AI is everywhere in legal practice, but the real transformation is yet to begin

“The legal sector is past the point of considering AI adoption. The challenge now is the gap between individual use and organization-wide deployment.”

Litify’s latest State of AI in Legal 2025 feels like a snapshot at a turning point. In three years, artificial intelligence has moved from hype to daily practice. In 2023, 23 percent of legal professionals reported using AI; in 2025, that figure is 78 percent. Adoption is moving faster than the transition to the cloud did back then. Yet the report reveals a more nuanced picture than simply “AI is now mainstream.”

The study is the third annual edition of the same trend study. Litify collaborates with a major independent market research firm that verifies and analyzes the results. The findings are based on hundreds of confirmed respondents from across the legal chain, with nearly equal representation from in-house counsel and full-service firms. The sample ranges from small firms with fewer than 25 employees to organizations with over a thousand people, and from juniors to managing partners. The margin of error is approximately 6 percent at a 95 percent confidence level.

This does not provide a theoretical vision of the future, but rather a fairly sharp image of where the sector stands today. That image is compelling: legal professionals themselves are embracing AI rapidly, while organizations struggle with policy, security, and genuine business impact. Consequently, the question is shifting from “are we going to use AI?” to “how do we extract structural value from it?”

AI is no longer an experiment

Those who still view AI as a toy for a few tech enthusiasts will find a different reality in this report. In two years, AI usage among legal professionals rose from 23 percent in 2023 to 78 percent in 2025, with over two-thirds of respondents indicating they use AI daily or weekly.

Most legal professionals primarily use general, accessible applications.

The report shows that:

• General systems such as ChatGPT are by far the most widely used.
• Built-in AI in office software, such as Microsoft Copilot, is rapidly gaining ground.
• Other integrated assistants like Google Gemini are utilized where they are included in existing systems.

These solutions are already present in the programs that legal professionals work in all day. The step to type a question into an AI is therefore small. This makes adoption easy, but says little about the maturity of the organization as a whole.

From individual experiments to organizational challenges

A clear tension runs throughout the report: the individual legal professional is often ahead of their own organization. Professionals consider technology important, as do employers, but in terms of policy, it often remains limited to isolated initiatives.

On the side of the legal professional, it is evident that:

• Many respondents consider AI important for performing their work effectively.
• Employees expect their organization to actively embrace modern technology, including AI.

On the organizational side, by contrast:

• Less than half indicate they receive sufficient training to use AI safely and effectively.
• Only a minority work within an organization with a clear AI policy; for a large group, policy is still “under development” or absent.

In practice, this results in a recognizable situation:

• Individual legal professionals use AI intensively, but often on their own initiative.
• Formal frameworks lag behind or are formulated so generally that they provide little guidance.
• Risks regarding privacy, confidentiality, and quality are present but are not structurally addressed.

The report makes it clear that this is not an exception, but the current average.

What legal professionals are actually doing with AI

When looking at the specific tasks for which legal professionals use AI, a recognizable pattern emerges. AI is primarily an accelerator for work they were already doing.

The most frequently mentioned applications are:

• Assisting with legal and case-oriented research.
• Summarizing case and file histories.
• Drafting, reviewing, and analyzing documents and contracts.
• Extracting data from documents to populate case files.

Litify describes this as “search on steroids”: valuable, certainly in busy practices, but still heavily focused on finding, organizing, and summarizing information. This aligns with the fact that many legal professionals use general, non-legal AI tools that are often not deeply integrated with legal platforms or document management systems.

Additionally, a second layer is emerging: a smaller group of organizations is already using AI for more process-oriented and legally specific tasks, such as:

• Building timelines around case files.
• Answering legal questions.
• Pre-screening new cases for potential conflicts or missing documents.

Such applications require integration with existing systems. AI must not only see the text you input but also the context of your case and document systems. Only then does true process support emerge, rather than isolated text production.

Beyond that, the third layer comes into view: agentic applications, in which AI independently performs entire steps in a workflow, for example:

• Automatically assigning cases to the appropriate staff member.
• Drafting draft invoices based on time tracking and case data.
• Actively contacting or reminding clients via various channels.

The report shows that only a small portion of organizations currently deploy these types of applications, but adoption is starting roughly at the level where generative AI stood in 2023.

The common thread: nearly everyone uses AI as an information and document engine, a growing group links AI to processes, and a small vanguard is exploring agentic possibilities.

Security, quality, and confidentiality act as a brake

Why is progress toward these more mature forms of AI use not faster? The main obstacles are well-known but remain dominant.

Litify shows that:

• Concerns about the confidentiality of client and case data remain high.
• Security and data privacy are cited by many respondents as major barriers.
• Doubts about quality, accuracy, and the risk of “hallucinations” play a significant role.

The authors draw a clear parallel with the introduction of cloud technology in the legal sector. At that time, the reflex was similar: client data belonged on in-house servers, not in an external environment. Today, cloud solutions are the norm, and the discussion has shifted to the configuration and security of those environments, rather than the principle itself.

For AI, the report outlines a similar path:

• Concerns are substantively real and consistent with the nature of legal information.
• The solution lies in clear prerequisites: policy, technical safeguards, control mechanisms, and targeted training.

Without that foundation, AI remains stuck in isolated experiments at the individual level. With that foundation, AI can safely and predictably become part of the legal operation as a whole.

The Litify AI Maturity Scale

To distinguish between “we use AI” and “we derive business value from it,” Litify introduces an AI maturity scale with four levels: Explore, Expand, Integrate, and Transform.

In broad terms, the scale looks like this:

• Explore
  Employees experiment individually with general “freemium” tools to perform simple tasks, such as summarizing or quick search queries. Policy rules are limited, and the impact is primarily noticeable at the individual level.
• Expand
  The organization makes its first investments in AI solutions for broader deployment. Tasks are selected more deliberately, but readiness is uneven: some teams lead while others lag behind. Positive effects are visible in parts of the organization but are not yet structural.
• Integrate
  AI is linked to existing platforms, case files, and workflows. There are clear policies, controls, and training programs. AI supports more complex, multi-step tasks, and leading organizational metrics, such as turnaround time and error rates, show demonstrable improvement.
• Transform
  AI is fully embedded and takes over complete tasks within clear boundaries and with appropriate oversight. Organizations see clear, measurable results: lower costs, accelerated revenue potential, and an adjusted cost base. AI is factored into budgeting, personnel planning, and strategy.

The report makes it clear that many legal organizations are currently between Explore and Expand, with a smaller group already moving toward Integrate. As organizations move higher on this scale, AI shifts from a handy tool to a structural engine in the legal infrastructure.

Individual gains, limited business impact

At a personal level, legal professionals experience clear benefits from AI. When asked about the most significant effects, respondents mentioned, for example, that:

• They can find the information they need more quickly.
• They can complete their work faster.
• They can provide better service to clients.
• They can deliver higher quality work and make better-informed decisions.

At the organizational level, the line is less distinct. The report shows that:

• Investments in AI have increased significantly in recent years, and AI tools are now at the top of the “purchase list” for legal departments.
• Only a small minority of organizations indicate that AI is already leading to clearly measurable business impact, such as structurally lower costs or noticeably higher revenues.

According to Litify, this ratio—high individual gain, limited business impact—cannot be permanent. As AI consumes a larger portion of the technology and innovation budget, executives will begin to treat it like any other investment: goals will be set, results measured, and decisions made based on return on investment.

What does this mean for your firm or legal department?

For Dutch law firms and corporate legal departments, the lessons from the report translate well into daily practice. Three movements are important in this regard.

Go beyond the isolated AI tab of individual legal professionals
As long as AI is primarily something legal professionals do “on the side” in their own browser tab, the effects remain fragmented.

The step toward more mature usage begins with deliberate choices:

• Determine which AI solutions belong to which workflows in your organization.
• Define which tasks are intentionally supported or partially performed by AI.
• Ensure everyone works with the same rules so that quality and risks remain manageable.

This requires involvement from practice group leaders, management, and IT, not just enthusiastic individuals.

Establish a solid foundation of policy, technology, and training
The report shows that the biggest constraint lies in concerns about confidentiality, quality, and security, combined with a lack of clear guidelines and training.

A mature approach therefore requires, at a minimum, that:

• A clear AI policy is established, aligned with professional rules, privacy legislation, and internal quality standards.
• Technical measures are implemented, such as secure environments, logging, role and access rights, and the separation of public and private models.
• Legal professionals receive targeted training with examples from their own practice, including the verification of AI output and attention to pitfalls.

Only when legal professionals know what is permitted, what is possible, and how to verify AI results will an organization feel confident deploying AI more broadly and deeply.

Link AI usage to concrete indicators

Finally, it is necessary to connect AI usage with measurable goals.

Litify emphasizes the importance of tracking “leading metrics,” such as:

• Turnaround time for standard documents and procedures.
• The number of cases a legal professional can handle without loss of quality.
• The number of correction rounds on standard products.
• Client satisfaction and response times.
• The balance between substantive legal work and administrative burdens.

Without these figures, AI remains an abstract improvement project. With these figures, you can specifically demonstrate where AI truly makes a difference and where adjustments are needed.

From handy assistant to structural engine

The State of AI in Legal 2025 depicts a legal sector that has embraced AI at a rapid pace but is still very much searching for the next phase. The vast majority of legal professionals now work with AI and experience real benefits in their own cases. At the same time, policy, integration, and measurable business impact are still unevenly distributed across the sector.

For legal leaders, this is precisely the mandate. Those who continue to view AI as an individual toolbox are leaving significant potential untapped. Those who dare to take the step toward integration, clear frameworks, and measured results are building a practice in which AI not only accelerates individual tasks but also enhances the organization’s effectiveness, resilience, and competitive position.