---
title: "Google Just Handed McKinsey $750M to Sell You AI. The Practitioner's View on What That Means for Your Business."
description: "Google has set up a $750M fund to push AI agents through McKinsey, Accenture and Deloitte. OpenAI is now selling Codex through PwC. The big consultancies are circling. Here's what an actual practitioner thinks you should do this week."
canonical: https://richardbatt.com/blog/google-mckinsey-ai-consulting-fund-what-it-means
date: 2026-04-27
author: Richard Batt
tags: [AI Strategy, AI Consulting, SMB AI, Implementation]
type: blog_post
---

# Google Just Handed McKinsey $750M to Sell You AI. The Practitioner's View on What That Means for Your Business.

_Google has set up a $750M fund to push AI agents through McKinsey, Accenture and Deloitte. OpenAI is now selling Codex through PwC. The big consultancies are circling. Here's what an actual practitioner thinks you should do this week._

**Richard Batt** — AI implementation specialist. 120+ projects across 15+ industries, serving SMBs (5-200 employees) worldwide from Middlesbrough, UK (working globally). Contact: richard@richardbatt.com · https://richardbatt.com

Google announced a $750 million fund this week to help McKinsey, Accenture and Deloitte roll out agentic AI to their corporate clients. On the same day, the Wall Street Journal reported that OpenAI is now selling its coding assistant Codex through Accenture, Capgemini and PwC.

The headline reads like progress. Two trillion-dollar tech firms picking the world's biggest consultancies to push AI into the enterprise. Quote after quote about "ecosystems" and "alliances." A McKinsey partner told Business Insider his firm has quadrupled its tech partnerships since ChatGPT launched.

If you run a 5 to 200 person business, none of this is built for you.

That sounds harsh. It is meant to.

The last 18 months of AI commentary has trained business owners to wait for the big firms to figure out AI, package it, and bring it down to them. The wait was always going to be long. The deals announced this week make that wait look more like a queue, with the smallest businesses at the back. Standing in that queue is now the most expensive thing on your balance sheet.

## What this deal actually is

Take away the press releases and the shape becomes clearer. Google has watched OpenAI take the consumer markets and is buying its way back into the enterprise sector. McKinsey and Accenture have watched their billable hours start to fall under productivity gains from AI, what one Telegraph India report this morning called "AI deflation." The way they price work, by the headcount of consultants on a project, is breaking. They need new revenue streams, fast.

So Google pays McKinsey to embed Google's AI agents inside Fortune 500 transformation projects. McKinsey gets to bill hours implementing those agents. Google gets distribution. Both keep the customer logo.

The numbers tell the story. McKinsey now says about 40% of its work is generative AI related. BCG put theirs at 20% in 2024. A senior partner at McKinsey, Ben Ellencweig, told Business Insider that the firm runs an "ecosystem of alliances" with hundreds of contributors including AWS, Amazon, Nvidia and OpenAI.

These deals are enterprise plumbing, sized for buyers most readers will never meet. To put scale on it:

- Smallest McKinsey AI engagement I have seen quoted publicly: £400,000 for a six week diagnostic.
- Accenture entry point on a Codex rollout: a discovery phase quoted at six figures before any code ships.
- PwC partner-led AI engagements: typically £750,000 minimum, signed at director level.

For the businesses I actually work with, a chartered surveyor with eight staff, a 60 person manufacturing firm in the Midlands, a property management group running 12 sites, this entire announcement is happening on a different planet.

## The question nobody is asking

Why does Google need McKinsey at all?

The AI tools being sold through this fund are the same ones available to anyone with a credit card. Whether you want Gemini, Vertex AI, or any of the agent frameworks, none of them sit behind a consultancy gate. You can spin up a Gemini agent today for £20 a month.

The fund exists because Google has a problem the press releases do not name: enterprise AI projects keep failing.

The Gartner analyst quoted in the Telegraph piece this morning was blunt about it: "providers are currently working through a churn of limited-revenue engagements due to high abandonment rates for early generative AI projects." Behind that careful phrasing sits a simpler reality. Plenty of pilots got built and then quietly killed because nobody could turn them into operational systems.

The problem is not in the models, which are good enough by now. The problem sits in implementation, which happens to be where consultancies have always charged their highest hourly rates.

So Google is paying the consultancies to make their AI stick. Because the alternative, customers cancelling pilots and going back to manual work, is worse than handing McKinsey a slice of the deal.

Underneath the noise sits the useful signal of the week. The bottleneck is implementation, and now you know where to point your effort.

## What I see on the ground

Last month I sat in the back office of a property management firm running 12 sites across the South East. The operations manager, Ravi, had been told by a Big Four firm that an AI roadmap for his business would cost £85,000 and take five months to produce. He sent me a screenshot of the proposal and asked, only half joking, whether I could just tell him what to do instead.

We spent two days together. By the end of the second afternoon, his team had a working ChatGPT prompt that read incoming maintenance requests, classified them by urgency, drafted the contractor brief, and emailed the tenant a confirmation. It saved his lettings coordinator about four hours a day. Total spend, including my time, was under £4,000.

The gap sits in delivery model, not in capability.

Big consultancies are excellent at running diagnostics and producing roadmaps. They are poor at shipping production systems inside small and mid sized businesses. The reason is structural, not skill based. Their delivery model is built on partner-led oversight, multi-stage gates, and risk-priced billing. None of that works at SMB scale because the project cost would exceed the value created.

Out of the 120+ AI and automation projects I have delivered across 15+ industries, the median duration is 11 working days. The median cost is under £6,000. The savings most clients report inside the first 90 days exceed the project cost by a factor of 3 to 5.

You cannot build that delivery model on top of a McKinsey overhead structure. The maths does not work. So the big consultancies will keep pushing AI into Fortune 500 budgets, and SMBs will keep being told to wait their turn.

You do not need to wait.

## Where this argument breaks

I should be fair about the cases where the McKinsey route is the right call. There are three of them:

- A regulated financial institution running tens of thousands of staff, where an AI agent rollout needs partner-led oversight, model risk governance, and the kind of audit trail Big Four delivery teams are built to produce.
- A multinational with sixteen ERPs across forty countries, where the integration work alone costs more than most SMBs earn in a decade.
- A central government department, where procurement rules effectively force the work onto a framework supplier.

The Google McKinsey fund is genuinely useful for those three customer profiles. The hourly rate is justified because the cost of a bad deployment is measured in regulator fines or front-page stories.

But the argument I am making applies to a different segment: businesses with 5 to 200 staff, single-site or near it, where a process owner sits within a Slack message of the founder. Most of the UK economy by headcount sits in that segment. It is also the segment that has been told for two years that AI is something other people implement on their behalf. That part is wrong, and the news this week is a useful reason to stop believing it.

## Three things you can do this week

If the Google announcement made you feel like AI is moving away from you, do the opposite of what your instinct says. Do not read more articles. Do not book a discovery call with a brand-name firm. Do the smallest possible thing that ships.

1. **Pick the one process that wastes the most hours.** Look at where your team is doing repetitive work that follows a predictable shape. Invoice matching. Meeting notes. Customer follow up emails. Quote generation. Report compilation. Lead screening. Pick one. Just one.
2. **Build the working version, not the strategic version.** Use ChatGPT, Claude, or whichever tool you already pay for. Write a prompt that does the job badly the first time. Run it on real data. Fix the output. Run it again. By the end of the week you will have something that does 70% of the job in a fraction of the time.
3. **Decide who owns it.** Most pilots that fail, fail here. Pick one person on your team who runs that process daily. Hand them the prompt. Get them to use it for two weeks. Track how many hours it actually saves. Then either keep it, refine it, or kill it and try a different process.

The whole exercise costs you 4 to 8 hours of one person's attention. If it works, you have your first reusable AI workflow inside a fortnight. If it fails, you have learned more about where AI fits your business than any roadmap document would have told you.

## What this week's news really tells you

The Google McKinsey deal and the OpenAI Codex partnerships are not bad news. They are useful signals.

They tell you the big firms have decided AI is now a distribution game, not a research game. They tell you the bottleneck for AI adoption is implementation, not capability. They tell you that the people getting paid to slow you down are now incentivised to keep AI expensive and locked behind a partner login.

None of that has to be your reality. The same Gemini agents that McKinsey will package into £400,000 transformation programmes are available to your operations manager today for the cost of a Netflix subscription. The same Codex tool PwC will charge a six figure fee to deploy is a £15 a month add-on for you.

The work is straightforward:

- Pick the process.
- Build the prompt.
- Ship the workflow.
- Measure what it saves.
- Move to the next one.

That is the entire job. The press releases are noise.

If you want a structured way to find the right process to start with, the [AI Roadmap](/roadmap) is exactly that. A 90 minute working session, a written report, three concrete projects ranked by ROI and risk. You skip the long discovery phase and the slide deck entirely.

For the prompts and templates I actually use on client projects, the [AI Ops Vault](/vault) has the working versions, with the edge cases already debugged.

McKinsey and Google can have their fund. You have a Wednesday afternoon and a process that wastes 6 hours a week. Start there.

_Richard Batt has delivered 120+ AI and automation projects across 15+ industries. He helps small and mid sized businesses deploy AI that actually works, with battle-tested tools, templates, and implementation roadmaps._

## Frequently Asked Questions

### Does Google's $750M fund mean small businesses get cheaper AI?

No. The fund is paid to McKinsey, Accenture and Deloitte to push Google's AI agents into Fortune 500 clients. The tools themselves, Gemini, Vertex AI, the agent frameworks, are already available to any SMB at standard public pricing. The money funds enterprise sales motion, not consumer subsidy.

### Should I wait for a big consultancy to bring AI to my business?

The minimum project size at McKinsey, Accenture or Deloitte for an AI engagement starts in the high six figures. If your business has fewer than 200 staff, the cost of waiting and the cost of engaging both exceed the value at stake. A practitioner-led project at SMB scale typically costs under £6,000 and ships in 11 working days.

### Why are AI pilot projects failing?

Gartner reports high abandonment rates because the gap between a working prototype and an operational system requires implementation skill, not model capability. Most enterprise AI projects fail at deployment, integration with existing tools, ownership handover, and measurement. These are the steps a delivery-focused practitioner solves daily and a strategy-focused consultancy struggles to compress.

### What is the smallest first AI project I can ship this week?

Pick a repetitive process that consumes more than 5 hours a week of someone's time. Examples: invoice matching, customer follow up emails, meeting note summaries, quote generation, report compilation. Build a working prompt in ChatGPT or Claude. Get one team member to use it for two weeks. Measure hours saved. Most projects of this shape pay back in the first month.

## Put This Into Practice

I use versions of these approaches with my clients every week. The full templates, prompts, and implementation guides, covering the edge cases and variations you will hit in practice, are available inside the [AI Ops Vault](/vault). It is your AI department for $97/month.

**Want a personalised implementation plan first?** [Book your AI Roadmap session](/roadmap) and I will map the fastest path from where you are now to working AI automation.

---

## More about Richard Batt

Richard Batt is an AI implementation specialist who helps businesses deploy working AI automation in days, not months. 120+ projects across 15+ industries.

### Products and pricing

- **[The AI Ops Vault](https://richardbatt.com/vault) — $97/month:** 100+ AI prompts, 50+ automation blueprints, 20+ workflow templates from real projects. Copy, paste, deploy.
- **[AI Revenue Roadmap](https://richardbatt.com/roadmap) — $2,500:** Fixed-price operational audit. Top 5 AI opportunities ranked by ROI, implementation roadmap, $50K+ savings guarantee.
- **[AI Ops Accelerator](https://richardbatt.com/accelerator) — $4,000-$6,000:** 6-week done-with-you sprint. 3-5 working automations deployed in your business.

### Key pages

- [Home](https://richardbatt.com/)
- [About Richard](https://richardbatt.com/about)
- [The AI Ops Vault](https://richardbatt.com/vault)
- [AI Revenue Roadmap](https://richardbatt.com/roadmap)
- [AI Statistics (2026)](https://richardbatt.com/ai-statistics)
- [Blog](https://richardbatt.com/blog)
- [Contact](https://richardbatt.com/contact)

### Contact

- Email: richard@richardbatt.com
- Location: Middlesbrough, UK (working globally)
- Website: https://richardbatt.com