Pragmatic use of AI in the modern legal practice

Real-world tools, ethical considerations, and practical strategies to improve efficiency, advocacy, and client service

2026 February

Artificial intelligence is no longer speculative in the legal profession. It is operational. It is embedded in tools we already use, layered into platforms we already pay for, and increasingly present in the workflows of firms of all sizes. Whether attorneys feel ready or not, AI has become part of modern legal practice.

Yet despite its rapid adoption, many attorneys remain uncertain – if not uneasy – about how AI should be used. The concern is not unfounded. Headlines about fabricated case citations, sanctions for improper reliance on AI-generated research, and opaque data practices have made many lawyers wary. Others fear that AI may erode professional judgment, commoditize legal work, or push firms toward efficiency at the expense of quality.

These concerns are legitimate. But they are also manageable.

The core issue is not AI itself. The issue is how AI is adopted, where it is deployed, and whether it is governed by intentional decision-making rather than convenience or hype.

I am not an AI evangelist, nor am I a skeptic. I am a practicing trial attorney who runs a growing and busy firm. Like many of you, I care deeply about the quality of my work, the trust my clients place in me, and the reputation of our profession. My interest in AI and advanced technology comes from a practical place: How do we deliver excellent legal work more efficiently while protecting our clients, without losing control of the process?

To frame that discussion, I often rely on an analogy that resonates with me personally – building a house.

A practical roadmap: Building the house

Broadly speaking, every well-built house follows the same basic phases:

1. Permits
2. Foundation and framing
3. Finish work

Legal technology – and AI in particular – can be adopted in the same way. Skipping steps leads to problems later. Moving too fast creates risk. Building intentionally creates leverage. This framework has helped me evaluate every AI tool, platform, and vendor we have considered.

Phase one: Permits – compliance, security, and ethical guardrails

Before you break ground on a construction project, you need permits. You need to make sure you are complying with building codes, zoning laws, and safety requirements. In the context of legal technology and AI, “permits” mean ethical compliance, data security, and client confidentiality.

The ethical reality of AI in law

AI tools raise serious and legitimate concerns for attorneys:

• Attorney-client privilege
• Confidentiality of sensitive client data
• Cybersecurity and data breaches
• Accuracy and reliability of outputs
• Whether client data is used to train models

Many of the most widely known AI tools – such as ChatGPT, Claude, and Gemini – are built on large language models (LLMs) that operate in what is colloquially referred to as an open universe. These systems pull from vast amounts of publicly available information across the internet and, depending on the platform and settings, may use user inputs to further train their models.

That creates a problem for attorneys. Uploading client documents, medical records, deposition transcripts, or case-specific facts into open-universe systems can risk violating attorney-client privilege and ethical duties.

Public AI models versus legal-specific AI

It is important to distinguish between public AI models – such as ChatGPT, Gemini, and Claude – and legal-specific AI tools designed for professional use, including CoCounsel, Lexis AI, Eve, and Harvey. Even at the enterprise or paid tier, public AI models are general-purpose platforms and are not built or marketed as legal- or medical-privacy-grade systems. Their providers do not claim that these tools satisfy attorney-client privilege standards, HIPAA requirements, or legal-specific confidentiality obligations. While enterprise plans may offer enhanced security controls and contractual assurances around data handling, they are still not purpose-built for regulated legal environments. However, this does not mean they lack value; quite the contrary.

Some consumer versions of public AI models do allow organizations to operate within a closed environment and prohibit the model from training on uploaded data, which can reduce certain security and data-retention risks; most have also instituted the appropriate safety measures within their models, e.g., SOC 2 Type II compliance. Still, these measures do not equate to HIPAA compliance or legal-grade privacy protections. As a result, while enterprise controls can mitigate risk, they do not always convert a general-purpose AI model into a compliant system for handling protected health information or privileged legal materials.

In fact, I have had direct conversations with individuals who have worked at these AI companies, and those conversations have consistently confirmed this reality: Public AI models are not designed to function as privileged legal workspaces. They are powerful language tools, but they were never intended to serve as secure repositories for confidential client records, medical files, or case-specific legal analysis. This is not a criticism of the technology – it is an issue of alignment. Using public AI models responsibly requires attorneys to understand what these tools are and, just as importantly, what they are not. Treating them as legal-grade systems creates unnecessary risk. Still, my intent is not to suggest that HIPAA compliance is a binary litmus test, but rather to encourage attorneys to make intentional, informed choices: understanding the tradeoffs between general-purpose and legal-specific AI tools, and selecting the level of risk, functionality, and peace of mind that aligns with their own practices and comfort level – just as they already do with every other aspect of their technology stack. A solid approach is to reserve public AI for non-confidential tasks and rely on closed-environment, legal-specific platforms when handling client data, privileged materials, and substantive legal work.

Of course, that does not mean these tools are useless. In fact, they can be extremely effective for:

• Drafting non-case-specific correspondence
• Rewriting or refining language
• Structuring arguments or outlines
• Marketing and educational content
• Internal firm documents

But attorneys must draw a clear line between general assistance and case-specific work involving confidential information, unless heavy redactions are made to the documents uploaded to public AI models; but then again, heavy redaction adds work to be done instead of eliminating tedious work. 

Security standards matter

Before adopting any AI or tech tool, firms should ask:

• Is the platform SOC 2 Type II compliant? – An independent, time-based audit that verifies an organization not only has appropriate security and data-protection controls in place, but that those controls operate effectively over a sustained period of time.
• Is data encrypted at rest and in transit? – When data is encrypted at rest and in transit, the information is protected by encryption both while it is stored on servers and while it is being transmitted between systems, preventing unauthorized access even if the data is intercepted or compromised.
• Is client data stored in a closed environment? – When client data is stored in a closed environment, the information is isolated within a controlled system with restricted access and is not shared with, exposed to, or used by external systems or for training public models.
• Is uploaded information used to train the model? – Preventing AI models from training on uploaded information is critical because it ensures confidential client data is not retained, reused, or incorporated into future model outputs beyond the scope of the specific task for which it was provided. 
• Is the AI company familiar with legal ethics and compliance?

Without satisfactory answers, you do not have permits, and so it may be best not to build on top of that…at least not with that particular model, or in that particular method of use. 

Phase two: Foundation and framing – case management systems

Once permits are secured, the next step is the foundation. In a law firm, the foundation is a case management system.

No AI tool can compensate for a weak or disorganized case management system. Platforms such as Filevine, Clio, Litify, Casepeer, and SmartAdvocate (to name a few) provide the structural framework for modern legal practices. They can manage:

• Client and matter data
• Deadlines and tasks
• Documents and records
• Communications
• Workflow automation
• Reporting and analytics
• …and much more!

AI is not a substitute for case management. It is an enhancement that sits on top of it.

Spotlight: Filevine’s AI capabilities and LOIS

In my practice, Filevine’s AI suite has been particularly impactful. Filevine has embedded AI directly into the system rather than treating it as a separate add-on. This matters because it allows attorneys to work and leverage AI within a secure, centralized environment, where the entirety of the case files, documents, notes, and information reside.

For example, when we upload medical records into a provider-specific card, the system:

• Automatically extracts CPT and ICD codes
• Identifies diagnoses and injuries
• Creates a medical chronology
• Generates a narrative summary
• Links information back to the source documents

Filevine also allows that data to be converted into draft demands. While these drafts are not final products, they certainly provide a strong baseline that attorneys can refine and personalize.

What once took hours – or days – can now happen in minutes. More importantly, it happens inside the case management system, without ever leaving the system, thereby reducing risk and friction. In fact, Filevine recently rolled out their built-in AI assistant, LOIS, designed to help attorneys and staff quickly extract, summarize, and interact with information already stored within a case file, allowing users to ask natural-language questions about documents, medical records, timelines, memos, notes, and key facts without leaving the Filevine environment. LOIS was built on the principles of Retrieval-Augmented Generation (“RAG”), which is a framework that enhances LLMs by connecting them to external knowledge bases, like Filevine, to provide more accurate information by grounding responses in retrieved facts rather than relying on training data. In the case of LOIS, it accesses all your specific case information to provide context-aware, accurate, and relevant responses that are grounded and retrieved from your case file, only. In this manner, LOIS becomes an AI assistant, at your beck and call to retrieve information at any moment. 

I have used LOIS countless times since the rollout, such that it has now simply become a part of my daily use. With one simple question inputted into LOIS, I can easily find any document, note, or information within any given case. I recently used it on a case that my firm was asked to take over for trial. I uploaded all the case files and began interacting with LOIS to obtain summaries of the case, files, law and motion, and medical treatment. All within minutes. This all serves as a much better, higher quality, and much quicker starting point than can be obtained otherwise…and all without ever leaving that same webpage.

Do not pay for AI capabilities and then fail to use them

Of course, while I have highlighted Filevine as an example, it is important to emphasize that it is not the only case management system incorporating AI-driven tools and capabilities. I reference Filevine simply because it is the platform my firm uses and the one with which I have firsthand experience. Many other case management systems – such as Clio, Litify, Casepeer, and SmartAdvocate, and others – are actively developing and deploying similar AI features, including document summarization, workflow automation, data extraction, and drafting assistance. The broader takeaway is not to switch platforms, but to take a closer look at the tools already available within your existing system. In many cases, firms are paying for AI capabilities they are not yet leveraging, and meaningful efficiency gains can be achieved simply by exploring and thoughtfully implementing the features already built into their current case management infrastructure. Do not leave efficiency on the table.

Phase three: The finish – where AI truly adds value

The finish work in a home is what people see. It is the beautiful color of paint, the style of base molding, crown molding, the entrance door, etc. It is also where AI generates the most excitement – and, if misused, the most risk.

Understanding large language models (LLMs)

Large language models like ChatGPT, Claude, and Gemini are advanced AI systems trained on enormous amounts of text data to recognize patterns in language in order to understand and generate human language. The ability to generate human-like responses enables them to summarize documents, draft text, answer questions, and assist with analysis across a wide range of tasks. They are flexible, powerful, and improving rapidly. 

Each model has its strengths. ChatGPT is known for being versatile with a creative writing style, making it useful for outlining arguments, refining language, drafting non-case-specific correspondence, and brainstorming themes for mediation or trial; among the three, ChatGPT tends to be the most versatile and consistent for general legal writing tasks, e.g., correspondences, emails, small memos, etc. Claude stands out for its ability to handle long documents due to its large context window (the amount of information a model can process and “remember” at one time, acting like its short-term memory), complex reasoning, and maintaining specific writing styles with a strong ethical focus, and is overall an excellent legal writer. Gemini has a deep integration with Google services and is great for understanding audio and video content. 

However, they also:

• Can hallucinate cases, citations, and facts
• Are not inherently legal-specific
• Require careful prompting
• Must be reviewed closely
• Not all models “think” or reason
• This is why legal-specific AI platforms are so important to also have as part of your AI tool bag. 

Closed-universe legal AI: designed for attorneys

CoCounsel was the first AI tool we adopted in our practice. What sets it apart is that it operates within a closed universe – the Westlaw database. This fundamentally changes both reliability and security.

Key features include:

• No training on your uploaded data
• Secure, encrypted environment
• Outputs limited to vetted legal sources
• Designed specifically for legal workflows
• We use CoCounsel for:
• Reviewing and summarizing medical records
• Creating detailed chronologies
• Preparing for depositions
• Identifying prior injuries or treatment gaps
• Drafting baseline demand letters
• Creating mediation and settlement summaries
• Contract and document analysis

Westlaw’s CoCounsel

One experience stands out. I once found myself with far less time than usual to prepare for a deposition (sound familiar?). Instead of my traditional approach, I uploaded the entire file into CoCounsel and ran targeted prompts – asking it to identify potential problem areas, inconsistencies, prior conditions, and timelines.

In a matter of hours, I was as prepared as I have ever been for a deposition. Not because AI replaced my understanding of the case, but because it accelerated the work that normally consumes days.

And, most importantly, CoCounsel is secure and meets all of the requirements with which we as attorneys are most concerned, including attorney-client confidentiality and HIPAA compliance. 

Furthermore, given that CoCounsel is now a part of Thomson Reuters (i.e., Westlaw), this means that the “closed universe” in which CoCounsel lives is the Westlaw universe, and the Westlaw universe alone. This lessens fears about hallucinations. To be clear, using a closed legal research environment such as Westlaw does not eliminate the risk of AI hallucinations, as those errors arise from the model itself, but it can meaningfully reduce certain risks by limiting outputs to vetted legal sources rather than the open internet. For that reason, AI – whether legal-specific or general-use – must always be carefully reviewed and independently verified by the attorney.

Of note, CoCounsel claims to use a multi-model AI architecture. Multi-modal AI uses several leading AI providers, e.g., ChatGPT, Claude, Gemini, to match the best-performing model for each specific legal task. What this means in practice is that instead of forcing every legal task through one pipeline, multi-modal AI selects the optimal model for each workflow. This flexibility ensures higher accuracy for legal-specific work and flexibility to evolve as better models emerge, without requiring attorneys to change vendors or retrain their teams. I would encourage readers to give this a trial run to confirm how it fits and meets their own needs and expectations. 

It is also worth noting that legal-specific platforms like CoCounsel often operate on AI models that are a generation behind the most cutting-edge public models, prioritizing stability, security, and reliability over immediate access to the latest capabilities. While this can mean slightly less-advanced language performance compared to general-use tools, many attorneys view that tradeoff as worthwhile when working with sensitive, case-specific, or privileged materials.

Lexis AI

Lexis AI offers similar capabilities within the Lexis ecosystem. Like CoCounsel, it operates in a closed environment and is designed for legal research, drafting, and analysis. Lexis+ AI also claims to use a multi-model AI approach. The takeaway is not which platform to choose, but that closed-universe legal AI is categorically different from open-internet tools.

Do-it-all options

Platforms like Eve, Harvey, Everlaw, and Anytime AI market themselves as comprehensive AI solutions. They aim to handle:

• Intake evaluation
• Medical chronologies
• Demand letters
• Discovery drafting and responses
• Document review

Some firms may find value in all-in-one platforms (from a pricing standpoint, they may make more sense depending on the size of your firm). Others may prefer best-in-class tools for specific tasks. My recommendation is simple: Take the meetings, try the demos. As great as a certain tool may sound, the best way to find out if it will augment your practice is for you to give it a test run. To date, I still find myself needing to put my hands on a product, test it out throughout a week or two, before I understand whether I will be able to benefit from it or not. At minimum, a meeting with a vendor will allow me to see them demo their tool so that I can envision what it will look like in practice. Ultimately, these tools are meant to be used as much as possible; and if I cannot see myself or my team doing that, then it might not be the tool for us. The fun part is when we run into a product that we were writing off, only to find out through a demo or trial that it will now be a mainstay. 

When Sharon (my sister, with whom I practice law) and I evaluate AI and technology, we always ask:

1. Will this make us more efficient?
2. Will it allow us to maintain excellence?
3. Will it improve the client experience?

If the answer is no to any of these, we pass. If the timing is just not right, well now we know with certainty what tool to get when the time comes. 

Deposition and transcript analysis – Steno, Inc.

Transcript Genius, offered by Steno, Inc. is one of the most accessible AI tools available. In fact, if you notice a deposition through Steno, you receive free access.

Its capabilities are many. Chat with your transcripts: Have an interactive conversation in a chat-like interface, where you can request facts, summaries, legal insights, etc. Interrogate multiple transcripts: Analyze and synthesize information across multiple transcripts; ask for, and uncover, contradictions, inconsistencies, etc. Customized transcript summaries: Tailored summaries of transcripts; adjusted for length, detail, and focus; hyperlinked citations. Iterate on responses: Refine responses in real-time, ask follow-up questions, request updates, etc. Find transcripts faster: Expanded search capabilities, including exact match and semantically related results (meaning that you can search transcripts for concepts, and not just keywords). You can also upload transcripts obtained from other sources or court reporters; you are not limited to transcripts obtained through Steno. 

Given that this is a free resource, through the use of just one noticed deposition through Steno, it becomes the lowest barrier of entry to AI tools at your disposal. This is an excellent entry point for attorneys curious about AI but hesitant to commit financially.

Medical chronologies and record analysis

EvenUp is widely known for AI-assisted demand letters with human review, and has been written about quite often. One lesser-known feature, though what I personally consider to be one of their strongest features, is their medical chronology/timeline generation.

When we are brought into a case late – especially close to trial – EvenUp allows us to upload thousands of pages of records and receive a clean, organized, visually intuitive chronology with citations. Nothing has come close to that yet, personally. That alone can justify its use in high-stakes cases.

PareIT was designed specifically with personal-injury attorneys in mind. Its HIPAA-compliant AI delivers structured medical chronologies, full case analysis, and interactive Q&A with the records, all in a very user-friendly platform. In fact, I would urge folks to give it a shot, as they have a free tier, where you can upload thousands of pages of medical records and receive a medical chronology for free. In addition to medical chronologies, it automatically organizes records into folders – billing, admissions, imaging, and more. This eliminates hours of staff time spent on manual organization.

Medical record retrieval and organization – ChartSquad

ChartSquad handles medical record retrieval for a flat fee, without per-page pricing. We have received tens of thousands of pages of records for a fraction of traditional costs. For example, I once received over 20,000 pages of medical records in a catastrophic-injury case for under $100, total! ChartSquad also provides a helpful dashboard to track all requests and now incorporates AI-generated summaries (in fact, they recently acquired PareIT, previously mentioned in this article). This eliminates countless hours of staff follow-up calls and administrative work.

And more tools!

Foundation AI focuses on document organization – automatically sorting files into structured folders, with minimal human touch. It removes a low-skill, high-time task from staff workloads, giving you back your time and helping improve the labor cost side of the ledger.

Client communication and intake automation

Smith.ai offers a hybrid of human and AI-assisted virtual receptionist services. We primarily use it after hours, so we never miss potential clients. Calls are recorded, summarized, and appointments can be scheduled automatically. 

We have also recently started using Speed AI to augment our phone and intake system. Speed AI is an AI-powered legal intake platform designed to work alongside a firm’s existing phone and intake systems by analyzing 100% of incoming calls to identify missed opportunities, score case quality of potential cases, and alert firms to high-value leads that may need follow-up, helping improve intake performance and conversion without replacing existing phone or CRM systems, giving firms visibility into intake quality, coaching opportunities for staff (which have already proven helpful), and data-driven ways to increase signed cases from calls that might otherwise have slipped through the cracks.

Whippy has transformed how we communicate with clients. It allows us to use a single firm number (in our case, our main office number) for texting while internally routing messages to the appropriate attorney or staff member. Clients no longer juggle multiple phone numbers (e.g., the main line, direct line to team member, case-specific SMS text number, etc.); they now simply have the one number that they can either call or text at any time. And, internally, we simply assign client numbers to the attorney or team member responsible for that specific client; and only that attorney or team member will get the messages whenever the client texts. Furthermore, Whippy has an app, such that communications now live inside our system rather than on personal devices. We also use Whippy for intake sequences and automated Google review requests – something that has significantly increased client feedback without awkward conversations.

Litigation drafting tools

Tools like EsquireTek and BriefPoint automate the creation of pleading shells and discovery responses. This work used to fall on junior attorneys or staff and consumed significant time. Automation frees those team members to focus on substantive legal analysis instead of formatting. Plus, in 2025, EsquireTek released “Omega,” its platform that fully automates the written discovery process for law firms, handling everything from client communication and drafting responses/objections to creating medical chronologies, promising completion in 25 days or less, with expert human review for accuracy. 

Content creation and marketing

AI has also transformed content creation. My sister Sharon manages our firm’s social media and podcast production (@kramirezlaw on Instagram and podcast “Art of Purpose w/ Kenny S. Ramirez” found on Spotify, Apple Podcasts, and YouTube), and people regularly comment on the professional-grade production quality of it all. I am the first to point out that she has an amazing skill and touch when it comes to production. Still, the volume of all she produces (given her day job as an attorney!) is only possible through AI tools that significantly reduce editing and production time. Tools like OpusClip, Adobe Premiere Pro, CapCut, and Canva are great places to start. For attorneys who want to build a presence but feel constrained by time, AI can remove friction without sacrificing authenticity.

Conclusion: Intentional adoption is the advantage

I often describe AI as the best first-year associate you could hire. It works fast. It handles time-consuming tasks. It produces strong first drafts. But it still requires supervision. It still requires review. Hyperlinked citations, verification, and attorney judgment remain essential.

AI will not replace lawyers. But lawyers who use AI thoughtfully will outperform those who ignore it. The goal is not automation for its own sake. The goal is leverage – freeing attorneys to do what only attorneys can do.

When adopted with the right permits, a strong foundation, and a thoughtful finish, AI becomes not a risk, but a powerful competitive advantage. I encourage you to demo and try out some of these tools, and any others available. Your practice will surely benefit, as will your clients. And please feel free to reach out at any time to chat about your experiences with AI and tech!

Kenny Ramirez is the founder and principal trial attorney of the Kenny Ramirez Law Firm, a catastrophic-personal-injury and wrongful-death practice based in the Inland Empire. He is the current president of the Consumer Attorneys of Inland Empire, and is an active board member and officer in various bar associations, including CAOC, HBAIE, WSBCBA, CAALA, and AAJ. In 2023, he received the CAOC Presidential Award of Merit for his expertise in negotiations on Senate Bill 235. He can be reached anytime at This email address is being protected from spambots. You need JavaScript enabled to view it.