A buyer’s guide to legal AI pricing
A practical guide to subscriptions, token-based pricing and emerging commercial models, and what they mean for legal AI buyers.
Legal software has traditionally been priced by subscription: customers pay a monthly or annual fee for a set number of users.
Yet with legal AI, the underlying costs are often measured in tokens (small units of text processed or generated by a model), so heavier use can cost providers more.
The legal AI market is offering a growing range of pricing models. But the bigger question is who carries the risk: the provider, or the customer?
We look at the pros and cons of each legal AI pricing model, what they mean for buyers and how the market is evolving.
The price of a token
Not all tokens are equal. The same task can cost very different amounts depending on the model used, how much reasoning it requires and how many tokens it consumes to reach an answer.
One recent study published by Microsoft found token consumption alone is not a reliable indicator of the quality or value of an AI-assisted outcome.
For buyers, the key is to look beyond the price per token to the total cost of completing a task, including model choice, agent behaviour and spending controls.
Passing token costs directly to clients may become more common, but whether clients will accept it is another question. They may rightly ask whether the usage was necessary, why a premium model was chosen, or why repeated agent activity should become their cost rather than the firm’s overhead.
The strongest pricing models connect the commercial arrangement to customer-recognisable value while providing an appropriate level of predictability. Depending on the service, this may include continuous access, completed workflows or clearly defined outcomes.
“Token expenditure and economic value are distinct.”
Model one: token-based pricing
Token-based pricing means customers pay according to how many tokens an AI model processes. Tokens are small units of text used when the AI reads a prompt or document and generates a response.
This model has a clear logic. Customers who use little pay less while those who run intensive workflows pay more. Its downside is that it transfers more of the financial risk to the customer.
The bill rises according to the amount of information sent to and generated by the AI. For a lawyer, input tokens may include prompts, documents, retrieved content and conversation history. Output tokens cover the answer and, depending on the system, parts of the model’s reasoning process.
It works particularly well for developers building their own products through APIs, as they can monitor usage closely, select models and optimise their systems around cost.
For a law firm lawyer or general counsel, tokens are less intuitive. A bill showing that a legal team consumed several billion tokens says very little about what the organisation received.
Costs can also become difficult to predict. Rapid growth may demonstrate adoption, but it also creates a commercial challenge. A provider cannot assume that token costs will remain stable while usage expands by an order of magnitude.
Using the most powerful model for every task may not be cost-effective. A better approach is to use the model best suited to the task, considering accuracy, security, speed and efficiency, while reserving more advanced models for work that genuinely requires them.
Model two: the predictable subscription
A fixed subscription model places most of the immediate usage risk on the provider.
The customer pays an agreed amount for access to the platform, usually based on the number of users, the products included and the length of the contract. The provider manages the underlying cost of different models, content sources and workflows.
This gives customers something increasingly valuable: cost certainty.
A lawyer can use AI for research, drafting or analysis without pausing to consider whether another prompt will increase the bill. A legal team can expand adoption without trying to forecast every document its people might upload.
Several factors can make this commercially sustainable:
- usage forecasting and cost management at scale
- efficient information retrieval
- intelligent model selection
- continued investment in content, security, product development and support.
The subscription model is particularly well suited to products that lawyers use regularly and where broad adoption is the objective. Its weakness is that customers may pay for licences that are rarely used. Procurement teams therefore need to examine adoption as carefully as price.
New AI workflows. Built for legal standards.
Model three: consumption without visible tokens
Consumption pricing does not have to expose the customer to raw technical units.
In this model, customers pay for the work delivered and costs attributed to the project that generated them.
This is an important development because it attempts to translate technical activity into a recognisable unit of work.
Rather than receiving a charge for millions of tokens, a customer might pay for a contract review, a due diligence project, a completed agent task or an investigation conducted across several sources.
The model is easier to connect to a matter, department or client. It also gives the provider revenue that reflects the intensity of the work rather than simply the number of people able to access the platform.
Agentic AI makes this increasingly relevant. A single user instruction may therefore trigger many actions. Consumption pricing reflects the activity beneath it. For subscription customers, this abstraction can also be an advantage: the provider manages the complexity of the underlying activity while the buyer retains greater cost certainty.
The challenge with this model is transparency. Customers will still need to understand:
- what constitutes a completed unit of work
- why one project costs more than another
- whether failed or repeated attempts are charged
- how human review affects the price
- whether the provider selects cheaper models where suitable
- what happens when a workflow exceeds its expected scope.
Replacing tokens with proprietary credits or units may make the invoice simpler without making the underlying price clearer.
Model four: the hybrid plan
Under a hybrid model, the customer pays a predictable base fee for the platform, security, support, integrations and normal usage. Additional fees apply only when the organisation uses expensive models, exceeds an allowance or runs unusually intensive workflows.
A plan might include:
- an annual enterprise subscription
- a defined number of users
- a shared organisation-wide usage allowance
- standard models included in the base price
- additional charges for frontier reasoning
- separate capacity for document-heavy work
- higher charges for autonomous agents
- spending controls at team or matter level.
Under this model, enterprise customers can buy predictable access for individuals while paying separately for applications and customised workflows.
The hybrid approach may appear to offer the best of both worlds, but it can also inherit the weaknesses of both. Customers may pay substantial base subscriptions and still face variable charges. A generous allowance may leave light users paying for capacity they never consume. A restrictive allowance may discourage adoption or create unexpected bills just as the technology becomes embedded.
The quality of a hybrid model will depend on whether the allowances reflect genuine customer behaviour and whether additional costs can be forecast before they are incurred.
Model five: pricing the workflow
Workflow pricing charges for a completed task rather than access or token use.
For example, a customer might pay per contract reviewed, document analysed or regulatory change assessed.
The benefit is commercial clarity: buyers can compare the price with the cost of doing the work manually or using an external provider.
The challenge is that tasks vary in complexity. A short contract is not the same as a complex cross-border agreement, and AI output may still require significant human review.
Workflow pricing works best when the task, quality standard and expected level of human input are clearly defined.
Model six: outcome-based pricing
Outcome-based pricing links the price to the value delivered rather than the amount of AI used.
That value might include time saved, lower legal spend, faster contract completion, reduced risk or increased capacity.
The appeal is clear: customers pay more when the technology delivers more value.
The challenge is measurement. Legal outcomes are often influenced by several factors, not AI alone. Outcome pricing is therefore most practical where the result is easy to define and measure.
In many cases, providers are more likely to use outcomes to prove value and support renewals or fixed fees, rather than calculate every invoice.
What buyers should look for beyond price
The Financial Times recently reported that legal teams are increasingly choosing different AI providers to solve different problems. For buyers, this makes the ability to connect models, trusted content and internal workflows increasingly important.
Law firms and legal teams that rely on widely available AI tools may find it harder to differentiate, while those that combine technology with their own expertise, data and workflows have a stronger point of difference.
A strong platform can bring these together in one place, including:
- multiple AI models
- authoritative legal content
- internal knowledge and data
- connections with third-party platforms and applications
- bespoke agents
- integrations with document and matter systems
- a secure interface for completing work.
The Financial Times argues that established legal technology providers such as LexisNexis have an important advantage through trusted legal content and expertise alongside proprietary data and partnerships.
Sean Fitzpatrick, CEO of LexisNexis’s global legal business, has explained that partners can license LexisNexis AI technology, allowing users to ask a legal research question from another platform, have it answered through Protégé and receive the response in the environment where they are already working.
This reflects an important principle: legal professionals should be able to access trusted legal intelligence within the workflows they choose.
Our customers need predictability
Subscription pricing gives organisations greater certainty over expenditure while allowing legal professionals to use the platform without assigning a cost to every prompt, document or underlying model action.
LexisNexis currently uses a subscription-based approach designed to provide predictable access as customer adoption grows. Its scale, authoritative legal content and multi-model architecture help it manage the technical complexity behind the service. The platform can retrieve relevant information efficiently and select the model best suited to a task, considering quality, security, speed and efficiency.
In a July 2026 interview, Sean Fitzpatrick said LexisNexis had no plans to move to consumption pricing, noting that customers value the predictability of the subscription approach.
Predictability is not simply a budgeting benefit. It also allows users to explore and embed AI-supported workflows without worrying that every additional prompt or agent action will produce an unexpected charge.
“While others are talking about moving to a consumption model, we don’t have any plans to do that, and our customers really like that.”
The return-on-intelligence problem
Counting prompts is not measuring value. A useful framework for evaluating legal AI moves through three levels — from what the platform is actually being used for, to how it is performing operationally, to what it is delivering for the business. Business impact provides the strongest basis for evaluating and communicating the return on AI investment.
Questions legal AI buyers should ask
Before selecting a provider, buyers should understand:
What is included in the quoted price?
Understand whether the price covers users, AI models, legal content, integrations, support and implementation.
How predictable will our total annual cost be?
Check whether usage, premium models or additional workflows could push costs above the headline price.
Which AI models and legal content are included?
Buyers should know what intelligence sits behind the product and whether authoritative legal sources are built in.
Can it work with our own documents, data and workflows?
The strongest platforms should fit around the organisation’s existing knowledge and ways of working.
What security, privacy and governance controls are included?
Legal teams need confidence that sensitive information is protected and AI use can be managed responsibly.
How will we measure whether it delivers value?
Look beyond usage figures to outcomes such as time saved, faster turnaround, improved capacity and reduced external spend.
How might the pricing model affect us in the future?
Consider how costs could change as adoption grows, AI agents become more common and teams rely on the platform for more complex work.
Certainty has a value
AI is reducing the cost of some forms of intelligence while creating a new category of expenditure that many organisations are still learning to manage.
Tokens are a useful technical measure, but they are not a measure of legal value.
Legal teams do not ultimately need more tokens. They need reliable research, faster reviews, better decisions and greater capacity. Law firms need a sustainable way to deliver work that clients recognise as valuable.
Each pricing model has trade-offs. Subscriptions offer certainty, but only if the product is used. Consumption pricing can appear fair, but costs may be harder to predict. Workflow pricing is easier to understand, but only when the task is clearly defined. Outcome pricing aligns price with value, but only when that value can be measured.
This makes adoption critical. A predictable subscription only delivers value when people actually use the product. LexisNexis combines AI with familiar legal research and workflow tools, supported by training, customer support and an established user base, helping organisations embed new capabilities into everyday work rather than treating AI as a standalone experiment.
For buyers, the strongest providers will make the complexity easier to manage — choosing the right technology, connecting it with trusted information, supporting adoption and providing confidence in both the result and the bill.
The legal industry has spent generations measuring work in six-minute units. It should be cautious about replacing one imperfect measure with another.
2026 research & thought leadership reports
The Mentorship Gap
AI can help new lawyers to search, summarise, draft, and reason. But can it teach?
Integrating AI into legal workflows
Anything-goes AI use won't raise your standards. Legal teams need to restructure their workflows based on suitable AI.
Death of the rainmaker
Love them or loathe them, rainmakers are a law firm's bread and butter. But will AI-powered legal solutions see the practice take precedence over the partner?
In CTO we trust
Chief Technology Officers at leading law firms face a new challenge: not just deploying AI, but making it trustworthy at scale.
Workflows that draft
Create outstanding drafts with workflows built on your documents and LexisNexis sources.
Pre-built & custom workflows
Move faster on everyday tasks and complex decisions with workflows that fit how you work.
Trusted content
Strengthen consistency with workflows that combine organisation knowledge and exclusive sources.