The 10 Types of Legal Work and Why AI Treats Each One Differently
Legal work breaks into ten distinct types. Each has different cognitive demands, risk profiles and AI fit. The map every legal AI buyer needs before choosing a tool.
1. Legal Drafting
This is the bread and butter. Contracts, briefs, motions, memos, opinion letters, board resolutions, partnership agreements — lawyers draft constantly, across every practice area.
A junior associate at a mid-size firm might spend three hours drafting a non-disclosure agreement that differs from the last one they drafted by exactly four clauses. A partner at a litigation boutique might spend an entire weekend writing a motion to dismiss. Both are drafting. Neither task is simple.
Where AI fits: AI can generate solid first drafts from prompts, templates, or prior work product. It can enforce consistency with a firm's style guide and produce jurisdiction-specific variations without the lawyer having to start from scratch every time.
Where it falls apart: When AI drafts like a non-lawyer. Generic output that misses jurisdiction-specific requirements, invents clauses that don't exist in practice, or produces text that reads like it was written by someone who has never set foot in a courtroom. Good legal drafting AI understands context — it knows the difference between a Delaware LLC agreement and a UK LLP agreement.
2. Contract Review and Analysis
Reviewing contracts to extract key terms, identify risks, flag non-standard provisions, and compare against market norms or internal standards. This is the work that keeps transactional lawyers up at night during deal season.
Picture this: a real estate portfolio acquisition with 500 leases. Each one needs to be reviewed for termination clauses, liability caps, governing law, assignment restrictions, and a dozen other data points. Miss a single change-of-control provision and your client could lose a key tenant the day the deal closes.
Where AI fits: AI can process hundreds of contracts in minutes, extracting structured data and flagging deviations from a client's or firm's standard positions. What used to take a team of associates two weeks can now be done in hours.
Where it falls apart: When it over-extracts or under-extracts. When it gives false confidence. When it fails to distinguish between a material deviation (uncapped liability) and a cosmetic one (a slightly different defined term for the same concept).
3. Due Diligence
If contract review is the daily workout, due diligence is the marathon. In M&A and corporate transactions, lawyers review thousands — sometimes tens of thousands — of documents to identify risks, liabilities, and issues that affect deal valuation or structure.
Due diligence is also where burnout lives. Large transactions can involve reviewing 50,000+ documents in a data room. Junior associates are thrown into this work during their first year and expected to surface issues that could cost their client millions.
Where AI fits: AI can scan entire data rooms, categorize documents, flag issues, and generate due diligence reports. It can surface a buried change-of-control clause in a vendor contract that a tired associate at 2 AM might miss.
Where it falls apart: When it treats every finding as equally important. A change-of-control clause in a key customer contract worth 30% of revenue is existential. The same clause in a minor office supply agreement is irrelevant. Good due diligence AI understands materiality.
4. Legal Research
Finding relevant statutes, regulations, case law, and secondary sources to support legal arguments or advise clients. This is foundational to almost everything lawyers do.
Traditional legal research — Boolean queries on Westlaw or LexisNexis — requires its own expertise. Junior associates learn to construct complex queries and pray they haven't missed a relevant jurisdiction. It's powerful but brittle.
Where AI fits: Natural language queries instead of Boolean logic. AI can also synthesize across jurisdictions and identify authorities that keyword searches systematically miss.
Where it falls apart: Hallucinated citations. This is the problem that has made headlines — lawyers submitting briefs with AI-generated case citations that don't exist. Good legal research AI grounds every citation in actual source material and never invents one.
5. Contract Negotiation and Redlining
After the first draft is exchanged, the real work begins. Lawyers compare versions, propose redlines, negotiate terms, and go back and forth — sometimes for weeks — until both sides can live with the result.
Anyone who has tracked the differences between version 7 and version 12 of a 100-page agreement knows the special kind of tedium this involves. And yet the work is critical: a single overlooked redline can shift millions of dollars in liability.
Where AI fits: AI can generate redlines based on a firm's established playbook, suggest alternative language when a counterparty rejects a position, and track negotiation history across multiple rounds.
Where it falls apart: When it can't distinguish between substantive and cosmetic changes. When it flags a formatting edit with the same urgency as a liability cap reduction.
6. Playbook Generation
Firms and in-house teams create playbooks — standardized positions on common contract terms. These playbooks guide junior lawyers through negotiations so they don't have to call a partner every time the other side pushes back.
Here's the problem: most firms don't actually have written playbooks. The institutional knowledge lives in partners' heads. When that partner retires or moves to another firm, the knowledge walks out the door.
Where AI fits: AI can analyze a firm's historical contracts to reverse-engineer their actual negotiating patterns and generate playbooks automatically. It can compare a firm's positions against market data and flag where they're being unusually aggressive or leaving value on the table.
7. Discovery and Document Review
In litigation, parties exchange relevant documents and information through discovery. E-discovery alone was the largest segment of the $31.59 billion legal tech market in 2024.
We're talking about teams of contract attorneys billing $200–500/hour to review documents one by one. An Am Law 100 firm recently reported cutting document review time by two-thirds using generative AI.
Where AI fits: AI can draft discovery responses, identify responsive documents, apply privilege reviews at scale, and manage massive document sets.
Where it falls apart: When the results can't withstand judicial scrutiny. A privilege call made by AI that can't explain its reasoning is a privilege call that will be challenged — and potentially waived.
8. Timelines and Chronologies
Building chronological narratives from documents, depositions, and evidence. This is critical for litigation strategy, trial preparation, and regulatory investigations — and it is one of the most tedious tasks in all of legal practice.
Where AI fits: AI can extract dates, events, and actors from thousands of documents and assemble chronologies automatically. What took weeks can take hours.
Where it falls apart: When it gets dates wrong or misattributes events. When it can't handle conflicting information. Good timeline AI flags contradictions instead of silently picking one.
9. Regulatory Advising and Compliance
Analyzing regulatory frameworks, monitoring regulatory changes, advising clients on compliance obligations. A multinational corporation operating in 40 countries needs to track regulatory changes across every one of those jurisdictions — in real time.
Where AI fits: AI can monitor regulatory changes across jurisdictions continuously, map regulatory requirements to specific business activities, and generate compliance checklists.
Where it falls apart: When it conflates proposed rules with final rules. When it misidentifies applicable jurisdictions. Regulatory work has zero tolerance for 'close enough.'
10. Deal Management and Advocacy Strategy
These two categories sit at the top of the complexity spectrum, where AI augments human judgment rather than replacing human labor.
Deal management is the coordination layer — tracking conditions precedent, managing closing checklists, coordinating across counsel. AI can generate and track checklists, identify dependencies, and flag items at risk of delay. But the strategic decisions remain human.
Advocacy strategy is the highest-value work in litigation: developing case theory, preparing for trial, analyzing judge tendencies. AI can analyze a judge's prior rulings and help anticipate opposing counsel's positions. But the creative leap — the insight that reframes the entire case — is still the lawyer's to make.
Why This Decomposition Matters
These ten categories of work sit on a spectrum. On one end, you have highly automatable tasks — document review, timeline building, discovery responses. AI can handle 80–90% of the work. The human role is quality control and final judgment.
On the other end, you have AI-augmented but human-led tasks — advocacy strategy, deal management, regulatory advising. AI provides data, surfaces patterns, and handles analysis. But the lawyer makes the call.
The money — and the real value for firms — is in the middle. Categories like drafting, due diligence, and contract review, where AI can do 70–80% of the work but human judgment is still essential for the last mile.
According to the 2025 Clio Legal Trends Report, the average lawyer bills only 3.0 hours in an 8-hour day. Their utilization rate is 38%. Nearly three-quarters of billable tasks are exposed to AI automation.
The Architecture That Works
The smartest legal AI companies have figured out something important: you can't build one model to rule them all. You have to build specialized systems for each category of work, optimize for accuracy within each narrow domain, and then bring them all back together into a single interface.
Type a case law question, and the system should pull up its research engine. Upload a merger agreement, and it should trigger the due diligence workflow. Ask it to draft an NDA, and it should activate its drafting module with your firm's templates pre-loaded.
The user shouldn't have to navigate menus, choose modules, or understand the underlying architecture. They should just describe what they need. The AI handles the routing.
This is what we're building at HAQQ. Not a chatbot. Not a single-purpose tool. A legal operating system that understands the full spectrum of legal work and meets lawyers where they are.
Because the goal was never to turn lawyers into machines. It's to give them better tools, so they can be more present, more strategic, and more human in the work that actually matters.