally to your peers about what they're thinking about,” said Powers. “Find those evangelists that can help you and want to talk about AI together. I have found groups that are like-minded, and those people will really help you stay positive and on an upward trajectory.”
