(From left to right) McKinsey’s Rodney Zemmel, PwC’s Dan Priest, Deloitte’s Jim Rowan, and EY’s Matt Barrington are at the forefront of AI strategy for clients.Matt Barrington/EY, Rodney Zemmel/McKinsey, Dan Priest/PwC, Jim Rowan/Deloitte, Elizabeth Fernandez/Getty, Tyler Le/BI
Consulting firms have become a destination for some companies looking to make the most out of AI.
AI leaders at these firms use tools like GPT Enterprise and internal chatbots like McKinsey’s Lilli.
BI asked AI leaders at several large companies to share tips for using the technology.
Working with artificial intelligence can sometimes feel more like an art than a science.
That’s why many companies are turning to consulting firms for guidance on how to maximize the technology.
Top firms are not only helping companies develop AI tools, upskill their workforces, and identify potential security weaknesses, but they are also creating chatbots and agents to organize their firm’s knowledge and streamline routine tasks. As a result, AI leaders at consulting firms tend to have a handle on AI strategies that can work for a broad range of tasks.
Business Insider asked AI execs at five top consulting firms — Deloitte, EY, KPMG, McKinsey, and PwC — to share their best tips for using AI in everyday work.
The AI leaders said they regularly used various AI tools, including models from OpenAI, Google, Microsoft, and Anthropic, as well as tools built internally, like McKinsey’s Lilli, EY’s EYQ, and ChatPwC, PwC’s internal version of ChatGPT.
Here’s how they use AI and some of their advice for getting the most out of it. Responses are edited for brevity.
Dan Priest, US chief AI officer at PwC: I do a lot of research with it. For instance, I was doing some analysis on labor productivity and how AI will improve labor productivity. The typical search would produce labor statistics. Well, AI, the big powerful foundation models, it’ll grab those labor facts and statistics, it’ll do analysis, it’ll show you trends, discontinuities, or cause analyses. It is much more robust. In terms of research and analysis, it emerges as a thought partner versus just a search engine.
You discover blind spots in your thinking. I was writing a policy and I thought it was pretty comprehensive, and I ran it through GPT Enterprise, and it found two other points in the policy that we should be adding.
Todd Lohr, head of ecosystems at KPMG: Part of my job as a leader is being able to synthesize information. AI is very helpful in that it has allowed me to understand trends and the marketplace and has enabled me to have a broader view as a leader and synthesize and ingest a lot more information.
It has also been helpful for communications in terms of preparing for meetings, follow-up from meetings, as well as correspondence.
Rodney Zemmel, global leader for McKinsey Digital and firmwide AI transformation: I’ve found it to be excellent at “level one” creativity and coming up with things you generally will not have thought of. It’s an excellent aid to brainstorming for our teams. I haven’t yet seen it as having true unbounded creativity, i.e., a new way of looking at the world. That won’t be far behind, though.
Dan Priest, PwC: I’ll give some context about what I’m trying to do, a short, punchy question, and then ask follow-ups that make them increasingly specific, and then you can adjust based on what you’re seeing.
During the week, I travel a lot, and if I get 100 or 200 emails in a day, it’s just really hard to keep up with every single one of them. I go into Microsoft Teams, activate Copilot, and ask it to review all messages in Teams and email and find the actions for me. I’ll just spend 15, 20 minutes at the end of the day, do the prompt “Identify emails that are addressed to me directly or that have an action for me,” and it produces the list. It’s not perfect, but it’s good at it.
I like to cook, and I don’t like to waste food in the refrigerator. So I will prompt, “Create a recipe with these ingredients,” and I’ll just list the things that I want to get rid of in the refrigerator.
Rodney Zemmel, McKinsey: Too many people are still using it to look something up. The trick is to have a dialogue with it and to get comfortable building agents that can execute simple tasks. Let AI handle the 80% of tasks we’re mediocre at, so we can excel at the exciting 20%, as one of my colleagues likes to say.
Matt Barrington, Americas chief technology officer at EY: Context management is paramount. I keep separate AI “workspaces” for different focus areas — such as technical Q&A or drafting client communications.
I also give the AI clear instructions about the style and depth of response I want, like “Provide a concise, bullet-point summary,” or “Act as a finance expert,” or “cite credible sources or references and provide links.”
Dan Priest, PwC: It is changing muscle memory.
I’ve spent a lot of years developing a certain writing style, a certain research technique, and I had to change that. And I am better for having changed it, but it doesn’t happen overnight.
It was just like anything that you learn, you have to be disciplined about learning it, and then it sticks.
Todd Lohr, KPMG: The biggest challenge is connecting all my individual data sources that are disparate. If I want to build my own personal AI, the challenge is having access to the right information and knowledge.
I have been deliberate about addressing this challenge when I took over in my current role. I put everything in one folder and personally curated the content that I agreed with and liked.
Matt Barrington, EY: The main challenge is keeping pace with the innovation. There’s a constant flow of new models, tools, and capabilities, and it can be tough to pinpoint which option is best for a given task. I follow newsletters, participate in AI-focused events, and learn from AI practitioners — but in my view, hands-on experimentation is the most effective way to stay informed and find what genuinely works.
Also, it is important to remember that while these models are confident and impressive, they can be wrong. Always validate the information and output before you utilize it.
Dan Priest, PwC: The questions have sort of shifted. A year ago, they were asking, “What’s the killer use case?” “What’s the most industrialized use case?” “What’s the use case that’s going to produce the most savings or the greatest deficiencies?” Now, the questions we’re getting are less about those technical use cases and they’re much more about “How do you evolve the business strategy to take advantage of AI capabilities?”
Rodney Zemmel, McKinsey: They want to understand how AI agents can integrate with their workforce, acting like talented interns who need proper training to be effective. We’ve also seen the conversation move from just productivity to growth and productivity, and to finding ways to do things better and faster than humans to doing things that no human could possibly do.
Do you have something to share about what you’re seeing in consulting? Business Insider would like to hear from you. Email our consulting team from a nonwork device at deliverable@businessinsider.com with your story, or ask for one of our reporter’s Signal numbers.
