Is McKinsey the key to our AI future?

Enterprise adoption, not AI model performance, is the bottleneck to AI progress

Everyone loves to hate on consulting firms. There’s been no shortage of memes making fun of corporations taking advice from 23-year-olds. Since LLMs have taken off, there are more grandiose predictions about how AI will put consulting firms out of business.

I’m here to defend the indefensible. Not only do I think consulting firms will thrive in the next few years, but I’ll go further.

If we want rapid AI adoption in the economy, we can’t just rely on AI native startups. Instead, it can only happen by getting large, non-tech companies¹ to rapidly adopt AI into their workflows and products, which is exactly what our consulting firms are useful for.

Model quality is not the issue; enterprise adoption is

The biggest blocker today to AI being used more ubiquitously by businesses and customers is not model quality. The latest models, like O1 pro, Deepseek R1, and O3 mini, are sufficiently good that it makes sense for the average person to be using them 5x more than they are today.

Instead, the biggest blocker is the enterprise application of these models! Your average corporation is way behind leading tech startups in using AI, both in empowering their workers with AI tools (ex: software engineers using GitHub Copilot/Cursor) and in building AI into their products. As an example, the models are good enough to solve a big chunk of support tickets at almost every corporation. How many of them have AI support up and running? This implementation gap means you could have AGI tomorrow, and it wouldn’t have nearly as big an impact on the real economy as we’d hope.

What is enterprise AI adoption today? It’s hard to say, though the answer is probably in between the government estimate of 6% (which looks at all businesses small and large) and the Ramp estimate of 25% (which has a highly biased sample of tech-driven businesses).²

Source

Consistent with these estimates, my anecdata from people outside of tech makes me think adoption is pretty minimal. Talking to friends at banks, hedge funds, and corporate law firms, it seems like many companies are still stuck in the “I use ChatGPT to edit my emails” phase of AI that’s >12 months out of date.³

The “clueless” corporations need help

Undoubtedly, some of this gap between AI capabilities and AI adoption is that people outside of tech aren’t “AI-pilled” enough and don’t fully grasp how important it will be.

But I don’t think that’s actually the biggest reason corporations haven’t adopted AI. A recent McKinsey C-Suite survey showed that almost every corporation knows AI will be transformative to their business and that they need a plan. But ~50% think they are moving too slowly, with “talent skill gaps” being the biggest issue (which I interpret as “we don’t know enough about how to do this well”).

My general framework for all of this is that there are really three kinds of companies today:

1. AI-native Companies: These are companies with AI at the core of the business. They were either started recently or have products that have heavily incorporated AI from the beginning. Companies like Cursor, OpenAI, Anthropic, Microsoft, Palantir, etc.

2. AI-fluent Companies: These are companies, mostly in tech, that are agile enough to incorporate AI into their products. Salesforce may not be an AI company, but it knows enough to figure out how to incorporate AI into its product.

3. The Clueless Non-Tech Companies: These corporations are not tech-savvy and never have been. They’ve always been last to jump on any tech trend and don’t know where to start when it comes to AI. Think of older companies like American Airlines, Caterpillar, Pfizer, Ford, Exxon Mobil, United Healthcare, etc.

To accelerate AI’s usage in our lives, we have two options. The first is that firms in #1 and #2 just take all the business from #3. But is that realistic? Are we going to have an AI-native American Airlines? An AI-native Pfizer? An AI-native Caterpillar?

I doubt it. These companies often have moats of some kind (regulatory, network effects, switching costs, etc.) that make them hard to topple and have allowed them to weather other technological shocks. And even if AI-native startups took on every one of these “clueless” companies, it might take too long. Tesla, one of the fastest-growing tech companies ever, took almost 20 years to become the dominant player against the old-school car companies it competed with.

If you really want AI adoption fast, we have to go with the second option: having existing corporations adopt AI en masse, both by embedding it in their products and allowing their workforce to leverage the best AI tooling in their work.

The AI Consulting Boom is happening

Which brings us back to McKinsey and other big consulting firms. Demand from corporations is driving a boom in AI consulting, as the NY Times wrote about last year:

IBM, which has 160,000 consultants, has secured more than $1 billion in sales commitments related to generative A.I. for consulting work and its watsonx system, which can be used to build and maintain A.I. models. Accenture, which provides consulting and technology services, booked $300 million in sales last year. About 40 percent of McKinsey’s business this year will be generative A.I. related, and KPMG International, which has a global advisory division, went from making no money a year ago from generative-A.I.-related work to targeting more than $650 million in business opportunities in the United States tied to the technology over the past six months.

The reaction to this from a lot of tech people is one big roll of the eyes, and that McKinsey teaching any company about AI is basically “the blind leading the blind.” The best AI talent isn’t going to McKinsey, so what do they know?

And while I get the sentiment, I think this misunderstands the role consulting firms play. These “clueless” corporations need people who speak their language and know how to push tech changes at large enterprises (which is a skill whether tech people like it or not). They need help dealing with the massive regulatory, privacy, and compliance issues (particularly in the EU) that often stop risk-averse companies like them from moving forward.⁴

And many of these corporations are even further behind than you think, so “AI consulting” work is actually just cloud and ERP modernization projects to get your data in a place where you can use LLM models effectively.⁵

The Palantir-ification of Consulting Firms

For all these reasons, I genuinely think traditional consulting firms will have an important role to play. However, I also think these firms will need to adapt and may not win out during this boom (despite my provocative title). That’s because consulting firms need to actually build AI capabilities for companies, not just advise. From the same NY Times article:

“In the mid-90s, C.E.O.s would say, ‘I don’t know what a website is or what it could do for my business, but I need it,’” Mr. Vaz said. “This is similar. Companies are saying: ‘Don’t tell me what to build. Tell me what you can build.’”

The article describes how consulting firms have been doing more of this, like McKinsey building an AI chatbot for the bank ING. But the results are, unsurprisingly, pretty mixed. I can imagine Palantir taking even more of this market given their focus on data, AI, and actually building great products for clients. (See this writeup for more on how they operate).

There’s also an opportunity for pretty much any reasonable technical person to become an AI consultant. You might not engage with the giant corporations of the world, but plenty of people on Twitter are doing quite well teaching workers and startups how to better leverage AI.

Scaling laws might be enough to get our AI models to AGI, but I don’t think they will solve AI adoption. To get that, we may need help from the consultants, slide decks in hand, helping the “clueless” corporations into the AI future.

1. Pretty much everything in this piece applies to getting government adoption of AI (even more so)

2. Note that this is enterprise adoption, not personal adoption. Personal adoption is higher with 39% of Americans being monthly users of GenAI products. But I think the point still holds because people using ChatGPT for one-off queries in a personal capacity can only have so big an impact on the economy.

3. I do have one friend in NYC who works at an European investment firm and only just got access to an enterprise version of ChatGPT they are allowed to use at work.

4. And yes, sometimes they may need to bring in McKinsey to tell them something that may be obvious to employees but need McKinsey’s name behind it. I’m not oblivious to criticisms of consultants, I just think they are overblown. Consultants, tell me where I’m wrong!

5. The Accenture CEO had this to say on a Q423 earnings call: “It is important to remember that while there is a near universal recognition now of the importance of AI, which is the heart of reinvention, the ability to use GenAI at scale varies widely, with clients on a continuum… Starting with a strong digital core [is important], from migrating applications and data to the cloud, [to] building a new cognitive layer, [to] implementing modern ERP and applications across the enterprise, to a strong security layer. And nearly all clients are finding it difficult to scale GenAI projects because the AI technology is a small part of what is needed.“