TL;DR:

(US, AI implementation) Northslope has been acquired by the OpenAI Deployment Company. Founded by Palantir alumni with ~150 forward-deployed engineers, it is ODC's second acquisition as it builds the practitioner force needed to close the gap between frontier AI capability and real-world deployment. [Link]

(UK, training) iTrain has been acquired by Harbor. The UK specialist in legal technology training and user adoption joins Harbor's managed legal services platform, extending its AI enablement capabilities into the UK market. [Link]

(US, federal AI services) Highlight Technologies has been acquired by Empower AI, a portfolio company of KKR. The employee-owned federal IT services provider deepens Empower's AI-powered government services platform. [Link]

(US, Salesforce consultancy) Dupont Circle Solutions has been acquired by CrossCountry Consulting. The Salesforce-focused technology consultancy expands CrossCountry's revenue operations and lead-to-cash capabilities. [Link]

(UK, AI/data consultancy) NextWave has raised a £3m investment from YFM Equity Partners. The financial services-focused data, AI and automation consultancy will use the funds to scale its AI-led transformation practice for Tier 1 institutions. [Link]

(CA, applied AI) AltaML has raised a US$6.1m strategic investment from VerticalScope Holdings, deepening a relationship that began when VerticalScope was an enterprise client. The investment backs AltaML's expansion across industry verticals. [Link]

(US, AI-native holdco) Thrive Holdings has raised $2bn, with OpenAI taking an equity stake. The Thrive Capital vehicle is acquiring accounting and IT services businesses and rewiring them with AI — the latest signal that OpenAI is not just selling API access but actively building ownership in the implementation layer. [Link]

Have we missed an interesting deal? Or do you want to correct details? Just hit reply to let us know.

OpenAI: Booz Allen Hamilton announced a new partnership to deploy secure, mission-ready AI for defence and national security clients — bringing deep mission expertise to OpenAI's frontier models for high-stakes government environments. [Link]

Anthropic: Fractal Analytics was elevated to Preferred Services Partner in the Claude Partner Network, unlocking direct Anthropic technical access for large-scale client deployments. [Link] UST became Anthropic's second Global Premier Partner, committing to train 20,000 employees on Claude and embed it across its engineering platforms and operations. [Link] AWS-pedigreed cloud consultancy Caylent is carving out its AI-enablement lane in the Claude Partner Network, translating its cloud deployment playbook into Claude implementations. [Link] Globant formalised a multi-year alliance to deliver enterprise AI through Claude-powered AI Pods — cross-functional delivery teams focused on secure, outcome-driven deployment. [Link]

Google Cloud: Cognizant expanded its partnership to pair Gemini Enterprise rollouts with its Frontier Certified Engineers — specialists who deploy and scale AI in live client environments. [Link]

Lovable: Officially launched a Solutions Partner Programme for agencies and consultancies building on its AI development platform. [Link]

Market signals: ISG's Q2 2026 Index showed global tech services ACV up 43% to a record $42.4bn — the fastest growth ever recorded, driven by cloud XaaS up 65%. AI demand is converting to booked revenue at scale. [Link]

In January, a company funded by Palantir's former CFO and Palantir's first independent board member raised a $22m Series A to build applications on Palantir's platform. In July, OpenAI bought it.

Axios broke the news on Tuesday, and by Wednesday the commentary had settled into a comfortable consensus: the labs are at war for FDE talent, and Northslope is the latest casualty. The talent war is real. It is also the least interesting thing happening here. The deal is full of details that don't fit the tidy narrative, and this week we'd rather catalogue the irregularities than repeat the consensus.

On the surface, Northslope looks like a rerun of DeployCo's founding acquisition. Tomoro brought roughly 150 forward deployed engineers; Northslope brings 137 people. Both were under three years old. Both had real, recognisable logo, enterprise customers with software in production.

The similarities end there. Tomoro was an OpenAI specialist, built by founders with strong consulting backgrounds who had already sold one firm to Accenture. It was also bootstrapped on zero venture capital. When DeployCo acquired it, the strategic rationale couldn't be questioned.

Northslope was somebody else's flagship. It was Palantir's first and only Vanguard: Elite partner, ranked - a 137-person firm - above Accenture, Deloitte and PwC in Palantir's partner hierarchy. It launched a Google Gemini Enterprise practice in April, but didn't have any disclosed relationship with OpenAI. It had raised from the innermost circle of Palantir alumni: funds run by Palantir's former CFO and first independent board member. Revenue grew nearly 7x in 2025, by the company's own account. In short, no obvious reason to sell now.

While FDEs are quickly becoming the hottest commodity in the AI Enablement world, we think that it can't be the only explanation for why this deal happened and what it means for others in the ecosystem. The most interesting bit are, as usual, in the footnotes.

Here is a strange fact about Palantir's Vanguard partner program: as far as Palantir's investors are concerned, it does not exist.

The program appeared quietly around November 2025 — no press release from Palantir, just a new page on the website and a pair of partner-issued announcements. Northslope was crowned the first and only Elite partner in December. Ten weeks later Palantir filed its FY2025 10-K. The word "Vanguard" appears nowhere in it. It has never been said on an earnings call. When a BofA analyst asked management in February about what they were seeing from "customers or software partners," they answered at length about customers and politely ignored the other half of the question. In two decades as a public and private company, partners have never once been a serious topic between Palantir and its investors.

Which is odd, because the official logic for the program is sound and Palantir repeats half of it constantly. Karp told investors in May that "we just cannot meet demand," that Palantir runs on 70 salespeople of whom "only seven actually even really sell." A company growing 104% in the US with a deliberately tiny salesforce and a services-heavy product needs someone to do the deployment work. The 10-K's risk factors admit as much, noting that partners give Palantir access to "a pool of qualified subcontractor personnel that we can call upon to enhance and augment our implementation and engineering services."

So why build the ecosystem and then refuse to talk about it? Here's what we suspect is going on, and it runs through Palantir's long war with short sellers.

In November 2025, Michael Burry disclosed puts with a notional value of roughly $912m against Palantir. Karp went on CNBC on earnings day and called him "batshit crazy"; the stock fell 8% anyway. Burry's most pointed attack, laid out later on his Substack, was an accounting one: he claimed Palantir splits its FDE costs "atypically" across R&D and sales and marketing rather than cost of revenue, flattering gross margins. His line: "If Palantir were following the accounting standards of Accenture or Deloitte, gross margins would collapse. The SaaS/software charade would be obvious."

The short case, in other words, is that Palantir is a services business wearing a software multiple, and the FDEs are the tell. Seen through that lens, a partner ecosystem is the cleanest structural rebuttal available: if third parties can profitably deploy your technology, it's a product, not a body shop, and the deployment labour sits on someone else's P&L. There is a certain comedy in the mechanics. Burry says Palantir's margins would collapse under Accenture's accounting; Palantir's answer is to hand the deployment work to Accenture. And to Northslope, the 90%-ex-Palantir firm that was, by construction, the perfect proof the model worked.

But the ecosystem had to exist without being discussed, because discussing it means admitting the deployment bottleneck, and putting the word "partner-delivered" anywhere near the revenue line invites exactly the questions Karp spends earnings calls pre-empting with "no intertwined economics" and "the numbers are pure."

The trouble is that "no intertwined economics" cuts both ways. Palantir designed the ecosystem so that customers pay Palantir for the platform and pay the partner separately for services. Clean for the software story. It also meant that when a buyer with $4bn arrived, nothing — no equity, no revenue share, nothing that bit — bound Northslope to Palantir. The 10-K's own risk factors warned that partners "may choose to pursue alternative technologies... in collaboration with others, including our competitors." Filed in February. Fulfilled in July, by the only company on the Elite page.

The open questions write themselves. Does Palantir eject Northslope from Vanguard: Elite, or quietly retire a tier that now showcases an OpenAI subsidiary? Did Palantir have any right of first refusal, and if not, why not? And was April's Gemini practice the tell that Northslope's investors had already priced its independence?

Let's look at the acquisition through the product lens. OpenAI launched Frontier in February: an enterprise platform for building, deploying and governing AI agents, complete with shared business context drawn from CRMs and warehouses, agent identity and permissions, onboarding "like employees." Dwarfed by the popularity of the consumer application, Frontier's capabilities somewhat resemble Palantir's AIP.

But Frontier is not yet Foundry. There is no fifteen-year data-integration and ontology stack underneath it, no government accreditation moat. It is an agent control plane sitting on top of existing systems, not a data operating system. On paper, Northslope's FDEs are the perfect people to close the gap by building the business-context layer, i.e. the part where an enterprise's actual operations get encoded into software. Neither OpenAI nor Northslope has said Northslope will work on Frontier: that's our guess, so hold it loosely. But it's not a long leap.

The detail that deserves more attention is what the deal says about where value sits in this market. OpenAI DeployCo bought a Palantir and Gemini shop. If the value in AI services were specialism — deep knowledge of a particular system or model — a Palantir specialist would be worthless to OpenAI. The acquisition only makes sense if the scarce asset is the meta-skill underneath the specialism: knowing how to deploy technology inside a complex enterprise, extract the business context, embed with operators, and ship against a mission, regardless of whose stack is underneath. Northslope proved the skill was portable when it stood up a Gemini practice in a quarter. "Palantir-native" was always a claim about people, not lock-in. But then what does this tell us about customer lock in?

Whether OpenAI is betting that Palantir's corner of the market — regulated, operational, mission-specific — is where enterprise demand is most proven, or if it couldn't find any worthy targets in its own ecosystem, is genuinely unclear. We'll be watching closely how this develops.

When OpenAI and Anthropic announced their services ventures in early May, one of the questions we raised in our deep dive was "Who the hell are they actually going to buy?". Since then, three implementation firms have been acquired: Tomoro (OpenAI), Fractional AI (the Anthropic-backed firm 98 people, notably smaller and quieter), and now Northslope. With that the target acquisition profile is slowly emerging: under three years old, roughly 100 to 150 people, platform-native skill density, named enterprise outcomes. Founders of young AI implementation firms now have a rough goalpost for what they are building towards.

The growth behind these firms deserves its own note, because services businesses are not supposed to grow like this. Northslope did ~7x in a year. The mechanics, from what we can observe: tight vendor relationships supply customer pipeline, logo-after-logo land grabs among large customers, aggressive hiring from the platforms' own alumni pool. A VC-company mindset applied to the services industry, bolted onto the unglamorous ability to actually deliver.

But with every answer comes a new question, and we'd encourage founders to hold them in mind when planning their next strategic move.

Will the labs integrate these firms or run them standalone? What happens when the lock-ups expire? The entire acquired asset is people and FDEs are the most poachable professionals in tech. Add two amplifiers: if the labs go public, acquired-company equity converts into something that moves 8% on a short seller's tweet, and retention math changes with the stock's volatility. And VCs are already circling anyone associated with these firms. The alumni-diaspora effect that produced Northslope out of Palantir will now run on Northslope itself.

Will customers be loyal to a brand or to a specific partner, as they are with traditional consulting firms? Will customers be happy getting locked into an ecosystem or will be see emergence of "independent" AI implementation consulting firms be able to implement different tech based on each project's specific brief?

If I was building an AI implementation business, that's what I'd be thinking about this weekend and beyond.