Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Fast and furious or cautious? AI requires both.
- Leaders from PwC and NBCUniversal share their AI journeys.
- Be ready with the right data before introducing AI.
Getting started and capturing the essence and value of AI can be exhausting. It also poses a vexing question: Is the way forward to move fast and furious, jumping on every model that comes online, or is it to proceed with a plan and extreme caution? Testimonials from two enterprise IT leaders suggest that a mix of both approaches is necessary.
Also: 5 ways to grow your business with AI – without leaving employees behind
Divergent experiences in launching and maintaining AI momentum were shared at a recent conference hosted by Section, a consultancy led by storied NYU professor Scott Galloway.
1. The human is the loop
For starters, don’t just hand over the keys to AI agents — any endeavor needs to remain human-instigated and human-led. “Stop being the human in the loop. The human is the loop,” said Scott Likens, global chief AI engineer at PwC.
Also: Building an agentic AI strategy that pays off – without risking business failure
Getting started means starting with the end user and working backward to determine the right tool for the job.
“Start with easily repeatable processes and data. Start with the pain point,” said Lasherelle Morgan, senior vice president of AI innovation and acceleration for NBCUniversal. “Don’t just bring in an AI tool. Ask, ‘what are you struggling with?’ ‘What are you spending five hours of your day on?'”
2. Experimentation is important
Be willing to experiment on a wide scale to determine where AI can deliver, Likens said.
PwC, for example, rolls out AI-driven experiments in one-day or five-day cycles. The challenge is to urge business leaders to think about the possibilities AI brings versus simply achieving 2% or 3% in cost savings. “All this talk of tokens just started a couple of months ago, and now all of a sudden there is a cost focus with AI,” Likens said.
“That’s the wrong way to look at it. Experimentation is really important and so easy nowadays, and you get feedback fast,” Likens said.
This calls for a radical shift in thinking among many business leaders, especially at the midtier level, Likens said.
“Many are not used to one- or two-week cycles; it’s a different architectural mindset,” Likens said. “Top executives and board members may be on board, as are new employees that have already been there. It’s that frozen middle, those experts and managers who don’t want to change their ways. It’s a human challenge.”
3. Blow up a bad process
For those excited about AI, there’s a perception that the technology will quickly fix gummed-up processes and speed up business decisions. There’s an important caveat to such thinking, Morgan said.
“You have to have clean data, and a workflow that is clean from start to finish. You need to literally get a pen and paper and write out the process, and show me who is the owner of that process. One thing AI is really good at is blowing up a bad process,” said Morgan.
With that pen and paper, find out from users: “what do you have to do repeatedly that you hate doing?” Morgan said. “Then work from there. Meet people from where they are. Also, where there is a lot of data that is easily repeatable as well. Start with things easily fixable by AI and start there.”
Also: How to build better AI agents for your business – without creating trust issues
At PwC, the pace of AI innovation and development has been furious, but underneath it is a well-planned data foundation formulated before AI was a big part of the picture. The company addressed data issues in regulated pieces of the business, such as accounting and auditing, Likens said.
Likens said the challenge was providing context to the data, which “usually sits in peoples’ heads. How do you extract tacit knowledge?”
Likens said the goal of PwC’s AI efforts is “tacit knowledge collection, telemetry, what agents are doing which then feeds into tacit knowledge collection.” That calls for a “focus on the architecture first, so it can scale for our people, so when they use it, they know it’s safe, they know what they can use it for, they have access to the right data they are allowed to have access to.”
4. Governance and guardrails
At NBCUniversal, governance and guardrails are essential pieces of the process. Still, the amount of oversight depends on the amount of risk to the organization, which Morgan refers to as the potential “blast radius.”
“With a use case as simple as an agentic tool that presets my lunch on my calendar or something like that, that’s low risk, we don’t need a human in the loop. If we have something agentic that is automatically sending our messages to consumers, that is a bigger deal because it is a higher risk for the company,” said Morgan.
NBCUniversal’s governance process consists of intake forms. These help Morgan’s team track and measure the potential impact of AI and agents on the organization.
Also: 96% of IT pros use AI now: Their top 7 agentic applications and biggest implementation roadblocks
At PwC, AI responsibility is centralized around “one percent of the organization — those are the deep AI engineers,” Likens said. “These engineers set standards, create the chassis that is trusted and safe. and do the builds. Then we have 10% who are hands-on builders distributed across the business. building for clients, who directly understand an industry or function.”
