For several years, the artificial-intelligence discussion inside professional services has focused primarily on what individual firms are building. Deloitte, PwC, EY, KPMG and their competitors have invested heavily in internal AI environments, proprietary tools, data platforms, technology alliances and workflow automation. The assumption underneath much of the debate has been familiar: each firm chooses its technology partners, invests in its own capabilities and gradually embeds AI into the way its professionals work.
A different institutional model is beginning to appear. Anthropic, Blackstone and Hellman & Friedman have created Ode with Anthropic, a standalone enterprise AI services company backed by a wider investor consortium that includes Goldman Sachs, General Atlantic, Apollo Global Management, GIC, Leonard Green & Partners and Sequoia Capital. OpenAI has moved in a similar direction through the OpenAI Deployment Company, a majority OpenAI-owned business launched with more than $4 billion of initial investment from nineteen investment firms, consultancies and systems integrators led by TPG. The participants include organisations such as Bain & Company, Capgemini and McKinsey & Company. These are not small implementation experiments. Capital providers, frontier-model companies, engineers and professional services firms are beginning to assemble around a new institutional layer. (Anthropic, Blackstone, and Hellman & Friedman Introduce Ode with Anthropic, an Enterprise AI Services Firm; OpenAI Launches the OpenAI Deployment Company to Help Businesses Build Around Intelligence)
Professional services has quickly become part of the experiment. Baker Tilly is working with Ode across advisory, tax and assurance. Citrin Cooperman has engaged the company across accounting, tax and advisory workflows. Multiplier Holdings takes the architecture further by acquiring professional services firms, building proprietary technology and AI capabilities for each business, while creating a common platform that can be used across the wider portfolio. Something unusual is happening. Powerful intelligence is becoming increasingly accessible, while the scarcer capability may become the ability to turn that intelligence into reliable professional production repeatedly. The real strategic question is therefore not simply who owns the AI. It is where the learning from using it compounds, and who ultimately controls the infrastructure built from that learning. (Baker Tilly and Ode with Anthropic Launch Initiative to Advance AI-Enabled Client Service; Citrin Cooperman Advisors Accelerates AI Transformation with Ode with Anthropic; Multiplier Holdings: About)
The Bottleneck Has Moved
Access to powerful AI models is becoming progressively less distinctive. Professional services firms can procure models from Anthropic, OpenAI, Google and others. They can provide secure internal access, deploy copilots and experiment with agents across research, documentation, analysis and workflow execution. Model quality still matters, and the frontier will continue moving quickly, but access to frontier intelligence increasingly tells us little about whether an institution can translate that intelligence into durable economic value.
The harder problem sits underneath the model. AI has to interact with existing data, methodologies, workflow platforms, security environments, regulatory controls and professional standards. A tax process that appears straightforward from the outside may contain dozens of systems, exceptions, approval steps and jurisdiction-specific requirements. Audit adds evidence, traceability, independence and accountability. Advisory work often contains much less standardisation. Turning intelligence into reliable professional production therefore requires considerably more than providing professionals with better tools. It requires redesigning how the work itself is performed.
That helps explain why Ode and the OpenAI Deployment Company are significant. Both frontier-model ecosystems have independently identified the gap between model capability and institutional deployment as an opportunity large enough to justify creating dedicated organisations around it. Their engineers work directly inside organisations, connect models to data and systems, redesign workflows and build the controls required to make AI operational. If that work remains bespoke, these businesses may ultimately resemble highly sophisticated implementation firms. If the learning accumulates across deployments, however, something more consequential begins to emerge. (Anthropic Partners with Blackstone, Hellman & Friedman, and Goldman Sachs to Launch Enterprise AI Services Firm; OpenAI Launches the OpenAI Deployment Company to Help Businesses Build Around Intelligence)
Three Different Bets on the Same Problem
The first model is already visible inside the largest professional services firms. Deloitte, PwC, EY and KPMG combine professional expertise, client relationships, internal technology capability, global delivery infrastructure and major technology alliances inside institutions they still fundamentally control. Their scale matters. A Big Four firm can invest in engineering, proprietary methodologies, data environments and workflow platforms that would be difficult for a smaller partnership to justify independently. More importantly, the lessons from redesigning one part of the organisation can remain available to other parts of the same institution.
Ode represents a different answer. Specialist AI deployment capability sits outside the professional services firm, while model providers, investors and professional firms participate around it. The ownership relationships make the structure particularly interesting. Hellman & Friedman backs Baker Tilly and Blackstone backs Citrin Cooperman, while both investors are also founding backers of Ode. In other words, investors are not only funding professional services platforms and separately funding AI companies. They are beginning to participate in an ecosystem in which the firms they own can consume capability from an infrastructure provider they also back. (Baker Tilly | Hellman & Friedman; Citrin Cooperman, a Leading Professional Services Firm, to Receive Significant Investment as Blackstone Acquires Stake from New Mountain Capital; Anthropic, Blackstone, and Hellman & Friedman Introduce Ode with Anthropic, an Enterprise AI Services Firm)
Multiplier takes the idea further by putting the professional firms and technology capability inside the same ownership system. The company acquires specialist accounting and tax businesses, retains their brands and leadership, and deploys technology teams directly into them to build proprietary capabilities around the needs of each individual firm. At the same time, those teams can draw on a common platform through which infrastructure, selected capabilities and learning can increasingly be reused across the wider portfolio. In August 2026, Multiplier announced a $35 million Series B after acquiring eight firms, with another four reportedly under signed term sheets. The objective is therefore not to standardise every acquired firm onto one common technology solution. It is to combine firm-specific proprietary capability with a platform that allows useful components and learning to compound across businesses. The three models are different institutional bets on the same problem: should the learning required to make AI productive remain inside the professional firm, sit inside a specialist provider, or be owned alongside the firms themselves? (Multiplier Holdings: Human Heart // AI Muscle; Multiplier Raises $35 Million Series B to Build a New Model for Professional Services)
The Real Asset Is the Learning Curve
The decisive question is not whether these organisations can build useful AI tools. Most serious professional services firms will be able to do that. The more interesting question is whether solving one problem makes the next one materially easier. A capability layer becomes strategically valuable when workflow knowledge, implementation patterns, control structures, integration methods and professional logic accumulate instead of being rediscovered every time another workflow is redesigned. The real asset is therefore not necessarily the software produced at the end of an implementation. It is the learning curve underneath it.
Multiplier makes this mechanism unusually tangible. Imagine that the first acquired firm requires substantial work to redesign a recurring accounting process. Engineers have to understand the workflow, identify exceptions, connect source systems, develop controls and decide where professional review remains necessary. The resulting technology may remain proprietary to that firm because its clients, systems and workflows are different. But parts of what was learned should transfer. The second firm starts with better infrastructure, reusable components and a team that has already encountered similar problems. The third adds another variation. By the tenth acquisition, the key question is not whether every firm is running identical technology. It is how much faster and better Multiplier can build the right proprietary capability for that firm because of everything learned across the previous nine businesses. If each implementation starts further along the learning curve, the platform is compounding.
The same possibility makes Ode strategically interesting. After working deeply inside one accounting firm, one tax practice or one advisory business, it knows more about deploying AI into that type of institution. After twenty implementations, it could possess patterns that no individual client sees because each firm only experiences its own transformation. That would create something close to cross-firm operating intelligence: accumulated knowledge about where professional workflows are similar, where they differ, which forms of automation work, which controls fail and how production models need to change. The strategic risk for a professional services firm is therefore not that an outside provider knows how to implement AI. It is that, after enough implementations, the outside provider may understand how to redesign parts of professional work better than any one of its clients does.
That changes the economics of ownership. If accumulated deployment knowledge reduces the cost and time required for future implementations, whoever owns that learning gains leverage. The external capability provider becomes more valuable with every deployment. Its bargaining position can strengthen. Professional firms may become increasingly dependent on capability they do not control. Differentiation can begin migrating from the institution serving the client toward the infrastructure that enables the institution to serve the client. This is closely related to what I described in The Platform Gravity Problem: Why Control Over Shared Systems Increasingly Shapes Power Inside Professional Services Networks: operational dependency can redistribute economic and institutional power without any formal transfer of legal ownership.
There is a credible alternative outcome. Deployment itself may become much easier. Frontier-model providers and enterprise software companies are rapidly building more capability into their products, while the large professional firms are accumulating substantial internal experience. Methodologies, risk appetites, regulatory requirements, client portfolios and data environments also differ significantly between institutions. If those differences dominate the similarities, cross-firm learning may compound much less than the platform thesis assumes. The AI capability layer could therefore prove to be a temporary response to a difficult stage of technological development rather than a permanent institutional layer. The next several years will begin to tell us which explanation is closer to reality.
But Professional Services Is Not Just Expertise
There is another limit to how far the capability layer can expand. Professional services firms do not simply manufacture outputs. They combine several different assets that are easy to collapse into one idea but behave very differently economically. I have argued in When Expertise Becomes Metered Infrastructure: How AI Is Changing the Economics of Professional Services that AI is beginning to change the economic form of expertise. Technical knowledge, recognised patterns, research, methodology and repeatable professional workflows can increasingly be embedded into models and systems. Expertise that once had to be recreated through professional labour can therefore become reusable infrastructure.
Judgment is different. Expertise helps a professional understand what the rules, evidence and recognised patterns say. Judgment becomes valuable when the evidence is incomplete, several answers are technically plausible, the circumstances are unusual or the consequences of being wrong are material. In Manufacturing Judgment: How Professional Services Build Their Most Valuable Asset, I argued that professional firms have historically developed this capability through repeated exposure to real client situations. AI can support judgment by giving professionals better information and alternative interpretations. But it can also remove some of the work through which younger professionals traditionally learned how to exercise judgment in the first place.
Then there is trust. A technically strong answer and excellent professional judgment still have limited economic value unless somebody is willing to rely on them. In Trust Capital: Why Professional Services Manufacture Judgment but Accumulate Trust, I describe this as the institution’s accumulated capacity to make clients, regulators, investors and other stakeholders willing to rely on its judgment. These are not clean boundaries between human and machine. AI will increasingly contribute to expertise and judgment, and technology itself may become trusted in highly standardised settings. The distinction is economic and institutional. Expertise is considerably more codifiable and reusable than judgment, while trust remains closely connected to accountability, relationships and institutional reputation.
Where Does the Firm End?
That distinction creates a much more interesting future than either “AI replaces professionals” or “AI simply makes professionals more productive.” Parts of professional capability can migrate into shared infrastructure without the entire professional institution moving with them. Expertise and workflow execution may become more technology-enabled and reusable. Judgment may become increasingly AI-supported but remain associated with accountable professionals. Trust may remain closest to the local firm, partner or regulated institution that the client ultimately relies upon.
Multiplier makes this possible architecture unusually visible. Proprietary technology and AI capabilities can be built around each acquired firm while a common platform underneath them provides infrastructure, reusable components and accumulated learning across the portfolio. The production systems of those businesses do not need to become identical for the platform to create leverage. The visible institutions can retain their brands, leaders, clients and differentiated ways of working while benefiting from technology capability that becomes progressively stronger because it has been developed repeatedly elsewhere.
That may be deliberate rather than temporary. Trust is expensive to rebuild and easy to damage. In The Transferability Problem: Why Professional-Services Roll-Ups May Be Discovering That Institutional Trust Is Harder to Scale Than Revenue, I argued that acquiring firms and revenue is much easier than transferring institutional trust from the acquired businesses to the platform above them. AI potentially widens that separation. Technology infrastructure and learning can compound across a portfolio much faster than trust can migrate from the individual firms within it.
The result could be a professional services institution with much more distinct layers than the traditional partnership ever required. A client relationship and trust layer remains close to the market. Professional judgment remains connected to individuals and regulated accountability. Firm-specific technology contains proprietary workflows and capabilities adapted to the individual business. A common platform underneath those firms provides reusable infrastructure and allows learning to accumulate across the wider institution. Frontier models and cloud infrastructure sit underneath it again. Different organisations may own each layer. That sits directly inside the argument of my Professional Services Transformation Theory: the economic architecture of professional services is changing faster than the organisational and governance structures used to manage it.
This is also why the debate about AI productivity is ultimately incomplete. Lower professional effort does not automatically create higher economic productivity if prices fall, technology costs rise, traditional leverage weakens or expensive infrastructure moves elsewhere in the organisation. I have explored those economics separately in The AI Cost Stack: Why Professional Services Firms Are Looking for AI Savings in the Wrong Place and The Professional Services AI Paradox: How the AI Platform Economy Is Colliding with the Partnership Model. The larger question here is structural: when production capability increasingly depends on reusable technology infrastructure, where does the professional firm itself begin and end?
What Happens Next?
The most useful evidence will not come from another AI announcement. It will come from what happens inside Baker Tilly, Citrin Cooperman, Multiplier’s portfolio and the companies working with the new deployment organisations over the next several years. If embedded engineering teams generate useful but isolated workflow improvements, the institutional implications remain limited. If the same teams begin producing reusable capability and accumulating learning that changes the economics of multiple firms, a new layer of the industry is genuinely forming.
The first test is therefore compounding. Does solving one professional workflow materially reduce the difficulty of solving the next? Does Multiplier become materially faster and better at building proprietary capability for its tenth acquisition because of what it learned from the first nine? Does Ode become faster and better at transforming accounting and tax firms because it has already worked inside others? Can the Big Four keep equivalent learning inside their own organisations, or will parts of the capability become more efficiently produced by specialised external providers? The answers can eventually be observed rather than merely predicted.
The second test concerns the institutional boundary. Where does judgment remain? Where does accountability sit? Does trust attach primarily to the local firm, the larger platform or eventually even the technology infrastructure itself? And where does the learning from every deployment accumulate? If professional firms retain client relationships, proprietary capabilities and accountability while another institutional layer increasingly captures reusable infrastructure and cross-firm learning, the balance of power inside the industry begins to change in ways that are much more consequential than another productivity improvement.
Closing Thoughts
Ode, the OpenAI Deployment Company and Multiplier are interesting not because any of them has already found the winning model. They are interesting because they reveal the emergence of a potential new institutional asset. Powerful intelligence is becoming increasingly available. The scarce capability may become the accumulated knowledge required to turn that intelligence into reliable, governed and repeatable professional production.
For professional services firms, the risk is therefore not simply that AI replaces professional work. It is that the learning required to redesign that work begins compounding somewhere else. Expertise may increasingly become infrastructure. Judgment and trust may remain closer to the profession. But the layer connecting the two could become one of the most valuable institutional assets in professional services.
The question is who will own it.
What This Means for Boards
Boards should resist treating AI strategy primarily as a technology-investment discussion. Access to frontier models will not be the durable differentiator. The more important question is whether the firm is developing an institutional capability to redesign professional work repeatedly, and whether the knowledge created through that process compounds somewhere that strengthens the firm rather than increasing its dependency on someone else’s capability.
That requires deliberate choices about boundaries. Which parts of the firm’s expertise should become reusable infrastructure? Which workflows should be rebuilt rather than incrementally automated? Which capabilities should remain proprietary to individual businesses? Which infrastructure should be common? Which capabilities are sufficiently strategic that their learning needs to remain inside the firm or the wider institution? Which can be sourced externally? If an external provider works deeply inside the firm’s workflows, what does that provider know after the tenth implementation that it did not know after the first, and who ultimately owns the economic value of that learning?
Boards also need to protect what does not migrate as easily. How will future professionals develop judgment when routine work disappears? Where does accountability sit when systems perform increasing amounts of analysis? What preserves institutional trust when production becomes more distributed? The technology questions are becoming easier. The strategic questions are increasingly about institutional boundaries, economic ownership and which capabilities the firm must still be able to call its own.
Continue Exploring
The Bigger Picture
The Professional Services Transformation Theory
Why changing economics, production models, ownership structures and governance are forcing professional services firms to rethink how the institution itself is designed.
AI, Expertise, Judgment and Trust
When Expertise Becomes Metered Infrastructure: How AI Is Changing the Economics of Professional Services
Why expertise itself is beginning to change economic form as more of it becomes embedded inside reusable technology.
Manufacturing Judgment: How Professional Services Build Their Most Valuable Asset
Why client work has historically produced both today’s output and tomorrow’s professional judgment.
Trust Capital: Why Professional Services Manufacture Judgment but Accumulate Trust
Why judgment only creates economic value when clients and other stakeholders are willing to rely on it.
AI Economics and Institutional Power
The AI Cost Stack: Why Professional Services Firms Are Looking for AI Savings in the Wrong Place
Why lower professional effort does not necessarily mean lower total economic cost.
The Professional Services AI Paradox: How the AI Platform Economy Is Colliding with the Partnership Model
Why reusable AI capability increasingly conflicts with partnership economics, incentives and governance.
The Platform Gravity Problem: Why Control Over Shared Systems Increasingly Shapes Power Inside Professional Services Networks
Why operational dependency can redistribute power even when legal ownership does not change.
The Transferability Problem: Why Professional-Services Roll-Ups May Be Discovering That Institutional Trust Is Harder to Scale Than Revenue
Why infrastructure and scale can centralise considerably faster than institutional trust.
Case Study 43: Citrin Cooperman and the Accounting Firm That Became Transferable
Why institutional capability and transferability matter as private capital moves through successive ownership cycles.
Sources
Company Publications
Baker Tilly and Ode with Anthropic Launch Initiative to Advance AI-Enabled Client Service
Citrin Cooperman Advisors Accelerates AI Transformation with Ode with Anthropic
OpenAI Launches the OpenAI Deployment Company to Help Businesses Build Around Intelligence
Multiplier Holdings: Human Heart // AI Muscle
Ode with Anthropic Acquires Casper Studios
Corporate and Ownership Sources
Baker Tilly | Hellman & Friedman
Independent Reporting
Multiplier Raises $35 Million Series B to Build a New Model for Professional Services