The Consultants Were Never Going to Die

The role of consultants in the AI era

Hey, Prasad here 👋 I’m the voice behind the weekly newsletter “Big Tech Careers.”

This week’s post comes from a special guest Sandipan Bhaumik, author of the agentbuild.ai newsletter, where he shares insights from his experience on how to get your AI workloads past the demo stage to production. I highly recommend following him on LinkedIn and subscribing to his newsletter.

In this article, he shares why the AI labs need the consultants.

If you like the article, click the ❤️ icon. That helps me know you enjoy reading my content.

Over to you, Sandipan!

---

Six months ago, the prevailing story was clear: AI had made the traditional consultancy obsolete. Why pay McKinsey day rates when a frontier model could synthesise strategy in seconds? The argument had enough surface logic to fill a dozen conference keynotes.

Then OpenAI launched DeployCo, a $4 billion joint venture with Bain & Company, Capgemini, and McKinsey as founding investors. Anthropic followed with its own enterprise venture, backed by Blackstone and Goldman Sachs, alongside a $100 million partner network anchored by Accenture, Deloitte, and PwC.

The firms that were supposedly extinct are now co-owners of the platforms that were supposed to replace them.

This deserves more than a moment of irony.

---

What the “death of consulting” argument got wrong

The argument was always a category error. It assumed the consultancy’s core product was analysis - that what Bain or McKinsey sold was structured thinking, market research, frameworks. If that were true, yes: a sufficiently capable model would have undercut them within a budget cycle.

But that was never the product. The product was access.

Access to the CFO who trusts the partner from decades of working together. Access to the board conversation that happens before the RFP is written. Access to the unwritten political map of a large organisation - who actually holds the budget, who will block the implementation, which division chief needs to feel heard before a decision can move.

That cannot be distilled into a model. It took decades to build. It lives in relationships, in dinners, in the credibility that comes from having been right (and wrong) in front of the same people, repeatedly, over time.

---

Why the AI labs need the consultants

Both OpenAI and Anthropic built something genuinely extraordinary. Then they hit a wall that had nothing to do with the quality of their models.

Enterprise AI deployment is not a technical problem at its core. The model works. The infrastructure can be provisioned. What fails consistently, expensively, and quietly is the organisational layer. The wrong executive owns the project. The business case was written to justify a decision already made. The frontline teams who are supposed to adopt the system were never asked whether the system solved their actual problem. The governance structure for when the AI is wrong doesn’t exist.

These are not engineering problems. They are change management problems, political problems, trust problems.

The consultancies have been solving exactly these problems - in exactly these organisations, with exactly these leadership teams for forty years.

DeployCo’s stated mission is to “identify where AI can make the biggest impact, redesign organisational infrastructure and critical workflows, and turn gains into durable systems.” Read that sentence without the AI framing. It is a description of what consultancies have been selling for so long.

---

The PE forcing function

There is a second mechanism at work that the “consultants are dead” narrative missed entirely.

Private equity now owns enormous parts of the mid-market. Those firms face mounting sponsor pressure to show AI-enabled value. And they want it as a measurable feature of the business at exit, not as a future roadmap item. That creates urgent demand, at scale, that the AI labs cannot service directly. The labs do not have the client relationships, the sector knowledge, or the change management capability.

Goldman Sachs appears as a partner in both DeployCo and the Anthropic venture. That is not a coincidence. Goldman controls the financial conversations that precede every major enterprise investment decision.

The AI labs are not buying technology capability from these firms. They are buying the room.

---

What this actually tells leaders

Two things are now demonstrably true, and they are in tension.

The AI labs built models that can genuinely change how organisations operate. They are also, structurally, unable to make that happen without the firms they spent two years threatening to displace.

For leaders making AI investment decisions, the implication is uncomfortable: the quality of the underlying technology is now largely secondary to the quality of the deployment. A superior model, poorly deployed into a politically resistant organisation with no clear ownership and no change management, will produce nothing. An adequate model, deployed by a team that knows the organisation’s real power structure and has the trust of its senior leadership, will produce results.

Technology has never transformed an organisation. Trusted humans, using technology, have. The AI labs just spent $4 billion learning what every failed ERP implementation already knew.

---

I would like to extend a big thank you to Sandipan Bhaumik for sharing his insights with Big Tech Careers readers.

I encourage you to subscribe to his newsletter, agentbuild.ai to keep upto date on strategies to take your AI workloads to production

---