Anthropic Just Started a Consulting Firm. Here's What It Means for Mid-Market Tech Decisions.

On May 4, Anthropic announced a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs, with additional backing from General Atlantic, Leonard Green, Apollo Global Management, GIC, and Sequoia Capital, to launch a new enterprise AI services firm.¹ The unnamed company will embed Anthropic engineers inside mid-size businesses to redesign workflows around Claude.²

Fortune called it Anthropic taking a shot at the consulting industry.³ That's accurate, but the more interesting question is what it means for the businesses that are going to be the targets of all this attention.

If you run a mid-market business, or you advise one, the day-after question is no longer "should I hire a CTO" or "should I bring in McKinsey." It's "what mix of these new options actually fits how my business works." This article walks through what just happened, who Anthropic is targeting, and how to think about your own technology partnership decisions when the model vendor is also the implementation firm.

What just happened

Anthropic and three Wall Street firms announced a new entity (still unnamed at announcement) backed by roughly $1.5 billion in committed capital. Anthropic, Blackstone, and Hellman & Friedman are putting in significant portions, with Goldman Sachs participating as a key investor.¹

The firm's stated model is to embed Anthropic engineers, the same kind of forward-deployed engineering talent Anthropic uses with its largest direct customers, inside client companies to redesign workflows around Claude. Goldman's global head of asset and wealth management, Marc Nachmann, framed it this way to CNBC: "There's a big shortage of people who know how to apply these tools into businesses and then transform them."² He added that the venture would help "democratize access to forward-deployed engineers" for companies that can't afford the talent, or the consulting fees, to build AI systems on their own.

The initial target sectors are healthcare, manufacturing, financial services, and real estate, with a heavy bias toward private-equity-owned mid-market companies.¹ The PE bias makes sense from the venture's distribution side. Blackstone, Hellman & Friedman, Apollo, and General Atlantic collectively own hundreds of mid-market companies that they can introduce to Anthropic's engineers without a sales cycle.

The structural play is interesting. This is not Anthropic launching a managed service. It's Anthropic plus the Wall Street firms whose portfolios constitute the customer base launching a vehicle that will land Anthropic engineers inside those portfolio companies, presumably at preferred rates and with guaranteed model capacity, in exchange for the kind of transformational AI work that has historically gone to McKinsey, BCG, Bain, Accenture, or the Big Four.

Why this matters even if you're not the target

The first reason: the announcement formalizes something that has been creeping up on the consulting industry for two years. The Big Four (Deloitte, PwC, EY, KPMG) and the strategy houses (McKinsey, BCG, Bain) have collectively poured over $10 billion into AI initiatives since 2023.⁴ BCG disclosed in April 2026 that 25% of its $14.4 billion 2025 revenue, roughly $3.6 billion, came from AI work. Accenture cut about 11,000 roles while committing $3 billion to AI and pledging 80,000 AI-focused hires. McKinsey reports that 25% of its fees are now outcome-linked, with the other three quarters still on traditional partner-hour billing.

The second reason, and the one that should land for any business owner reading this: AI is restructuring the question of who you hire to fix your technology. The traditional answer was "a consultant for strategy and a contractor or vendor for implementation." The new answer Anthropic is offering is "the model vendor's own engineers, embedded, redesigning the workflow around the model itself." OpenAI is doing a similar thing with its forward-deployed engineering service, with starting projects reportedly priced from $10 million.⁵

Both moves are responses to a real problem, and Nachmann is pointing at it correctly. The talent gap is the bottleneck. Most businesses do not have anyone on staff who can actually implement an AI agent against their accounting system, their CRM, their inbox, and their phone in a way that doesn't break the next time the vendor ships an update. The consulting incumbents have been slow to put bodies on this work because their economics depend on partner-rate billing of senior strategists, not staff-rate engineering.

Anthropic is offering to fix that bottleneck for the mid-market by sending its own engineers. The trade-off, for the business on the receiving end, is what they are giving up in exchange.

What you give up to get an Anthropic engineer in your office

This is the part that doesn't show up in the announcement coverage.

Vendor lock-in is the obvious one. When Anthropic engineers redesign your workflow around Claude, your workflow runs on Claude. Switching to a different model later is not an off-the-shelf operation; it's another redesign. The companion piece to this article, The AI Subscription You Bought for Your Business May Not Cover the Tools You're Actually Using, covered what happened on April 4 when Anthropic re-priced how third-party AI tools consume subscriptions. The same pattern applies here on a longer timescale. The vendor sets the price. You absorb the change.

Your data leaves your building. The Anthropic engineering model assumes Claude is the brain. Claude runs in Anthropic's cloud. Customer records, financial data, support transcripts, vendor agreements: all of it flows through the model to be reasoned over. For some businesses that's fine. For others, regulated or otherwise, it isn't. The piece on the small business case for local AI inference lays out the math on the alternative. The point here isn't that the Anthropic-services model is wrong; the point is that it presupposes a deployment posture that you should have evaluated before signing.

Business context is not a forward-deployed engineer's strength. A talented Anthropic engineer can implement an agent against your CRM in a week. But the question of which workflow to automate first, which one will move revenue, which one will get adopted by the staff who actually have to use it, that's a business question, not an engineering question. Goldman's pitch (democratizing access to forward-deployed engineers) is framed as a talent-supply story. Talent supply is necessary but not sufficient. The deployments that ship value are the ones where someone first sat with the operations team for a week and learned where the actual constraints are.

Retention is by project, not by relationship. When the engineering team rotates off, the institutional knowledge of why your workflow was designed a certain way leaves with them. This is the structural problem Anthropic's model inherits from the Big Four play it's competing against. The standard fix is documentation. The actual outcome is documentation that nobody reads, leaving the next person to redo half the analysis.

The five questions to ask before signing with anyone

This applies whether you're evaluating Anthropic's services firm, a Big Four engagement, a fractional CTO, or a vendor's professional-services team. The questions are the same.

1. Whose model are you building me on, and what's my exit?

If the answer is "ours, and you can't easily move," that's a strategic decision, not a technical one. Make it knowingly. The April 4 Anthropic policy change is what happens when the vendor adjusts the price after you've committed. If your engagement is built on a single model vendor's stack, your exit cost is not zero.

2. Where does my data go, and who can read it?

Not "is it encrypted in transit." Where does it sit at rest, who at the vendor has access to it, what's the retention schedule, what's the audit trail when something goes sideways. If the answer involves "we use industry-standard practices" without specifics, the answer is "they don't know."

3. Who's responsible for this six months from now?

After the implementation is done, when the integration breaks because an upstream vendor shipped a backwards-incompatible update, who fixes it? Is it a support ticket queue? A retainer? An internal hire? "We'll come back if you want" isn't an answer. That's just another consulting engagement.

4. What's the operating cost after handoff?

Implementation cost is usually quoted up front. Operating cost is usually understated. AI workloads, in particular, have variable cost driven by usage, model selection, and the vendor's pricing. Get the operating-cost projection in writing and ask what triggers re-pricing.

5. What does the documentation look like, and is your team going to use it?

The number-one failure mode for outsourced technical work is undocumented deployment. The number-two failure mode is documented deployment that the staff who have to operate it have never been walked through. Insist on a handoff session, recorded, with the people who actually do the work, not just the manager who signs the invoice.

Where this leaves the mid-market

The Anthropic-services play is a real option that did not exist last year. For PE-owned mid-market businesses, it's likely to become a default path, at least for a while. The economics work for everyone involved: Anthropic gets distribution, the PE firms get an implementation channel, the portfolio companies get engineers they couldn't afford on their own.

For businesses below the PE-portfolio tier, the picture is mixed. Anthropic's services firm will not, by economics, send engineers to a 30-person service business. The Big Four will not either. What's left is the model that has always existed and is unchanged by this announcement: practitioners who know your business, your industry, and your stack, and who treat the vendor relationship as one input rather than the entire shape of the deployment.

That's the version I run, but this article isn't a pitch. The point is that the question every business owner should now be asking is the one in the section header above. If your answer to "whose model am I building on" is "I'm not sure," the Anthropic-services announcement is the moment that uncertainty stops being free.

What I'd do this month

If you operate a business that's started using AI, or you're about to start: pick one engagement you're considering and run the five questions against it. Then decide. The decision is yours, and it should be made with the answers in front of you, not after the engineering team is already inside.

If you advise other businesses, particularly in the mid-market, this announcement is going to come up in conversations starting now. Be ready to walk owners through what they're actually signing up for. The Wall Street press release does not explain it.

Sources

1. The joint-venture announcement, structure, and target sectors. Per the Anthropic and Blackstone press releases of May 4, 2026, plus coverage at CNBC, Bloomberg, and Blackstone's release. Total committed capital ~$1.5 billion. Initial sectors: healthcare, manufacturing, financial services, real estate. Heavy PE-owned-target bias confirmed by CNBC. ↩
2. Marc Nachmann quote on talent shortage and democratizing forward-deployed engineers. Per Marc Nachmann, Goldman's global head of asset and wealth management, in his CNBC interview accompanying the announcement. ↩
3. Fortune characterization of the venture as "Anthropic taking a shot at the consulting industry." Per Fortune coverage of the May 4 announcement. ↩
4. Big Four and strategy-house AI investment numbers. Per Future of Consulting's 2026 update, BCG's April 2026 disclosure (25% of $14.4B 2025 revenue from AI), Accenture's 11,000 role cuts and $3B AI commitment, McKinsey's 25%-outcome-linked figure. ↩
5. OpenAI forward-deployed engineering pricing. Per Finance Story coverage of OpenAI's enterprise consulting service launch. Project floor reported as $10 million for custom GPT-4o models, AI workflow agents, and domain-specific copilots. ↩