Professional services firms — law firms, consulting firms, accounting practices, financial advisory businesses, marketing agencies, engineering consultancies, and the full range of businesses that sell expertise and time — have always had to make decisions about how to handle costs incurred in the service of client work. Copying costs, research database subscriptions, expert witness fees, travel expenses, and software tools used in client engagements have generated billing policy questions for decades. The answers firms have developed are embedded in their engagement letters, their billing practices, and their client relationship cultures.
Consumption-based AI pricing introduces a new version of this question with characteristics that prior expense categories do not share. AI consumption costs are incurred at the level of individual queries and interactions — not as flat subscriptions or discrete invoices, but as micro-costs that accumulate continuously across every use of AI tools in client work. They vary by model tier, context length, and usage pattern in ways that are difficult to predict and difficult to attribute precisely to specific client matters without deliberate tracking infrastructure. And they touch a data governance dimension that copying costs and travel expenses do not: the AI tools generating the consumption costs are also processing client confidential information, which creates obligations around how that usage is tracked, attributed, and disclosed.
Understanding how consumption-based AI pricing interacts with client billing requires working through the three billing models available to professional services firms, the tracking infrastructure each model requires, and the client communication approach that treats AI cost handling with the transparency that client relationships and professional conduct standards expect.
The Three AI Consumption Billing Models
Pass-Through at Cost: Transparency With Infrastructure Requirements
The pass-through model treats AI consumption costs as a direct expense of the client engagement — the firm tracks AI usage attributable to each client matter and bills those costs to the client at the firm’s actual cost, with no markup. This is the most transparent billing model and, in many professional contexts, the most defensible from an ethical standpoint: the firm is billing the client for the actual cost of a tool used in their matter, the same way it would bill for a research database search or a specialized software license purchased specifically for the engagement.
The pass-through model’s transparency is also its primary challenge: it requires the tracking infrastructure to know, with reasonable accuracy, how much AI consumption was attributable to each client matter. In a firm where attorneys, consultants, or advisers use AI tools throughout their workday across multiple client matters simultaneously, attributing AI consumption costs to specific matters requires either matter-specific AI sessions that track usage by client, or a post-hoc allocation methodology that estimates how AI usage was distributed across the matters worked on during a billing period. The former is technically more accurate but requires workflow discipline from professionals who are accustomed to flowing between matters fluidly. The latter is operationally simpler but introduces estimation that may be difficult to defend if a client scrutinizes the AI cost allocation.
For law firms, the pass-through model raises a specific ethical question: whether AI consumption costs are billable expenses at all, or whether they are overhead costs of doing business that should not be passed to clients separately. Bar association ethics guidance on billing for technology costs is not uniform across jurisdictions, and the question of whether AI usage fees constitute billable disbursements — similar to LexisNexis or Westlaw research costs — or overhead expenses — similar to the firm’s internet connection or word processing software — is actively being addressed by bar association ethics committees in multiple states. Law firms implementing pass-through billing for AI consumption costs should review current ethics guidance in their jurisdiction before adopting this model.
For non-legal professional services firms, the pass-through model is generally permissible as a matter of contract law — if the engagement letter specifies that AI usage costs incurred in the engagement will be billed at cost as a separate line item, and the client accepts those terms, the billing is contractually authorized. The practical challenge is ensuring that the engagement letter language is specific enough to create clear client expectations about what AI costs will be billed and at what rates, so that AI usage charges on invoices do not create surprise or dispute.
Markup as a Profit Center: Revenue Opportunity With Disclosure Obligations
The markup model treats AI consumption costs as a billable service with a margin built in — the firm tracks AI usage attributable to client matters, charges the client a rate per AI interaction or per AI usage period that exceeds the firm’s actual cost, and retains the margin as revenue. This is how professional services firms have historically handled other technology-enabled services: research database searches billed at rates that include a margin above the per-search cost, document review platforms billed at rates above the actual software cost, and similar models in which the technology service is bundled with the professional judgment that makes it useful and priced accordingly.
The markup model generates more revenue per client engagement than pass-through, and it compensates the firm for the investment it has made in selecting, configuring, and governing the AI tools that clients benefit from — investment that the raw consumption cost does not reflect. A firm that has spent months building an effective AI governance program, configuring AI tools for its specific practice area, developing the prompt libraries and workflow integrations that make AI productive in client work, and training its professionals in AI use has created service delivery value that the vendor’s per-token charge does not capture.
The disclosure obligation associated with markup is straightforward but important: clients who are being charged a marked-up AI usage rate should understand that they are being charged for AI usage and that the rate includes a service margin. This disclosure does not require itemizing the firm’s cost and margin separately — it requires that clients know they are being charged for AI as a service component of the engagement. Engagement letters that authorize AI usage billing should specify the billing mechanism (per query, per session, per period, or as a percentage surcharge on professional fees) without necessarily disclosing the specific markup percentage, but should be clear enough that a client reviewing their invoice can understand what the AI-related charges represent.
For both pass-through and markup models, the client data governance implications of AI usage tracking deserve specific attention. The systems used to track which AI interactions were associated with which client matter may themselves contain client information — the queries submitted, the documents processed, the outputs generated — that is subject to the firm’s confidentiality obligations. AI usage tracking infrastructure must satisfy the same data governance standards as the AI tools themselves, which means managed AI deployments with proper audit logging designed for client matter attribution rather than cobbled-together tracking from consumer AI tool activity logs.
Absorbed in Professional Fees: Simplicity With Cost Management Requirements
The absorbed model treats AI consumption costs as overhead — embedded in the professional fees the firm charges, not separately itemized or disclosed as a line item. Under this model, AI tools are part of the firm’s service delivery infrastructure, the same way its research subscriptions, office software, and professional development are overhead costs that inform the firm’s rate structure without being separately billed. The client pays professional fees that implicitly reflect the efficiency and quality benefits of AI-assisted work, without seeing AI as a separate cost component.
The absorbed model is the simplest billing approach from an operational standpoint — it requires no per-matter AI usage tracking, no invoice line items, and no client communication specifically about AI costs. It also eliminates the billing ethics questions that pass-through and markup models raise in regulated professional services contexts. The entire question of whether AI usage is a billable expense or overhead becomes moot when AI costs are absorbed into professional fees rather than separately billed.
The cost management challenge of the absorbed model is the inverse of its simplicity benefit: without per-matter tracking, the firm does not have visibility into how AI consumption costs are distributed across its client base. Some clients may generate substantially higher AI consumption costs than others — due to matter complexity, document volume, or intensive AI-assisted research requirements — without those costs being reflected in their billing. If AI consumption costs are material and unevenly distributed, absorbed billing may create cross-subsidization that underprices high-consumption matters and overprices low-consumption ones relative to actual service delivery costs.
Managing the cost risk of absorbed billing requires monitoring total AI consumption costs relative to total professional fee revenue and adjusting rate structures if AI costs grow to the point where they materially affect the firm’s economics. This is a portfolio-level cost management function rather than a matter-level tracking function — simpler to execute but requiring consistent monitoring to ensure that AI consumption cost growth does not erode the margins that absorbed billing was designed to maintain.
Tracking Infrastructure for Matter-Level AI Attribution
Pass-through and markup billing models both require matter-level AI usage tracking — the ability to attribute AI consumption costs to specific client matters with sufficient accuracy for billing purposes. Building this tracking infrastructure is one of the primary operational benefits of managed AI deployments over consumer AI tool use for professional services firms.
Consumer AI tools do not provide matter-level usage tracking. They generate aggregate usage data organized by account or user, not by the client matter the usage was associated with. A firm that attempts to implement pass-through or markup billing for AI usage from consumer AI tool accounts must either rely on professional honor systems (each professional self-reports AI usage associated with each matter), estimate allocation from billing records (allocating AI usage proportionally to hours billed per matter), or forgo the billing model because the tracking is not feasible. None of these approaches is adequate for a firm that takes client billing seriously.
Managed AI workspace deployments built for professional services can be configured with matter codes or project identifiers that professionals select when beginning an AI session, creating usage records that attribute each AI interaction to the specific matter for which it was performed. These matter-attributed usage records are the billing source data that enables accurate pass-through or markup invoicing — and they are also the audit trail that supports the firm’s billing if a client scrutinizes AI usage charges.
The FTC’s guidance on business pricing transparency addresses the disclosure requirements that apply when businesses charge customers for services and service components — including the transparency obligations that apply to professional services billing practices and that inform the disclosure standards for AI cost billing in client engagements.
The NIST AI Risk Management Framework provides the governance architecture for AI systems used in professional services contexts — including the audit logging and data governance functions that make matter-level AI usage tracking possible within the confidentiality architecture that client data protection requires, ensuring that AI usage tracking serves billing purposes without creating the additional data exposure that poorly designed tracking systems generate.
Professional services firms that build deliberate AI consumption billing policies — selecting the billing model that fits their client relationships and professional conduct obligations, implementing the tracking infrastructure the model requires, and communicating AI cost handling clearly in engagement letters — are managing AI as a professional service component with the same intentionality they bring to other aspects of their client value proposition. Firms that do not address AI billing policy are making a default decision — typically absorption without cost management — that may not serve their economics or their client relationships as AI consumption costs grow with adoption.
