The brilliant bottleneck
When AI labs partner with the world's largest consulting firms, it's a smart distribution play. But it also narrows the advice before the conversation begins.
Last week the Wall Street Journal ran a piece with a headline that made me pause mid-scroll: "AI Needs Management Consultants After All." The gist: OpenAI and Anthropic are striking deals with major consulting firms — from management consultancies like McKinsey and BCG to Big Four firms like Accenture and Deloitte — to push AI adoption into the enterprise. The consulting firms bring the client relationships, the change management muscle, the industry fluency. The AI labs bring the models.
It's a smart play. Genuinely brilliant, actually. If you're an AI company trying to reach thousands of large enterprises simultaneously, you don't hire thousands of salespeople. You partner with organizations that already sit in those boardrooms every day. It's a distribution strategy disguised as a consulting partnership, and it solves a real problem: most companies want to adopt AI but don't know where to start or who to trust.
But here's where it gets interesting.
When your consulting partner is contractually aligned with one model provider, the advice starts to narrow before the conversation even begins. The recommendation becomes Claude or GPT — not because it's the best fit for your specific workflows, your data architecture, your regulatory environment — but because that's the product the partnership was built to sell. The consultant's job shifts, subtly, from solving your problem to deploying their partner's solution.
Enterprise AI done well often means stitching together multiple models, or building something bespoke, or concluding that the best answer for a given workflow isn't a large language model at all.
And that's a meaningful distinction. Enterprise AI done well often means stitching together multiple models, or building something bespoke, or concluding that the best answer for a given workflow isn't a large language model at all. That kind of thinking requires independence. It requires advisors whose economics aren't tied to any single provider's adoption targets.
I want to be clear: I admire what Anthropic has built (we work with Claude every day). And the WSJ piece speaks to something real — the complexity of enterprise AI adoption is significant enough that even the labs themselves can't do it alone. That's a validation of the entire consulting layer. The question is whether the consulting layer can stay objective when it's funded by the technology it's supposed to evaluate.
What this means for you
For any organization reading this and wondering what it means for you: ask your advisor who's paying them. Not in a cynical way — in a practical one. The best AI strategy starts with your operations, your customers, your constraints. It ends with the right tools for the job, however many vendors that involves. If the answer is always the same product, that should tell you something about the process, not the product.
Choose advisors who work on your behalf.
