5 AI & Legal Tech Predictions for 2025 and How to Prepare

We at Pocketlaw have seen impressive growth and adaptation of AI in the Legal Tech space during 2024, where AI & Legal Tech are at the top of the agenda for many in-house legal teams. This is something that we have witnessed firsthand with our customers’ increased PLAI usage since May (+1260%) this year and also from our various AI breakfast events during the year with high attendance.

From our point of view, 2023 was the year of showing the added value of using generative AI for Legal Tech, and 2024 was the year to use generative for real in legal workflows. Then, 2025 will be focused a lot on scale.

In this article, we have gathered predictions (both from our internal experts but also from broader surveys for key trends in the adoption of Large Language Models (LLMs) and generative AI) for where AI-driven legal tech solutions and tooling are heading, from agentic workflows to the hyper-personalisation of legal tech solutions. We have tried to make our predictions as grounded and relevant as possible from a legal tech perspective; no AGI here, unfortunately.

1. Agentic Workflows: The Future of Automation in Legal Tech

Our first prediction is that Agentic AI will shift from isolated tasks to automating legal contract workflows. This is driven by the rise of Agents - autonomous AI systems capable of completing legal workflows with minimal human intervention.

All the major LLM providers (as well as open-source providers) such as OpenAI with their O1-model, Google Gemini 2.0 with its agentic features and computer use powered by the latest Claude models show advances in agentic capabilities during this year, and will likely continue during 2025.

Looking at the industry in general, there are a lot of talks about so-called vertical agents, and we will likely experience a plateau of sorts in terms of the rate of improvements for the foundational models used, which are driving the LLM development mainly as we have soon come to exploit the internet as a source of training data, see more in IIya Sutskever’s recent talk.

The trend and future trajectory for AI Agents have also been highlighted in several studies/surveys this year from various independent providers such as:

• LangChain (2024) -  State of AI Agents: Trends and Challenges

• Langbase (2024) -  State of AI Agents in Enterprise Application

• Menlo Ventures (2024) - The State of Generative AI in the Enterprise

As mentioned in the survey, “LangChain (2024) - State of AI Agents: Trends and Challenges,”:

In 2024, AI agents are no longer a niche interest. Companies across industries are getting more serious about incorporating agents into their workflows.

The survey also states some interesting statistics regarding AI agent production usage:

About 51% of respondents are using agents in production today

The Langchain study continues with mentioning the following:

Encouragingly, 78% have active plans to implement agents in production soon.

Although not all our products here at Pocketlaw require Agentic workflows, we do still use AI agents and related concepts quite heavenly in our AI-assisted Contract Review offering, where these types of AI agents tend to shine and provide higher accuracy compared to chains of LLM calls.

However, even if Agents and Agentic workflows show a lot of promise, there are still issues and barriers to getting Agents into production, which both studies above mention. Some of the problems they mention are connected to infrastructure, cost, and wider adoption.

However, from a legal tech point-of-view, we see the following challenges:

1. Improved Reasoning Capabilities: Current agents and LLMs need help with multi-step reasoning and maintaining context over long workflows (i.e., some memory capabilities). At Pocketlaw, as a legal professional, you are always in control, and we have inbuilt mechanisms for you to steer the output of the LLM or agent. For instance, in our AI contract review, you can rerun each rule individually and provide additional context to steer the agent's behaviour.

2. Domain-Specific Intelligence: In a legal context, Agents must be tailored to the nuances and intricacies of legal work, such as jurisdictional differences, ethical concerns, or regulatory requirements. At Pocketlaw, we use our internal legal domain expertise to fact-check and ground the responses of LLMs and Agents. In the context of contract review, you can steer and incorporate this behaviour in your legal playbook.

3. Trust and Accountability: Accountability is essential in a highly specialised field such as legal tech. LLMs are known to be non-deterministic by design, so we at Pocketlaw always provide reasoning traces to explain the agents’ actions further. We always build solutions with a human-in-the-loop mindset, meaning that you will always have the final say before an LLM or Agent can make any decisions.

In 2025, we will see a lot of talk all about agents of various forms and how these can improve legal workflows. Although there are some challenges, the blend or mix of human-in-the-loop combined with advanced automation will be unbeatable.

At Pocketlaw, we believe that agents and agentic workflows will be a great companion to you as a legal professional by augmenting and helping you, not replacing you.

2. Hyper-Personalisation Will Reshape Legal Service

Our second prediction is that Hyper-Personalisation Will Reshape Legal Service. A few factors drive this. Firstly current trends indicate that the cost of intelligence will continue to reduce drastically, for instance, with the latest LLama 3.3 release you can run GPT-4 type models at almost no cost locally.

Of course, running this in a production scenario is entirely different. This indicates that providing the best and most accurate legal service will be imperative. While leveraging internal or external documents or databases is critical for legal tech solutions, the search and recommendation experience will become equally important.

The concept of hyper-personalisation is not new but is instead a natural continuum in industries where data and AI are adapting at a much faster pace. Consider why you sometimes buy certain goods or clothes from a particular store or listen to music on various streaming platforms.

Recommendations and the hyper-personalised nature of the client vs. customer engagement have been key factors here. Why shouldn’t legal tools have a similar transformation to better gather for your needs as a legal professional?

Some potential scenarios of how this could manifest that we see:

1. Customised Workflows: AI systems will adapt to individual lawyers' unique writing styles and preferences while automating repetitive tasks with their specific working methods. For instance, with the PLAI assistant today, you can draft documents or suggested clauses based on your specific way of writing or talking. Or learn from your previous interactions.

2. Client-Centric Features: Tools will deliver more tailored legal strategies and instant insights for clients analysing historical data, case-specific details, and past and future (prediction) outcomes to create relevant, actionable recommendations. For instance, at Pocketlaw today, you can build custom dashboards based on extracted metadata from your document repository and create actionable insights such as alerts and triggers for when e.g., contracts of a specific contract value expire. We also believe that time to insight should be easy and accessible for anyone, so you can now use natural language (NLP) to get custom and personalised insights from your contracts.

3. Adaptive Learning: AI tools will further evolve based on user interactions by employing short-term and long-term memory to make these agentic systems truly personalised and intuitive. Of course, with data privacy in mind, you can decide a time to live for these “memories”.

Hyper-personalisation will have a large impact on the end-to-end user experience for legal tech tools. By making them more intelligent, context-aware and adaptable to your unique needs and problems as a legal professional.

3. Legal Professionals As Legal AI Strategists

Our third prediction is that legal professionals will become more pivotal in the role of Legal AI strategists. As AI evolves, legal professionals will pivot to adapt to this role, leveraging AI in their day-to-day lives to amplify their impact rather than being replaced by it.

One key factor is that the legal spaces have generally been slow adaptors when embracing new tech tools and solutions. However, this dynamic has completely shifted with the advent of LLMs and generative AI.

Remember, AI will not replace legal professionals; legal professionals augmenting themselves with AI will replace those who don’t.

The Langbase (2024) - State of AI Agents in Enterprise Application study also shows what we at Pocketlaw have seen this year a large increase in the usage and adoption of LLMs for legal use cases.

Fig 1. What use cases are you using LLMs for? Source: Langbase.

The study shows that LLMs are primarily used in software development (87%), followed by text generation and summarisation (59%), marketing and communication (50%), IT & Operations (48%), customer service (43%), HR & Finance (26%), and legal use cases (15%).

The number one use case from this study is very relevant for Legal cases; for instance, a key usage we see of our PLAI assistant together with the tools of comparing various documents with each other or drafting and suggesting clauses. To succeed in these use cases, you need to shift or pivot your thinking to be more of an AI strategist.

However, looking at Menlo Ventures (2024) - The State of Generative AI in the Enterprise. Generative AI spending by department is the highest for IT or product & engineering (19-22%). Still, legal spending is only 3% (a prediction within a prediction here is that we will see a 3-5x increase here). This means there is a lot of potential here to use Generative AI in even more legal workflows, where these so-called AI strategists will be essential stakeholders, adopters, and gatekeepers.

The Legal AI Strategists will possess a unique skill set, including:

1. Enhanced Decision Making: AI systems will provide actionable insights and predictions, allowing legal professionals to refine their strategies, be grounded in more extensive sources of information, and make better-informed decisions during negotiations or litigation. At Pocketlaw today, the PLAI assistant or our repository insights can be helpful here, and you can store your favourite prompts and know-how in our prompt library solution.

2. AI Literacy: A good understanding of AI capabilities will be necessary, and we will likely see more AI training from various actors for legal professionals. At Pocketlaw, we did a round of prompt engineering training for legal professionals this year, for example. But as a legal professional, it is also crucial for you to effectively understand, critique, and manage AI tools to ensure the best and most relevant output for your needs.

3. Client-Centric Innovations: With more usage of AI-powered platforms such as Pocketlaw, automating routine or boring stuff for you, legal professionals will have more capacity to focus on personalised advisory roles with deeper client relationships and focus on providing even more tailored legal solutions.

The shift or role of “AI strategist” is not exclusive to the legal domain; many other fields have or will shortly experience a similar type of evolution of the main occupation. This will allow them to focus on strategic, high-value-creating activities instead of non-value-creating mundane tasks.

4. The Cost Of Intelligence Will be Close to Zero

Our fourth prediction is that the cost of intelligence will reach or be near zero. As LLMs, both open source and closed source, get better, faster, and cheaper, more use cases and higher adoption will likely be reached. A key factor for this trend is the increased commoditisation of LLM and AI intelligence and economic of scale factors for the large LLM providers.

Looking into the LangChain (2024) - State of AI Agents: Trends and Challenges study one of the highest barriers in putting, e.g. more agents in production is the cost aspect.

LangChain's report identifies the biggest limitations of putting more agents into production as performance quality (41%), cost (18.4%), safety concerns (18.4%), latency (15.1%), and other factors (7%).

Fig 2. What is your most significant limitation of putting more agents in production? Source: LangChain.

The trend is also exemplified in the current market share of LLMs, which is compared to 2023 vs 2024 in the study Menlo Ventures (2024) - The State of Generative AI in the Enterprise:

Fig 3. Change in market share for all the major LLM providers 2023 vs 2024. Source: Menlo Ventures.

The report shows OpenAI's market share dropping from 50% in 2023 to 34% in 2024, Anthropic rising from 12% to 24%, Meta holding steady at 16%, Google increasing from 7% to 12% and Mistral AI slightly decreasing from 6% to 5%.

There are no public records on what LLM providers are most used in a legal tech setting. I do expect that the distribution is somewhat similar to the graph above. All major providers, including OpenAI, Anthrophic and Google, have done quite a lot of cost-cutting over the year and noticed the cannibalisation of OpenAI to the other providers. We will likely see Meta share grow more next year due the the increased performance and fine-tuning availabilities of the LLama model family.

Some potential implications that we see for legal Tech:

1. Lower Barriers to Entry: Startups and smaller firms will have more opportunities to develop and explore specialised AI tools or solutions for Legal Tech, which will drive an increase in competition and innovation. However, these must be built with the legal domain in mind, as Legal Tech will only give additional value if it is not only the next cool tech but with an emphasis on solving real issues and legal use cases. This is why we at Pocketlaw believe the real magic happens when AI technologists and domain experts collaborate to build use case-tailored legal tech solutions.

2. Focus on Differentiation: With the cost of intelligence reaching close to zero or AI is becoming a more or less standardised commodity, tech solutions and or platforms will need to have other USP and differentiation factors to make them enjoyable for users:

3. User Experience (UX): Intuitive platforms seamlessly integrate with existing workflows.  At Pocketlaw, we take pride in providing a superior and intuitive user experience on our platform. Any new feature we add needs to adhere to our strict design guidelines and principles. Our word plugin moves our PLAI offering closer to your premise, and our Web API makes it possible to integrate with other solutions, such as Salesforce, seamlessly.

   • Data Privacy: Robust compliance and security measures are critical and should be explicitly tailored to the legal domain. We have built Pocketlaw with privacy by design; your data is your data and no one else’s. We are also GDPR compliant and SOC2 compliant.

   • Domain Expertise: Legal domain knowledge will continue to be very important. Legal Tech tools must be specialised for the unique nature of legal workflows, whether the LLM is fine-tuned or context-tuned via solutions such as RAG.

4. Holistic Solutions: Providers will likely shift from offering standalone tools, such as supporting only an AI assistant or supporting only parts of the CLM journey. Instead, we will see the rise of more comprehensive platforms that combine multiple capabilities, from drafting a contract to sending it to eSigning for various parties, to generate insights on what happens after the signing process. At Pocketlaw we have always had a platform-as-a-service (PaaS) mindset and will continue expanding our capabilities.

As the cost of intelligence approaches zero, the legal tech field will become more competitive than ever. Success will no longer hinge on having AI capabilities, but rather on a holistic product offering that is seamless, secure and specialised to solve real problems in the legal workflow.

5. Legal Tech Will Likely See Some Consolidation

Our fifth and final prediction is that we will likely see some consolidation in the legal tech space. Legal Tech has experienced much growth in recent years, driven by the introduction and application of generative AI. This has also led to a fragmented market with many different providers starting from various angles, such as CLM or purely AI-based solutions, e.g., contract analysis. In the coming years, there will likely be some consolidation in the legal tech market.

Some scenarios that could unfold:

1. A CLM-focused platform acquires AI native company to accelerate the integration of AI, LLMs, and agentic workflows into the legal process.

2. AI-first or agent startup partners with or acquires CLM providers to expand their feature set and close the gaps in comprehensive legal workflows.

3. Larger technology companies acquire or start partners with legal tech providers to integrate them into their offerings. We have already seen this with consultancy firms such as PwC.

4. Major tech firms like Microsoft, Google or Amazon, are launching native legal tech solutions to compete directly with existing providers.

5. Some legal tech companies are exiting the market due to increased competition and reduced access to venture capital.

Note that these are hypothetical scenarios that could happen, not necessarily occur. However, (1) - (4) could be seen as a push to drive continued momentum toward building end-to-end legal tech platforms rather than isolated AI-enabled features. (5) could be a side-effect of this. Ultimately, this consolidation trend is inevitable as both user adoption increases and the market continues to saturate and mature.

This will simplify user adoption for legal professionals, allowing end-to-end platforms to be selected instead of having to handle various fragment tools. Legal tech players who anticipate the shift and act decisively will be well-positioned to lead the industry into the next phase of growth and innovation, where we at Pocketlaw will continue to innovate and grow with our clients.

Conclusion

As we look ahead to 2025 (and beyond), it’s clear that the legal tech industry is poised for a transformative change driven by the maturation of AI technologies and the evolving market dynamics. Below are some of the key takeaways from our five predictions for 2025:

1. AI Will Scale and Specialise: Agentic workflows and advanced AI tools will become more capable of handling complex legal tasks, offering legal professionals more efficiency and accuracy when using these tools. At the same time, maintaining a human-in-the-loop approach for accountability and trust.

2. Hyper-Personalization as a Competitive Advantage: Legal Tech platforms that embrace client-centric features, adaptive learning, and tailored workflows will stand out in an increasingly crowded market. Customisation will be the key to delivering exceptional value.

3. Legal Professionals Will Lead the AI Evolution: Deep domain knowledge will continue to be a competitive advantage to gather for the unique needs of the legal domain, where legal professionals will have a paramount role.

4. The Democratization of AI: With the cost of intelligence approaching zero, more organisations—large and small—will have access to cutting-edge tools, fostering innovation and heightening competition. Holistic solutions will win over more fragmented offerings.

5. Consolidation Will Drive Integration: The legal tech space will likely see mergers and partnerships that push towards comprehensive end-to-end platforms, which will simplify user adoption and further solidify AI’s role in legal workflows.

At Pocketlaw, we remain committed to being at the forefront of these changes, delivering innovative, user-focused solutions that empower legal professionals to navigate the evolving landscape. From our CLM solutions to AI-powered PLAI offering to offer a genuinely end-to-end legally tailored experience.

Hopefully, some of these predictions will make you think and prepare for what tools and solutions you need as a legal professional. And introduce some questions, thoughts and checkpoints before you decide what tool to use.

The future of legal tech isn’t just about adopting AI - it’s about using it to amplify or augment human expertise and more creative ways compared to how it has been done before. Together, we can build a smarter, more efficient, and accessible legal industry for all.