From Seats to Outcomes: The New Economics of Legal AI

For more than three decades, the Software-as-a-Service (SaaS) industry was built around one of technology’s most successful business models: charging customers per seat. This pricing model gave customers predictable costs based on team size because it was based on the number of licensed users. SaaS followed this across industries including customer relationship platforms, enterprise collaboration tools and LegalTech platforms. The economics were compelling with revenue growing alongside headcount in a linear fashion while the marginal cost of serving an additional user was minimal. This model is beginning to break down in the age of Agentic AI.

Unlike traditional software, agentic AI does not merely assist professionals in doing legal work such as research, drafting and contract review. It can perform end-to-end workflows. However, every contract reviewed or due diligence report generated involves highly variable costs such as tokens and compute costs. The more you use an AI-native product, the more expensive it gets because of these variable costs. Thus, AI-native software is defined by a different set of economic principles than traditional SaaS.

This shift is appearing in Legal AI. Legora recently announced that its flagship product, Agent Pro, would be shifting to a consumption-based pricing model. Rather, than just charging law firms and in-house departments for accessing the tools, these companies are charging based on the consumption of AI (Eg: Credits and Tokens etc.). Consumption-based pricing and outcome-based pricing should not be confused. Outcome-based pricing would potentially involve charging based on completed legal work such as number of contracts reviewed, due diligence reports drafted or term sheets drafted.

This piece argues that Legora’s pricing shift should not be understood as an isolated commercial decision. As Legal AI becomes increasingly agentic, the industry is likely to evolve from charging per seats, to charging for AI consumption and ultimately to pricing based on completed legal outcomes.

The SaaS model that defined LegalTech

Before Generative AI (“GenAI”) entered the profession LegalTech products including Contract Lifecycle Management (CLM), E-discovery, Document Review and Practice management were priced on a per-seat basis. This was because software was just a tool for lawyers to perform tasks through. The lawyer produced legal work while software acted as an enabling layer. Furthermore, more users meant more revenue.

SaaS tools pre-GenAI era did not perform the work themselves. A CLM platform organized agreements and helped classify templates. Legal research tools surfaced relevant authorities and lawyers still had to “Boolean search” to find relevant case laws. The economics were compelling for such companies as well because software companies were able to generate predictable, recurring revenue because growth tracked closely with headcount. Other than this, once the software had been developed and deployed the cost of serving an additional user was marginal. This created the high-margin, recurring characteristics of SaaS businesses that made it attractive for investors and founders.

Larger firms that employed more lawyers had to buy more seats thereby leading to a linear relationship between increase in revenue and the number of seats. However, Agentic AI breaks that assumption because these tools are able to perform end-to-end workflows of multiple lawyers. With Agentic AI, lawyers act as orchestrators of such systems where software produces the work. Thus, charging per-seat would mean that these companies are being penalized for using efficient technology.

The Economics of Agentic Legal Work

Traditional SaaS scales users. Agentic AI changes the economics of legal work. Every AI interaction carries highly variable costs such as tokens and compute costs. A contract review, due diligence report generated or petition drafted carries multiple AI inference costs. Thus, the additional cost of serving a user is no longer negligible and depends largely on the complexity of the legal workflow.

SaaS businesses selling Legal AI tools could no longer justify seat-based pricing because these tools could perform tasks done by multiple lawyers. Unlike traditional SaaS, costs now increased alongside usage. Legal work is particularly token-intensive because it involves reviewing multiple contracts, drafting petitions and generating various versions of a term sheet. This requires a tool to have multi-document analysis, long context window, iterations/variations and retrieval. The defining characteristic of AI-native software is that every additional unit of value delivered incurs an additional cost. Agentic AI allows firms to perform more legal work without a proportional increase in headcount.

Monetising Compute, Not Licenses

If AI-native software companies incur costs every time legal work is performed, charging a fixed-fee irrespective of usage becomes increasingly difficult to justify. Legora’s decision is not only a pricing innovation but a result of changing economics of AI-native software.

The key insight to draw from here is that Legora is charging for AI usage as opposed to software access. Legora’s consumption-based pricing model is based on the number of credits that is used coupled with the dashboard that can assist in tracking real-time spending based on matter and client. Furthermore, thresholds can be set on how much to spend limiting spend. This model better aligns with customer usage. In doing so, pricing becomes better aligned with changing economics of such work.  

Consumption-based pricing brings with it its own benefits such as LLM optimization, clearer ROI measurement and prompt discipline but it has drawbacks as well. This pricing model is aligned with AI usage but clients and law firms are increasingly concerned about the outcome or value generated. Thus, this is why consumption-based pricing is to be viewed as a transition phase. The next challenge for Legal AI companies is to shift simply from monetising usage to showing how their tool helps deliver legal work at scale efficiently.

Under consumption-based pricing, controlling inference costs becomes a competitive necessity because every token consumed affects the provider's margins.

The Race to Tokenmaxx

Consumption-based pricing fundamentally changes the basis of competition in Legal AI. Under the traditional SaaS model, companies competed by acquiring more customers and expanding software licenses. However, under the consumption-based model every AI interaction simultaneously generates revenue and incurs inference costs. This means that cost-efficiency becomes a strategic advantage from an operational concern.

The result of this is what may be termed as “tokenmaxxing”. In traditional discourse, “tokenmaxxing” refers to the phenomenon of maximising token usage or in other words consumption of AI to win an internal gamified dashboard as has happened in Uber and Amazon. For the purposes of this article, I use the term "tokenmaxxing" differently, to mean extracting the greatest amount of legal value from every token consumed.

The objective is no longer to just build a capable Legal AI platform but to deliver the same quality of legal work with fewer tokens, lower inference costs and more efficient compute utilisation. A platform that drafts a contract or reviews a due diligence report with half the inference cost of a competitor will enjoy a structurally superior unit economics. Maximizing AI consumption is not maximizing AI value.

Token optimization is one of the next big strategic advantages in Legal AI. Legal AI companies will start to invest in model-routers, fine-tuning open-source models, prompt optimization and context compression to reduce such costs without comprising on quality. Law firms and in-house teams will develop prompt discipline and specifically the need on when AI-first workflows are cost-effective than traditional workflows as well.

Ultimately, consumption-based pricing rewards usage but efficiency becomes a competitive advantage. As inference costs such as tokens become an increasing part of unit economics, the winners will not be the companies generating the most tokens but those delivering the greatest amount legal work for lesser inference costs. This shift in unit economics is what will eventually transition the industry into outcome-based pricing.

Selling Legal Outcomes, Not Compute

Legora’s shift towards consumption-based pricing should not be viewed as the destination of Legal AI business models but as a transition towards a fundamental change in how legal services are monetized. Traditional SaaS monetized software access through seat-based pricing. Consumption-based pricing monetizes through AI usage such as credits (billing) and tokens. As Agentic AI performs end-to-end workflows, the next logical evolution is to monetize completed legal outcomes.

Clients do not purchase AI because they want to consume more tokens. They purchase AI to review contracts faster, complete due diligence more efficiently, draft petitions with greater speed or automate compliance workflows. The economic value therefore lies not in the AI consumed but in legal work successfully delivered. The desired legal outcomes have to be achieved accurately, efficiently and at predictable cost.

Outcome-based pricing presents its own challenges. Pricing legal work by outcomes requires Legal AI system to consistently reliable outputs that both the client and vendor can evaluate objectively. Until, these capabilities mature, consumption-based pricing is likely to remain the bridge between traditional SaaS subscriptions and fully outcome-based pricing models.

The evolution of such pricing models indicates a shift in what Legal AI companies sell as well. Yesterday, they sold access to software. Today, they’re selling AI usage. Tomorrow, they’re likely to sell completed legal outcomes-based number of contracts reviewed, due diligence reports drafted and petitions drafted etc. The companies that will likely succeed in this transition will not be the ones with the powerful models, but those capable of delivering the greatest amount of legal value at the lower cost and ultimately, those able to price the outcome rather than the compute.

---

This article has been authored by Harshith Viswanath, LegalTech Fellow at the Indian LegalTech Network and a law student at the National Academy of Legal Studies and Research (NALSAR) University, Hyderabad.