Accounting firms have moved quickly from AI experimentation to widespread adoption, and the result is often a fragmented AI stack. Firms adopted AI one use case at a time: a tool for general productivity, another for professional research, another for audit and document work, and increasingly, AI embedded in individual applications and workflows.
That speed made sense. Firms wanted immediate efficiency gains, and employees adopted the tools that helped them work faster. But as these AI features accumulated, each became its own technology decision, with its own cost, integration requirements, data considerations, workflow implications, and governance questions.
The result is that AI adoption has outpaced firm-wide strategy and control. Thomson Reuters’ 2026 Future of Professionals research found that 81% of tax and audit professionals now use AI regularly in their day-to-day work, while 35% say they use AI tools their firm has not authorized. AI is already in use across the organization, but not through a coordinated, centrally governed approach — and that gap is business risk.
This is why buying the latest AI feature is unlikely to create a lasting competitive advantage. Everyone can buy features. Any firm can license a general-purpose model, add AI-powered research, automate document review, or turn on a new copilot. Those capabilities create real short-term productivity gains, but they are quickly becoming table stakes.
The competitive advantage comes from what a firm builds underneath those features. A true AI foundation connects intelligence to the firm’s own data, workflows, permissions, institutional knowledge, and governance. It lets AI capabilities scale across the organization instead of being deployed as isolated point solutions, and it creates a common base on which new capabilities can be added, governed, measured, and improved over time.
The firms that pull ahead will not be the ones with the most AI tools. They will be the ones that build the strongest foundation underneath them — and for accounting firms, that foundation is Firm AI: intelligence mapped to how your firm actually operates.
Buying features worked. It won’t give you an edge anymore.
An AI feature is easy to justify. It solves one visible problem, the pilot is cheap, and the demo lands. So the firm buys it, then buys the next one the same way. For a while, that was the right instinct — early efficiency gains were real, and being fast mattered.
What nobody adds up is the hidden tax on that approach. Each feature needs its own connection into the firm’s systems. Each one reopens the same governance questions about ownership, audit trails, and ethical walls. Each one has to be monitored and re-integrated whenever an underlying system changes. Buy ten features and you have paid that bill ten times over, because nothing built for one is reused by the next. The result is a drawer full of tools and no compounding return on any of them.
And the edge those features once gave you is disappearing. When every firm can license the same off-the-shelf AI, the tool itself stops being a differentiator. General-purpose AI is becoming table stakes. What separates firms now is not whether they have AI, but whether they have built something underneath it that competitors cannot buy off a shelf.
A foundation is what lets AI scale
The barrier to AI at scale is not access to AI. It is everything underneath it — and the evidence is clear.
Both Deloitte and KPMG have independently measured the same wall. Deloitte’s 2026 Tech Trends research finds only 11% of organizations have successfully deployed AI agents in production. The gap is not a technology problem; it is organizations trying to layer AI onto processes designed for humans instead of redesigning workflows and the operating model around AI. Deloitte’s research on agentic transformation makes the missing foundation even more explicit. Among leaders trying to scale agentic AI, 72% name the lack of a unified, accessible data foundation as a barrier, 70% cite the ability to trust and govern agents, and 67% point to the cost and complexity of integration. The constraints on scaling AI are foundational.
KPMG’s Global AI Pulse research lands on the identical figure from the value angle: just 11% have reached the stage of deploying and scaling agents in ways that produce enterprise-wide outcomes, even as organizations plan to spend an average of $186 million on AI in the year ahead. As its Global Head of AI and Data Labs put it, value does not come from launching isolated agents, and adding more AI capabilities does not automatically translate into enterprise value. The organizations that scale have the underlying data, integration, governance, and operating model to connect those capabilities to how the business actually works.
Gartner adds the cost dimension. It predicts that more than 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value, and inadequate risk controls.
When every AI capability is implemented independently, every deployment carries its own integration work, data connections, governance model, permissions, change management, and measurement. The cost and complexity accumulate alongside the number of tools. A foundation reverses that math. The first AI workflow requires investment in the underlying architecture — connecting the firm’s data, systems, workflows, permissions, institutional knowledge, and governance. But once those capabilities exist, the next workflow does not rebuild them. Each additional use case builds on what is already there, so your tools get cheaper as you add them rather than more expensive.
That is the difference between a feature and a foundation. A feature keeps what it learns to itself. A foundation shares it with everything built afterward, and knows what a conflicts clearance is, what an engagement is, and what an independence obligation means at your firm. This is what Firm AI means in practice: not generic intelligence pointed at your data, but intelligence mapped to your firm’s operating model — its roles, its controls, its definitions, its way of working.
A foundation is what protects the firm
Scaling AI without a foundation does not just cost more. It creates exposure. When staff run client work through ungoverned tools, there is no independence screen and no record of what was done — and the Thomson Reuters figure above shows this is already happening in more than a third of firms.
A foundation puts governance in one place instead of reinventing it per tool. Accountability, permissions that mirror the firm’s ethical walls, and a complete audit trail live in the layer itself and apply the same way to every AI action. The answer to ungoverned AI is not to ban it — that only pushes people toward unapproved tools — but to route firm work through a layer where the controls the firm already applies to client work also apply to the AI acting on it. As audit-quality, independence, and governance requirements tighten, that difference moves from good practice to a requirement.
A foundation is what lets you compete
For accounting firms, choosing features over a foundation is quietly becoming a competitive fault line. Three pressures are arriving together.
Talent. Firms cannot hire their way to capacity, so AI has to scale across the firm rather than stay isolated in individual tools.
Consolidation. As private equity and consolidation reshape the profession, operational infrastructure increasingly matters in diligence. Disconnected tools look like cost. A governed, scalable AI foundation looks like an asset.
Scrutiny. As audit-quality, independence, and governance requirements tighten, firms need AI that is controlled, auditable, and embedded in the right workflows — not tools operating outside the firm’s oversight.
A foundation answers all three at once, because capacity, governance, and auditability live in the layer itself. Features answer none of them for long, because each one solves its own slice and leaves the firm-level problem where it was. When general-purpose AI is available to everyone, the firm that has built intelligence around its own operating model — one that makes better decisions, faster — is the one that pulls ahead of competitors still buying features one at a time.
The Intapp view
This is the layer we build. Intapp Celeste is the governed foundation for Firm AI in accounting. It connects the systems a firm already runs so that intelligence maps to the firm’s own operating model rather than sitting beside it in a separate tool. Client acceptance, independence and conflicts clearance, and relationship intelligence run as governed workflows on that shared layer, with accountability and the audit trail built in rather than bolted on. Every workflow stands on the same foundation — so the firm scales AI meaningfully, competes on more than table stakes, and stays protected from business risk.
See how one governed foundation runs firm-wide workflows end to end.