Why AI Implementation Works Best With A Team Approach

This is the published version of Forbes’ CIO newsletter, which offers the latest news for chief innovation officers and other technology-focused leaders. Click here to get it delivered to your inbox every Thursday.

---

Artificial intelligence can bring about the next technological revolution, but it’s only as good as the data that goes into the system. And while companies are developing systems that can create realistic writing, music, pictures and videos using simple text prompts, many writers and artists are not happy about where that data comes from.

Musicians have joined in the fray, with the Artist Rights Alliance sending a critical letter to AI developers this week asking them to stop training AI on existing music. The letter, signed by more than 100 artists, including Billie Eilish, Nicki Minaj, Elvis Costello, Jon Bon Jovi and the estates of Frank Sinatra and Bob Marley, accuses AI developers, tech companies and digital music platforms of working to “infringe upon and devalue the rights of human artists.” AI, the letter says, can create music that lessens the value of the songs created and performed by people.

Other types of creators have already taken action against AI developers for training models on their content without permission. The New York Times sued OpenAI and Microsoft at the end of 2023, claiming the companies used millions of copyrighted articles to train generative AI models. The Authors Guild sued OpenAI last September, accusing the company of using copyrighted works for AI model training without compensation or authorization. Thousands of nonfiction authors also filed lawsuits against the company in November. It’s not clear at this point where any of this will go, but the creators’ point of view is being reinforced through this latest letter. Those who create content for a living want to be compensated for any uses— especially for a computer system trained to mimic exactly what those creators do for a living.

It also drives home a difficult point: It’s hard to get quality, non-copyrighted data to train an AI system. Adobe has tackled this issue in its AI-powered image generator Firefly by training its system only on Adobe Stock images, openly licensed content and content in the public domain. It also has a compensation model for Adobe Stock contributors whose content is used for image training. While Adobe already had access to a lot of this content—which is high value and proprietary—there could be elements of this arrangement other AI developers could bring to the table. They could partner with publishers, record labels and licensing firms from the get-go to ensure that all content creators are fairly credited and compensated, working more closely with them than “borrowing” content without permission.

AI does much more than create images, writing and video, and its purposes can be targeted to specific company needs. CIOs play key roles in this conversation, but there are others who should also be a part of the process. I talked to ServiceNow Chief Digital Information Officer Chris Bedi about how to successfully bring AI to a business. Excerpts from our interview appear later in this newsletter.

POLICY + REGULATIONS

Big tech companies are scrambling to change their existing software packages to comply with EU antitrust laws and regulations. Microsoft announced this week it will remove its Teams video chat application from its Office 365 package for all customers worldwide. Teams has been part of the general Microsoft package since 2017, and rivals have complained this gave Microsoft an unfair advantage. The company began selling the two products separately in Europe last year, and decided to globalize the change this week to ensure customer clarity and make it easier for global companies to make large software purchases.

Some of these changes to comply with European law may go further than companies had previously considered. Forbes senior contributor David Phelan reports on comments made by Margrethe Vestager, the EU’s commissioner for competition policy about how Apple hasn’t made changes to its iPhones that were required by law, including making its Photos app removable. Phelan notes that Apple’s Photos app is currently integrated throughout the phone’s operating system, with different apps connecting to Photos in different ways. Unlike unbundling an application from a purchased package—like Microsoft is doing, or allowing messages from other platforms—like Meta’s WhatsApp has done, allowing additional developers to make something to replicate the Photos app and integrate it into Apple’s operating system could represent a significant reprogramming and recalibration.

ARTIFICIAL INTELLIGENCE

Many companies are announcing AI integrations that can make their software and platforms more powerful, user-friendly and customizable. Salesforce is rolling out a beta version of Einstein Copilot for Tableau, which the company said can “make everyone a data expert,” Forbes senior contributor John Koetsier writes. This AI assistant, which works in the data management and visualization platform, can help users probe deeper into information by suggesting relevant questions for the data, automatically extracting important pieces of data, and remembering previous questions to help users contextualize their work. Koetsier writes this new AI assistant is basically a natural-language interface that can help anyone get specific answers from large datasets, even if they aren’t experts in writing queries or data analysis.

Microsoft is working on an AI-powered chatbot for Xbox. Forbes senior contributor Paul Tassi writes this will be an animated character—potentially similar to the company’s earlier Windows assistants Cortana and Clippy—that can help players with audio or typed requests for support tasks, including technical issues or refund requests. Microsoft is already adding its AI-powered Copilot assistant to its Windows operating system for computers, and it released its first computers with dedicated Copilot keys last month.

But AI integrations aren’t all about chatbots. Artificial intelligence and machine learning can also make existing systems more efficient, and they may be added to a system in a way that isn’t obvious to the user. Forbes senior contributor Adrian Bridgwater writes about “infused” AI applications, including Nerdio’s upgraded AI-powered optimization. This functionality lets generative AI “learn” how to automatically and proactively reallocate cloud computing power to different users and functions as needed.

NOTABLE NEWS

OpenAI has revealed another new generative AI tool, but like its text-to-video generator Sora, it’s keeping this one from the public for now. OpenAI says its Voice Engine can use text and audio samples to create “natural-sounding speech that closely resembles the original speaker.” However, the high potential for misuse of this technology—especially in an election year—has kept it inside for now. OpenAI says it is engaging with U.S. and international partners across government, media, entertainment and education to gain their feedback. However, OpenAI wrote in a blog post that the company seeks to get explicit and informed consent from speakers whose voices will be used by the tool, and it has a set of safety measures to brand all Voice Engine-created content as AI-generated.

BITS + BYTES

ServiceNow CDIO Chris Bedi On How To Maximize AI Planning

Generative AI can maximize efficiency and boost productivity at many parts of a business. ServiceNow has been integrating AI into many of its applications, which are used for a wide variety of business functions. Chris Bedi, ServiceNow’s chief digital information officer, talked to me about who should be involved in making decisions to bring AI to a company. This conversation has been edited for clarity, length and continuity.

When a company is trying to figure out how to use AI, who should be part of the conversation and how should they start planning?

Bedi: It’s a cross-functional team, for sure. For me, it starts with what outcomes are we trying to drive? It could be a speed outcome: we want to get something done faster. It could be an experience outcome. It could be a productivity outcome. …The ‘squad,’ so to speak, has to be the CIO organization because all the tech is underpinned by the CIO organization. Cyber[security] has to play a role here to making sure the data is not prone to the different attack vectors that have surfaced for gen AI. Legal. …Me and the general counsel in partnership [have created an] AI Ethics Council. All of these functions need to come together to really move forward.

Companies need to be operating at two speeds with AI. One speed is, ‘Let’s incrementally improve outcomes we’re already striving for.’ It’s fairly easy to get organized around that because there are already teams working on it. AI becomes another lever to accelerate it. Then, I’ll call it speed two, which is taking a step back [with a] clean sheet of paper and saying, ‘OK, if we were to draw this up from scratch with an AI-first approach, what would this look like?’

I’ve often seen organizations trying to conflate the two and then not much comes out because they’re trying to solve for both at once. I do think these are different speeds—speed two, which is the complete reimagination, requires a lot more thinking and a lot more change management. It’s a lot more sometimes threatening to people, as well, because it’s fundamentally reshaping their jobs. But if we have that three-year north star defined, even if we don’t know how to get there, then speed one and all these incremental improvements can start to accrue to that vision.

How does a cross-functional team impact implementation of AI?

What’s interesting about AI is the technical implementation is actually pretty quick. Then it becomes change management, training and measuring outcomes. Did we get the experience right or did we need to tweak the UI in a certain way? In the end, [it’s] measuring how quickly are we getting to those business outcomes. They may be measured in months, not quarters and years, like some of other long-running IT projects, because results are so apparent very quickly.

Given how nascent we are in the AI journey, there’s certain things—I call it healthy experimentation without permission—that I think CIOs should actually be dealing with. Really exploring the possibility [of] how much can we stretch this? How much can we redefine the role of a content writer and marketing, for a recruiter in HR, or a salesperson servicing their customer. What could we do in an experiential way?

What is the difference between implementing more organization-wide AI changes and smaller experiments that may lead to future implementations?

The cross-functional team is critical to get to the finish line, which is the business result we’re looking for. [But companies need] room for experimentation, looking at new technologies, because it’s moving incredibly fast. CIOs need to have a team constantly experimenting with it.

I think about the speed of innovation with generative AI, with things like text-to-code, text to automated workflows. We’ve seen in the last three to four months an overall 5% productivity gain for all software engineers, which isn’t small. We have a 52% acceptance rate on the code that generative AI is putting out there, so it’s really helping speed innovation, too. I think CIOs need to look to AI to say, ‘How do I speed up innovation in my organization using generative AI tools?’ But also, ‘How do I lower the technical barrier to being able to participate in innovation?’

FACTS + COMMENTS

Free blue checkmarks began appearing on many accounts on X, the social platform formerly known as Twitter. Before the platform was purchased by Elon Musk, a checkmark indicated it had verified the owner of the account. Under Musk, these checkmarks were taken away and available for purchase.

2,500: Followers an account needs to have to get an automatic blue checkmark now, according to a post from Musk

$84: Annual price for a Premium X account, which includes a blue checkmark

‘Please don’t judge for my complimentary blue check’: Actor Mark Hamill posted Wednesday

STRATEGIES + ADVICE

Developers need time to write code, and there are some ways to help them get uninterrupted opportunities to do that.

Among all of its functions, AI can be used to help make your work teams stronger. Here are some ways to do it.

QUIZ

Which app earned the most revenue last year?

A. YouTube

B. Disney+

C. TikTok

D. LinkedIn

See if you got the answer right here.