AI for Legal Departments
Any legal department can benefit from artificial intelligence use, and you can also spend a lot of money investing in AI tools that nobody uses. The goal of this article is to outline the different types of AI solutions available, the critical components for effectively implementing them in legal departments, and offer guidance on modernizing your legal department without wasting time or money.
AI adoption will always bring certain risks, but mitigations are available when the AI is implemented appropriately. It is important to approach AI with a pragmatic understanding of the limitations in AI accuracy, the needs of your business, and the range of tools available.
Ironically, it all starts with the people in your legal department.
Step 1: AI as a Tool Requires Human Adoption
A tool is worthless if nobody is using it. This isn’t a new challenge in legal tech, AI has simply given an old problem a new face. Adopting the right tool for your legal department requires collaboration with the individuals it will support. A top-down mandate typically results in low adoption, quiet resentment, and the fear of being replaced by a machine.
It’s worth saying plainly:
The only legal professionals AI will replace are the ones who merely copy/paste into and out of AI tools without applying their judgment.
Some individuals, however, will never adopt or use artificial intelligence tools. Legal teams can be creatures of habit, and I’m certain that there are lawyers who refuse to send emails. Winning these steadfast individuals over on the productivity and efficiency benefits of AI may require extra steps, and the most useful strategy for this education is live demonstration of AI tools solving an actual problem.
Step 2: Understanding the different types of AI tools
There are several categories of AI tools relevant to legal departments. At a high level:
| AI Type | Description | Example |
|---|---|---|
| General-Purpose AI Tool | An AI tool whose sole purpose is to generate responses to input. These typically look like chatbots | chatGPT Gemini Claude |
| AI Features in Existing Software | AI features added into existing tools | Ironclad Microsoft Word (w/CoPilot) Many more |
| Wrapper Software & AI Backend | Software with a custom user interface and backends to a third party AI tool (such as chatGPT). | Cursor AI Many more |
| Legal Specific AI Tools | Legal specific AI tools | Harvey Lexis+ AI |
| Internal or Hosted AI Models | Models hosted on internal machines or isolated AWS instances | AWS Bedrock + Models Custom Hardware/Open Source (Llama) |
Each of these categories has different costs, tradeoffs, accuracy levels, and implementation requirements. Assessing the size, maturity, and needs of your legal department is essential to selecting the right type of solution.
Step 3: Understanding your business, your legal department, and your current software tools.
A fortune 100 company with over 200k employees will have different legal needs than a startup. When reviewing potential AI tools, it is important to understand the variety of tools available, and the specific improvement you’re trying to capture. Reviewing your current software licensing is important as well, because you may already license the solution to your problem, or have opportunities to enable AI features within those tools at lower costs.
While a full Legal Specific AI Tool may not be ideal for many companies, almost all legal departments can benefit from some AI usage. A general-purpose AI tool can bring substantial productivity to the general operations of your legal department.
| General-Purpose AI Tools (Gemini, chatGPT, Claude, Perplexity, etc) | |
| Pros | Cons |
| Low cost subscriptions | Less accuracy |
| No integration costs | Limited access to internal data |
| Immediate updates/innovations | Less control over updates/innovations |
| Multi-functional, not limited to legal tasks | |
A general-purpose tool, configured and licensed properly, can generate summaries and status reports from the largest of documents, organize data, write scripts for excel functions, and simplify jargon in writing to communicate with a broad audience. It can turn your entire legal department into a team of amateur coders, developing proof of concepts, compliance apps, and automated intake forms in minutes. While General-Purpose AI tools will have limitations with accuracy and specific functionality/access to data, the versatility and adaptability of the general-purpose AI tools at these costs can’t be denied.
| AI Features in Existing Software | |
| Pros | Cons |
| Lower cost subscriptions | Vendor lock-in (year-over-year price increases) |
| Lower implementation/integration costs | Limited versatility outside of the purpose |
| Purpose/task driven AI | |
| Access to existing data | |
AI features in existing software can be an advantageous solution for solving specific legal problems at lower costs. Corporate email within Google workspace can benefit from Gemini identifying the five (5) most important emails, Ironclad can take a first pass of redlines based on a contract playbook, or identify the contracts with Alaska as the choice of governing law. These tools can provide access to your specific legal data, and support legal specific use cases with less disruption.
While these AI features may be helpful for many corporations, you must understand your legal department’s needs and your business. If the specific features you are looking at cost $100k to license with a $50k professional services implementation SOW, sometimes General-Purpose AI tools may fulfill the need at substantially lower costs.
Sometimes software is built on top of the AI tool, similar to the features in existing software, but for a specific (sometimes legal) purpose.
| Wrapper Software & AI Backend | |
| Pros | Cons |
| Sometimes lower cost subscriptions | Vendor lock-in (year-over-year price increases) |
| Sometimes interchangeable AI Models | Integration costs |
| Software designed for AI usage | Limited versatility outside of design |
| May include access to existing data | |
Identifying these as software “wrapping” AI sometimes carries a negative connotation. However, these types of AI tools can offer additional advantages, such as user selection of the specific AI model underlying the tool. This permits flexibility for the legal department to utilize the latest and innovative AI models that are offered for those systems. Looking for this AI model selection is important when examining these types of tools.
This has brought us to the legal AI tools, the ones that were purposely built for legal professionals to use.
| Legal Specific AI Tools | |
| Pros | Cons |
| Higher accuracy in legal tasks | Higher costs |
| Designed for legal use | Vendor lock-in (year-over-year price increases) |
| May provide access to existing data | |
| Access to legal specific data | |
These include a variety of tools such as Lexis+ AI, Harvey, and many others specifically designed with the purpose of legal department use. These tools have the greatest access to legal training/data provided by the vendor and deliver the highest levels of legal accuracy in their output. However, the cost per seat of these licenses tend to be the highest cost. Licensing these tools makes sense for some corporations, while others may not see their return on investment.
However, if you have a highly technical individual on your legal team, running AI models internally or hosted may be the most cost effective choice.
| AI Models (internal or hosted) | |
| Pros | Cons |
| Usage-based costs (very low costs) | Limited support |
| Closed environment | Technical complexity |
| AI Model Versatility (for updates/innovations) | |
| Unlimited versatility | |
| Access to corporate data | |
Running an AI Model within local computer hardware or AWS provides a legal department with a low cost knowledge base that has access to its own corporate data within a closed and secure environment. This means contracts, memos, drafts, and confidential information can be used with AI features such as retrieval-augmented generation (RAG) and multi-context protocol (MCP) to create customized AI chatbots for your legal department. Your team could ask your knowledgebase to draw connections between all rights/restrictions in your contracts for each classification of data, or identify favorable alternative language your colleague drafted for a similar contract within seconds.
While versatile, these systems require implementation and support. Managing these systems is feasible by a single technical individual who understands your legal department. If you have a legal operations professional or team, this option may make the most sense to your department.
Final Decisions
Many vendors will pitch an AI tool for your legal department, but evaluating how it is built, how it will be implemented, and its limitations will be critical for identifying systems that actually benefit your legal team. Before licensing any tools, run a pilot and gather feedback. A short trial period will reveal issues, expose workflow gaps, and highlight whether the tool fits your culture.
In the end, the best AI solution will always be the one your legal team will actually use.
Thanks for reading.
-Shawn