AI adoption: What enterprises need to know before they do it

Image created by DALL·E 3.

AI has become so ubiquitous in the enterprise world that organisations fear missing out on its opportunities. However, misconceptions about the technology and its proper use are still prevalent.

So, how then can enterprises swiftly integrate AI into their daily operations? Is AI adoption really as scary and complex as many believe?

Michael Keithley, Chief Information Officer of United Talent Agency (UTA), and Michele Buschman, Chief Information Officer of American Pacific Mortgage (APM), addressed these concerns during the panel titled “The future of digital transformation: Evolving your business with AI” at Zoom’s Work Transformation Summit APAC event.

A tale of two businesses

In the case of UTA, there were two challenges to AI adoption, the first of which was fear of the unknown.

“AI is just a tool. You’ve got to step back and realise that it’s just a tool; demystify it. If you drive a Tesla, or a modern car, you use AI. If you use a smartphone, you use AI. It’s not the Bogeyman. There’s so much press going on, and so much hyperbole around, that one of the biggest challenges is to make people understand what AI is and isn’t,” Keithley remarked.

The second challenge, which was a direct result of the first, was that employees were hesitant to experiment with the technology.

“As I was trying to go department by department to get the leadership buy-in on this, I would pull up ChatGPT on my laptop, and I’d walk over to their office and I said, ‘You know, this is great. Do you use it?’ And I found out that they haven’t really used it,” the CIO recalled.

To put their mind at ease, Keithley allowed employees to try out ChatGPT and become comfortable using it.

“I said ‘Let’s do something. What’s going on in your world? It doesn’t have to be work.’ Then I tried to guide them through a prompt that would be relevant to them, leading to that ‘aha’ moment. And once they got that ‘aha’ moment, they were bought in; they saw the power that’s relevant in their lives, in their world. And then they became big supporters, and I got the green light from that,” he continued.

To make matters worse, the recent strikes in Hollywood were due to the use of AI, among other things. According to Keithley, this affected people’s prejudice about the technology.

Michele Buschman, Chief Information Officer, American Pacific Mortgage. Image courtesy of American Pacific Mortgage.

On the other side of the pond, mortgage firm American Pacific Mortgage couldn’t wait to deploy AI fast enough. However, as a financial institution, it had to study the technology thoroughly or risk heavy sanctions from regulators.

“Everyone loves the functionality, and they gotta have it, which is great. I didn’t have to sell that to the business. But what they didn’t understand is how it works and what those risks would be to our data,” Buschman said.

Before deploying AI, the mortgage firm worked with its vendors in figuring out where the controls are, and what parameters will be implemented to ensure data protection.

“I would rather take 30 days, make sure we have a policy framework, that we understand what the risks are, and we really understand the controls we have over our data, rather than me just saying, ‘Okay, go for it, turn it on.’ And tomorrow, we’re on the front page of the paper. So for me, the challenge was getting the business to understand the risks, and why I needed to slow play a little bit,” she added.

Turning on the jets

To protect its data, UTA developed a private GPT based on Microsoft Azure’s version of OpenAI’s ChatGPT. 

“That way, we can have our own data in there and not worry about PII, or confidential data, and some of those other things. So, we made the shift from the public version of GPT to a private one, and that’s where things have really gotten exciting for us, as we start adding our own data, augmenting what you can get with the public one, but feeling secure and comfortable with it. That’s where the magic happens,” Keithley said.

Moreover, the company started an incubator programme, where Keithley and his team visited each department, gamifying the experience for everyone.

“The department gets to upvote or downvote the value of the prompts that people are submitting. Quickly, over the course of a week, with some prizes, people get excited about participating,” he continued.

As a result of the programme, UTA now has four or five prompts that drive significant value for the business, the Chief Information Officer added.

Meanwhile, APM decided to use AI to generate quicker answers to both internal and external queries.

Michael Keithley, Chief Information Officer, United Talent Agency. Image courtesy of United Talent Agency.

“Loan officers often send questions about scenarios, like a buyer’s credit history. Our chatbot can handle some answers, but many require deeper analysis. We have what’s called a scenarios team for this. Traditionally, an individual from this team would find the information and answer the question,” Buschman said.

To improve this process, the company expanded the knowledge base of their chatbot and connected their software with the credit underwriting guidelines of agencies such as Fannie Mae and Freddie Mac.

“Now, our service person, when confronted with a question that couldn’t be answered, can go in, and be more specific in the prompt or question they’re asking. The system will go to the website and bring back the information with a deep link to where that information can actually be seen. What might have taken someone 20 or 30 minutes is now reduced to five minutes. Even better, the deep link is included. So, if they need more information or want to verify it, they have direct access to the source of that data for validation,” she elaborated.

AI assist

Apart from their internally developed AI projects, both UTA and APM are leveraging Zoom’s recently launched AI companion, which has proved especially helpful for summarising meetings.

“The fact that the host sets up the meeting, they can decide whether they want to distribute the summaries to others, edit them, or whatever. It’s been really seamless and has  generated a lot of value,” Keithley said.

For APM, using Zoom’s AI companion has reduced the need for manual note-taking, which allowed people to focus on the meeting itself.

“You don’t lose track of action items, as it breaks them out for you. It provides both an executive summary and a detailed summary. The amount of productivity enhancement we’re going to see organisationally, as we continue to push deeper into all the different feature sets available to us, is amazing,” Buschman shared.

Additionally, the APM CIO discussed the company’s future vision for their Zoom usage.

“We’re planning some custom development to add our intranet as a tab within the application, so that this can be the communications hub. Adding the AI companion layer will enhance this further, with capabilities for reading your emails, listing action items from those emails, looking at your documentation, and analysing recordings of both meetings and phone calls,” she said.

To conclude, Buschman advised enterprises to update their vendor due diligence questions to better understand how their data is being managed. She also recommended starting small, acknowledging that not all organisations can afford to take on ambitious projects simultaneously.

“See what you can leverage within your software provider and start experimenting. If your organisation isn’t initially excited about it, work with a few individuals and show them what it can do. It’ll spread like wildfire across your organisation,” she suggested.