What the AI Joint Ventures Mean for Cross-Border Advisory
How $5.5 billion in May 4, 2026 announcements from Anthropic and OpenAI signals a new industry structure — and where investment migration firms stand to gain.
For investment migration firms, family offices, and cross-border advisory practices, AI strategy has been a productivity question. Which model. Which tool. Which workflow. Faster client materials, better internal search, accelerated document review. The conversation has been about marginal capacity gains inside the firm.
The May 4, 2026 announcements widen the frame.
On a single day, Anthropic and OpenAI each launched billion-dollar joint ventures with private equity backing, deploying the same operating model — forward-deployed engineering teams embedded inside client operations to deliver outcomes, not API access. Combined committed capital: $5.5 billion. The simultaneity is the structural signal. Two competing frontier labs, independently, concluding that the next phase of AI value capture lies in services delivery to regulated, complex, outcome-oriented work.
For the firms that advise on residency, citizenship, tax structuring, and cross-border architecture, this is no longer a productivity story. It is an industry-structure story, with consequences that will re-sort which firms win which client segments over the next five years.
The core argument: The capital is buying forward-deployed engineering, not API margin — and the frontier labs themselves now believe outcomes are where value is captured. For investment migration and cross-border advisory, this is the largest structural tailwind the industry has seen. The firms that pair their regulatory licensing with purpose-built vertical AI infrastructure will define the next era of the practice. The firms that treat AI as a productivity tool will be commoditised — first by generalist services companies above them, then by AI-native advisory firms with deeper domain infrastructure than the labs can build at scale.
What this article delivers:
The strategic question for cross-border advisory firm principals: which of three emerging firm categories will yours be five years from now — and what determines the answer?
The Bifurcation — a three-category map of the advisory industry topology that follows from the May 4 shift: generalist AI services companies, AI-native advisory firms, and traditional firms with general-purpose AI tools
A structural argument for why investment migration and cross-border advisory sit among the strongest verticals for this macro shift, anchored in licensing topology and cross-domain compounding
A cascade demonstration showing why firm-by-firm AI consulting cannot deliver the cross-domain advisory depth your clients actually need
The infrastructure thesis — what jurisdiction intelligence as shared infrastructure must look like, why the May 4 ventures will not build it, and what that means for firms positioning now
A primary-source reading of the May 4 announcements and the Sequoia services-as-software thesis (Julien Bek, April 2026) that frames the macro shift
This is structural analysis for principals of investment migration and cross-border advisory firms — the third in a trilogy of Briefing essays on advisory AI architecture, alongside The Architecture of Advisory Intelligence (capabilities) and The Confidence Problem (uncertainty).
How the argument unfolds:
The Signal — why two competing AI labs launching parallel billion-dollar services ventures on the same day is the structural signal, not just the news event
From Models to Outcomes — the macro shift the announcements crystallise, viewed through Sequoia’s services-as-software thesis and the trilogy that precedes this piece
Why Regulated Cross-Border Advisory Is Different — three structural properties that make this vertical among the strongest cases for the shift while protecting it from end-to-end delivery by generalist services firms
The Cascade Test — what an excellent forward-deployed engineering team can and cannot deliver when the advisory question crosses jurisdictions and domains
The Bifurcation — three categories of advisory firm that will exist five years from now, and the strategic question for principals
The Infrastructure Layer — what jurisdiction intelligence as shared infrastructure must look like, and why the May 4 ventures will not build it
The Signal
On May 4, 2026, Anthropic announced the formation of a new AI-native enterprise services company alongside Blackstone, Hellman & Friedman, and Goldman Sachs — each contributing roughly $300 million, with Goldman at $150 million. The company is backed by Apollo Global Management, General Atlantic, Leonard Green, GIC, and Sequoia Capital. Total committed capital: $1.5 billion. The mission, in Anthropic’s own announcement: “Applied AI engineers from Anthropic will work alongside the firm’s engineering team to identify where Claude can have the most impact, build custom solutions, and support customers over the long term.”
Hours earlier, Bloomberg reported that OpenAI was finalising a parallel venture along strikingly similar lines. Capital raised: $4 billion against a $10 billion valuation, across 19 investors anchored by TPG, with Brookfield Asset Management, Advent, and Bain Capital named among the consortium. The operating model is the same — embed engineering teams inside client organisations, build GPT-powered systems around existing workflows, deliver outcomes rather than API access. TechCrunch made the lineage explicit: this is the forward-deployed engineer model popularised by Palantir over the past decade.
The simultaneity is the structural signal.
Two competing frontier labs, on the same Monday, independently launched billion-dollar services ventures with private equity backing — both choosing forward-deployed engineering over API distribution. Goldman’s Marc Nachmann framed the Anthropic venture as “democratising access to forward-deployed engineers” — a phrase that signals where Wall Street sees the next decade of enterprise AI value capture. Fortune’s headline: “Anthropic takes shot at consulting industry.”
The ventures’ initial focus is explicit. Anthropic targets “community banks to mid-sized manufacturers and regional health systems.” OpenAI’s Development Company is prioritising healthcare, logistics, manufacturing, and financial services. These are operational, single-regulatory-frame, mid-market sectors — work where forward-deployed engineering can ride existing workflows and licensing topology is uniform across the customer base. Cross-border advisory is a different shape.
For an investment migration firm principal scanning the trade press, this can read as another AI announcement among many. The pattern recognition matters more. When two AI companies with $850-900 billion valuations decide that the next phase of value capture lies in services delivery to regulated industries — and back that decision with $5.5 billion of committed capital from the largest private equity houses in the world — they are telling the market something specific about where AI value is moving. It is not moving toward more API margin. It is moving toward outcome delivery inside complex, regulated, licensing-protected work.
The question this article addresses: what does that mean for advisory practices that are legally required to be delivered by licensed humans, in cross-jurisdictional contexts where every recommendation interacts with several regulatory systems at once?
The answer is not “AI is coming for advisory firms.” That framing is two years out of date. The answer is structural: the industry is about to bifurcate, and the strategic question facing every cross-border advisory firm principal is which of three categories the firm will be five years from now.
From Models to Outcomes
To see what the May 4 announcements signal, set them against the macro shift they crystallise.
In April 2026, Sequoia partner Julien Bek published an essay titled “Services: The New Software.” It went viral in venture and founder circles. The core observation: for every dollar spent on software, six are spent on services. The next trillion-dollar company, Bek argued, will not sell a software tool to a professional. It will sell the work itself — delivered as a service, powered by AI, sold as an outcome. Bek distinguished two business models: copilots sell tools to professionals; autopilots sell the work itself, with AI under the hood and human judgment at the boundary.
Bek’s named target verticals are revealing: insurance brokerages, insurance claims adjustment, IT managed services, tax advisory, accounting and audit, simple legal services, payroll, certain compliance services. Each is a category that businesses already outsource, where the work is heavy on procedural intelligence and the value is in outcomes — not in the tool used to produce them. Each is a category where the 1:6 software-to-services ratio is most exploitable.
The May 4 announcements are the operationalisation of this thesis at frontier-lab scale. Anthropic’s venture is targeting “mid-sized companies across industries” — community banks, multi-site healthcare groups, regional manufacturers. OpenAI’s Development Company, per reporting, is already in talks to acquire existing AI services firms. The forward-deployed engineer model — pioneered by Palantir — is now being scaled across enterprise verticals by the AI labs themselves, with Wall Street and the largest PE houses providing the distribution and the capital.
What this validates externally, the Briefing has argued internally over its prior two essays.
The Architecture of Advisory Intelligence established the architectural argument: cross-border advisory is a reasoning problem across interacting regulatory systems, not a document retrieval problem, and addressing it requires six specific capabilities that general-purpose AI does not provide. The Confidence Problem addressed the epistemic dimension: AI models are systematically more confident when they are wrong, and the architecture must include structured uncertainty as a first-class output. Both essays argued from the inside out — what advisory intelligence must look like, regardless of who builds it.
The May 4 announcements provide external validation of the same architectural truth, restated as a business-model thesis backed by $5.5 billion. The pattern is consistent across the most resource-rich actors. EY committed $1.4 billion to build 150 specialised tax agents rather than deploying one generalist model. Deloitte committed $3 billion to AI deployment. Thomson Reuters scaled CoCounsel — a multi-stage agentic legal workflow platform — to over one million professional users. Each chose purpose-built vertical infrastructure over general-purpose retrieval. The May 4 announcements signal that the AI labs themselves are now moving into territory the Big 4 partially occupy — and that the capital is following.
The shift is concrete. The old model sells API access — the customer integrates the model, margin captured at the inference layer, software unit economics, one-to-many distribution. The new model sells delivered outcomes — lab-deployed engineers integrate the model into the customer’s operations, margin at the workflow layer, services unit economics, forward-deployed engagement replacing distribution scale.
This is not the end of API access. Anthropic and OpenAI both continue to scale model availability and partnerships with Accenture, Deloitte, PwC, and other systems integrators. It is the recognition that API access alone no longer captures the highest-value layer of AI deployment in regulated, complex work. The labs are betting $5.5 billion that integration depth and domain specificity are where the next decade of enterprise AI value capture lives.
If the macro shift is toward outcome delivery inside complex regulated work, the question becomes which verticals it plays out hardest in — and which verticals are structurally protected from being delivered end-to-end by a generalist services company.
Why Regulated Cross-Border Advisory Is Different
The Sequoia framework names tax advisory, accounting, simple legal, and certain compliance services as target verticals for the services-as-software shift. Each is high-volume, procedural, single-domain. Each fits the autopilot model: heavy on procedural intelligence, light on cross-system judgment, deliverable as a productised service inside a single regulatory frame.
Cross-border advisory is structurally different.
Three properties make it among the strongest cases for the macro shift while simultaneously protecting it from end-to-end delivery by a generalist AI services company.
Multi-jurisdictional licensing topology. Investment migration is governed by overlapping licensing regimes that no AI company can satisfy without licensed humans in the loop. Immigration advice is regulated through the OISC regime in the UK, MARA in Australia, CICC in Canada, and equivalent authorities elsewhere. Caribbean and European citizenship-by-investment programmes require licensed local agents — Malta, St. Kitts, Antigua, Dominica, Grenada, and others each operate their own agent licensing regimes. Tax advisory licensing varies by jurisdiction (Chartered Tax Adviser in the UK, Steuerberater in Germany, expert-comptable in France, CPA-track regulation in the US), with securities and corporate-formation advisory layering additional regulation.
This is not bureaucratic friction. It is regulatory architecture — the structure that determines who has standing to advise. A generalist AI services company can build forward-deployed engineering capacity inside any kind of firm. It cannot, structurally, deliver licensed cross-border advisory without licensed advisors. The licensing is the moat, and it is per-jurisdiction.
Cross-domain compounding. Sequoia’s named verticals tend to operate within a single regulatory domain. Cross-border advisory is the multi-domain compound. A single client’s strategy may engage four to seven jurisdictions simultaneously through residency, asset location, corporate seat, banking infrastructure, citizenship pathways, and estate planning — with each domain’s rules interacting through the others. The reasoning surface is the interaction space between several regulatory systems, with treaty networks, substance requirements, and beneficial-ownership rules cascading across them.
The work is high-stakes in ways most service categories are not. A wrong analysis of UK exit tax interacting with an Ireland-Portugal treaty provision does not produce a 5% efficiency loss. It produces stranded capital, a failed application, an unexpected tax liability that may take years to unwind, or a banking compliance event that propagates across the entire architecture. Recently changed rules amplify the failure surface. Italy’s flat tax for new residents tripled in January 2026 from EUR 100,000 to EUR 300,000. Spain’s Golden Visa closed to new applicants in April 2025. Portugal’s NHR was reformed in 2024. The UK’s non-dom regime was abolished in 2025. The reasoning surface is shifting under any system that cannot track freshness.
Ticket-size economics. Engagement sizes range from $50,000 to $500,000 or more, with senior advisor time as the binding scarcity. The economics that make Sequoia’s 1:6 software-to-services ratio compelling — high services spend relative to the cost of software amplification — apply with full force in this vertical. AI that meaningfully amplifies a senior advisor’s reasoning capacity has compounding returns at this price point. The same infrastructure that produces marginal gains in a $5,000 advisory engagement produces structural advantage in a $250,000 one.
Three structural reasons follow for why generalist AI services companies will struggle to deliver cross-border advisory end-to-end.
First, the knowledge surface is industry-wide, not firm-specific. What an immigration lawyer knows about Portugal’s Digital Nomad Visa is what every other immigration lawyer knows. The infrastructure required to make that knowledge structured, fresh, and reasoning-ready is shared vertical infrastructure — not a firm-by-firm forward-deployed engagement. A consulting team embedded inside one advisory firm for six months cannot build the equivalent of sustained vertical operation across 60+ programmes and multiple regulated domains.
Second, cascade detection requires sustained domain operation. The cross-domain interactions — UK exit tax cascading through Irish holding company treaty access cascading through UAE substance requirements cascading through Swiss banking compliance — emerge from sustained engagement with the regulatory architecture, not from a consulting engagement that ends. They demand structured representations of each regulatory system, maintained continuously. This is operating infrastructure, not consulting output.
Third, regulatory liability sits with the licensed advisor. The forward-deployed engineering team can build powerful tooling around a firm’s workflow. It cannot bear the regulatory liability that attaches to advice. The licensed advisor who signs the analysis carries it — and the value of vertical AI infrastructure is in making that licensed advisor measurably more effective, not in displacing the licensure.
This addresses the natural counter — that an advisory firm could hire one of the new AI services companies to provide the engineering layer while retaining the licensing itself. The combination produces real value, but not the value the cross-domain advisory tier requires. The firm gets accelerated internal tooling — better search across its own materials, faster drafting, more consistent junior output. It does not get the industry-wide vertical infrastructure that cross-domain reasoning depends on, because the engagement model produces firm-specific tooling, not sustained vertical operation across the regulatory architecture of an entire industry.
The asymmetric implication: the more regulated and complex the domain, the higher the ROI of vertical AI infrastructure to a licensed practitioner — and the harder it is for a generalist services company to compete end-to-end. Investment migration sits among the strongest examples of this dynamic. Other cross-border regulated verticals — international wealth structuring, complex private banking, cross-jurisdictional estate planning — share parts of the same profile. Investment migration is distinctive in compounding all three dimensions of licensing topology, cross-domain reasoning, and ticket-size economics simultaneously.
The framework is structural. What does it look like when the abstraction meets a concrete advisory question?
The Cascade Test
Consider a scenario any cross-border advisor will recognise in adjacent forms.
A UK-based founder operates a software consultancy through a UAE free zone company, with an Irish intermediate holding company managing European client contracts. She is evaluating Portuguese residency through the Digital Nomad Visa. Her advisory team — an immigration lawyer, a tax advisor, a corporate structurer — is competent within each domain. She asks her firm’s AI tool a reasonable question: “What do I need to consider for a move to Portugal?”
Now imagine her firm has just completed a six-month engagement with a frontier-lab forward-deployed engineering team. The team is excellent. They have indexed the firm’s jurisdiction guides, programme briefings, treaty summaries, compliance memos. They have built a Claude-powered tool that pulls relevant material in seconds and drafts client-facing summaries in a fluent professional register. The productivity gains are real. The system answers the question competently at the surface level — pulling Portuguese Digital Nomad Visa requirements, Portuguese tax residency rules, UAE free zone regulations, Irish holding company considerations. Each retrieval is accurate. Each is sourced from legitimate jurisdiction materials. The synthesis reads as professional advisory output.
It is also dangerously incomplete.
The move to Portugal establishes Portuguese tax residency. This triggers UK exit tax provisions on unrealised capital gains under TCGA 1992 s.25 — applicable to her equity stake in the software consultancy. The exit-tax calculation interacts with the Ireland-Portugal double tax treaty, which determines withholding rates on dividend distributions from the Irish holding company. The Portuguese residency also shifts the management-and-control analysis for the UAE free zone company — if she is directing operations from Lisbon, the UAE entity’s tax-free status depends on whether it can still demonstrate adequate substance in the Emirates. The Irish intermediate company’s treaty access depends on beneficial ownership analysis, which shifts when the beneficial owner’s residence changes. Her Swiss private bank will re-evaluate her compliance profile upon residency change, potentially triggering banking dynamics that operate independently of client behaviour.
Each fact in this cascade is individually retrievable. The cascade itself is not. It emerges from the interaction between regulatory systems, not from any document within them. The forward-deployed engineering team — competent, well-funded, deeply embedded in the firm’s workflow — has built a system that retrieves correctly within each domain but cannot model the interaction space across them. The synthesis reads as integrated analysis. It is actually a patchwork of single-domain answers that misses every cross-domain dependency that matters.
The hidden dependency most readers — even experienced ones — will not initially see is the Swiss private bank’s compliance re-evaluation. Banking response to a residency change sits in the space between the immigration, tax, and corporate domains. It operates on its own schedule, often independent of any choice the client makes about the move itself. A system that does not model banking as a domain with its own reasoning logic will miss it entirely, and the founder may discover the dependency only after onboarding friction begins at the bank.
The agentic version operates differently. A tax-cascade modeller traces the UK exit-tax exposure through the treaty network. A regulatory-intelligence agent flags that the Ireland-Portugal treaty’s Article 10 dividend provisions interact with the beneficial-ownership analysis the residency change triggers. A substance-monitoring agent surfaces the UAE entity’s recalibrated substance position. A compliance co-pilot tracks the Swiss banking response. None of these is mature in the sense of being deployed everywhere. The architecture is emerging — components proven at scale, integrated multi-domain advisory operation in formation. The Architecture of Advisory Intelligence mapped the six capabilities such systems must deliver. The Confidence Problem mapped the epistemic architecture they must include. Both are observable in production systems today, in early form.
The boundary the example reveals is the one that matters for the bifurcation. A forward-deployed engineering team is excellent at firm-specific workflow integration — building tooling around how a particular advisory firm currently operates, accelerating that firm’s existing process. The structural integration of multi-jurisdictional regulatory reasoning is something different. It is shared vertical infrastructure — not consulting output, not firm-internal tooling, not a project that ends when the engineering team rotates off. It is sustained vertical operation across the regulatory architecture of an entire industry.
Both kinds of AI capability are valuable. They are not the same kind of investment, and they will not be delivered by the same kind of firm.
The Bifurcation
Three categories of advisory firm will exist five years from now. The distinctions are emerging, not established. The strategic question for principals is no longer “should we adopt AI?” — that question is settled, and any firm not adopting AI in some form has already chosen its position by default. The question now is which of the three categories the firm will be by 2030.
Category A — Generalist AI services companies. Anthropic’s enterprise AI services company. OpenAI’s Development Company. The wave of similar ventures that will follow, backed by frontier labs and private equity. Well-capitalised, lab-aligned, deep on engineering. Their economics favour cross-vertical scale over single-vertical depth, and their delivery model is firm-by-firm forward-deployment across portfolio companies. They will win the regulated services market where work is closer to operational automation than cross-domain advisory: claims processing, structured compliance, document automation, single-domain analysis at volume. In cross-border advisory, they will excel at the productised edge — visa application drafting at scale, programme Q&A, document collection — while struggling to reach the cross-domain tier that depends on sustained vertical operation.
Category B — AI-native advisory firms. The new category. Licensed professionals operating with vertical AI infrastructure that delivers what cross-border advisory actually requires — structured jurisdiction knowledge with provenance and freshness, domain-specialised reasoning, cascade detection, traceable output, structured uncertainty signals. Crucially, consuming Category A services does not, by itself, position a firm in Category B. Category A delivers firm-specific tooling around existing workflows. Category B requires vertical infrastructure — built once, maintained continuously, consumed across many firms in the industry. These are different infrastructure layers, not different levels of the same one. The competitive shape is different from Category A. Smaller engineering footprint. Deeper vertical depth. Regulatory protection that compounds over time. Higher gross margin per engagement. Scale through agentic capacity rather than headcount. A three-person AI-native firm can operate at the effective capacity of a thirty-person traditional firm because the infrastructure is doing the structural reasoning work — and the principal’s time is reserved for judgment and client relationship work that cannot be infrastructure-served. The category is emerging, not established. Early instances exist; it is not yet recognised at industry scale.
Category C — Traditional advisory firms with general-purpose AI tools. Most firms in the investment migration and cross-border advisory space currently sit here, often without recognising the categorisation. They have deployed ChatGPT Enterprise, Claude for Business, Microsoft Copilot, or built internal chatbots over their document repositories. The productivity gains are real and observable. They are also strategically thin. Two sub-paths emerge. Firms that adopt purpose-built vertical infrastructure — building internally, partnering, or licensing — move toward Category B. Firms that continue with general-purpose tools face progressive margin compression: from Category A on operational tasks, and from Category B on the high-margin cross-domain advisory work that defines the practice’s premium positioning.
The strategic implication is not “abandon what you have built.” It is “decide which category you want to be by 2030 and start positioning now.” The infrastructure decisions made between now and 2027 will determine which category a firm occupies five years out. The window for being early to Category B is open. It will narrow as the category formalises and the AI-native firms gain regulatory comfort, client trust, and operational track record.
Three signals over the next 24-36 months will indicate the pace of category formation.
The first is regulatory enforcement. The EU AI Act’s high-risk obligations take effect August 2, 2026 — three months from this writing. Advisory AI operating in migration processing and essential private services contexts is likely classified as high-risk under Annex III. The Architecture of Advisory Intelligence mapped the requirements in detail. The point for this analysis: a Category C firm running a general-purpose chatbot can probably produce paper compliance — audit logs, disclaimers, human-oversight documentation — but has no structural resilience against the underlying obligations. When a regulator or client asks why a specific recommendation was produced, the system cannot answer from architecture. A Category B firm absorbs compliance as a property of the architecture itself — provenance, traceability, freshness, uncertainty signalling are how the system operates, not features added to satisfy a checklist. The distinction between paper compliance and structural resilience will become a competitive boundary, not just a regulatory one.
The second signal is capital deployment from the May 4 ventures. Anthropic’s enterprise AI services company and OpenAI’s Development Company will deploy through 2026-2027 across mid-market regulated industries. The verticals they target first, the depth they reach, and the verticals they cannot penetrate will reveal where Category A’s competitive boundary actually lies. The expectation here is that they will scale operational AI services aggressively, while cross-border advisory at depth will remain unreached by their model.
The third signal is the rate at which licensed practitioners build or partner for vertical infrastructure access. Category B is forming through two pathways. Firms that build their own vertical infrastructure — capital-intensive, multi-year, justified only at sufficient volume and specialisation. Firms that consume vertical infrastructure provided by category-specific platforms via API, MCP, or embedded workflows — faster, vendor-dependent, the dominant pathway for firms below the build threshold. The relative growth of these pathways will shape the structure of the vertical infrastructure layer itself.
The bifurcation is not a prediction of certainty. It is a structural reading of a shift already in motion. Capital deployment is real. Architecture choices are being made now. The firms that recognise the category question early have time to position for it. The firms that treat it as a future concern will discover that the position is being chosen for them by the choices they decline to make.
The Infrastructure Layer
If Category B is the strategic destination for licensed cross-border advisory firms, the infrastructure question follows directly: where does the vertical infrastructure come from?
The frontier-lab joint ventures will not build it at depth. Their model — forward-deployed engineering across portfolio companies — is firm-by-firm and cross-vertical. Building the kind of jurisdiction infrastructure investment migration requires — 60+ active programmes, freshness propagation across regulatory change, treaty-network reasoning, cascade modelling across tax-immigration-corporate-banking interactions — is sustained vertical operation. It fits Bek’s autopilot framing precisely: heavy on domain intelligence, with judgment delivered by licensed professionals.
The missing piece in the current advisory landscape is jurisdiction intelligence as shared infrastructure — structured, provenance-tracked, continuously maintained data and reasoning capability that any advisory surface can consume. Not a chatbot over a firm’s document store. Not a copilot embedded in a partner’s workflow. Infrastructure: programmes, treaty networks, regulatory changes, eligibility logic, fee schedules, cascade reasoning across domains — built once, maintained as continuous operation, consumed by many advisory firms through API, MCP, or embedded workflows.
This is the layer Category B firms will operate against. Some will build their own — justified at sufficient volume and specialisation, capital-intensive, multi-year. Most will consume infrastructure provided by category-specific platforms, the way traditional advisory firms today consume professional data services, regulatory databases, and compliance platforms — only with deeper reasoning capability, structured rather than text-only delivery, and freshness propagated rather than periodically refreshed.
The May 4 announcements signal that capital is now flowing to whoever can deliver work. The same logic applies one layer down. Capital will follow infrastructure providers serving licensed advisory firms in regulated, complex verticals — because that is where the outcome-delivery layer is going to be built at depth. The frontier labs will deploy their engineering capacity broadly. The vertical infrastructure layer will be built by firms whose entire operation is sustained engagement with a single industry’s regulatory architecture.
A note on perspective: Sovara — the platform behind this publication — is building vertical AI infrastructure for cross-border advisory. The structural argument in this essay reflects what that work has made visible, not the other way around. The category is forming; other platforms will enter as the macro shift accelerates. The reason I write the Briefing is that I think the structural picture is more useful to advisory firm principals than any single provider’s pitch — including ours
The future of cross-border advisory is not advisors versus AI. It is AI-native advisory firms versus everyone else.
The Framework, Condensed
The Bifurcation
The macro shift to services-as-software — operationalised by Anthropic’s and OpenAI’s May 4 joint ventures, framed by Sequoia’s 1:6 thesis — combined with the licensing topology of cross-border advisory, produces a three-category map of the industry five years out.
Category A — Generalist AI services companies.
Profile: Frontier-lab-backed, PE-funded, cross-vertical engineering depth.
Wins: Productised regulated work, single-domain at volume, operational automation.
Loses: Cross-domain advisory at depth, which requires sustained vertical operation they will not build.
Category B — AI-native advisory firms.
Profile: Licensed practitioners paired with vertical AI infrastructure and agentic delivery.
Wins: Cross-domain advisory at depth, regulatory-protected work, scale through infrastructure rather than headcount.
Loses: Productised commodity work at lower price points than Category A.
Category C — Traditional advisory firms with general-purpose AI.
Profile: Existing advisory firms running ChatGPT, Claude for Business, internal chatbots.
Wins: Existing client relationships, brand, transitional period through 2027.
Loses: Progressive margin compression from above (Category A on operational work) and below (Category B on advisory work).
Three application questions for advisory firm principals.
A managing partner of a fifty-person immigration practice can apply the bifurcation directly. Current state: senior advisors do hours of jurisdiction research per matter; junior output is inconsistent; programme data goes stale between engagements. The decision is not which AI tool to deploy. It is which infrastructure layer to operate against. Building internal vertical infrastructure is capital-intensive and multi-year, justified only at substantial volume. Partnering with or licensing a vertical infrastructure provider is faster, vendor-dependent, and becoming the dominant pathway for firms below the build threshold. Deploying a general-purpose chatbot over existing materials is the Category C path, with margin compression following on the schedule the wider macro shift is setting.
A multi-family office serving cross-border clients can apply the same framework. Adding AI-native advisory capability is a build / partner / acquire question, in roughly that order of capital intensity. Scale and time-to-capability strongly favour partnership. Build is justified only when the office’s volume and specialisation profile justify dedicated infrastructure ownership.
A founding team building a cross-border advisory firm in 2026 has the unusual option to operate as Category B from inception. A smaller licensed footprint paired with vertical infrastructure consumption produces different unit economics from a traditional firm built on headcount expansion. The May 4 announcements suggest capital is now flowing to firms positioned to consume rather than rebuild this infrastructure.
The framework’s logic is structural. Macro AI capital deployment shifts value capture toward outcome delivery. Cross-border advisory’s licensing topology and cross-domain compounding mean the depth required for that delivery cannot be supplied by generalist services companies. The advisory firms that pair regulatory licensing with purpose-built vertical infrastructure win the cross-domain advisory tier. The firms that treat AI as a productivity tool are commoditised from both directions. The strategic question is no longer about AI adoption. It is about category position. The window for early positioning is open. It will not stay open indefinitely.