An AI paralegal is an AI system that does recurring legal work end to end, the way a paralegal would. You send it a request by email, Slack, or Teams. It gathers what it needs, drafts, chases people, routes documents for signature, and tracks progress. Lawyers step in only for judgment calls.
Think of it as an AI employee that specializes in legal work.
This guide is about AI that takes recurring legal work off an in-house team's plate and returns it done. That makes it different from AI legal assistants, which help lawyers draft faster.
One request, start to finish
Here's what that looks like for a typical request.
On a Monday morning, a sales lead posts in Slack: "Can we get an NDA out to our new prospect today?" Nobody in legal has to chase it.
- Alaro opens the matter, checks the counterparty on Companies House, and sends your NDA.
- The prospect replies with their own paper instead. Alaro redlines it in Word with tracked changes, using your playbook positions.
- One clause falls outside the playbook, so it goes to your lawyer with a short note and a suggested position. Everything else is handled.
- Once the final version is approved, Alaro sends it for signature through DocuSign, chases the signatory, and files the signed copy.
- The sales lead gets the answer in the same Slack thread: signed and filed.
Your lawyer spent a few minutes on one clause. An AI legal assistant would have helped with the redline, but someone would still have chased the counterparty, sent it for signature, and filed it.
"Most legal work inside a company isn't hard. It's the routing, the chasing, and the follow-up. We built Alaro so a legal team can hand that work over the way they would to a good paralegal, and get it back done."
AI paralegal vs. AI legal assistant vs. human paralegal
In-house teams usually weigh four options for routine legal work. The difference that matters most is who does the work and who checks it.
| AI legal assistant (e.g. Harvey, Legora, Claude, ChatGPT) | Human paralegal (hired) | Virtual or outsourced paralegal | AI paralegal | |
|---|---|---|---|---|
| What you get | A faster draft, summary, or answer | A team member who does the work | Work done by an external person, task by task | Work done from request to completion |
| Who does the work | Your lawyers, with AI help | The paralegal | The provider's staff | The AI paralegal, with lawyers stepping in for judgment calls |
| Who handles the chasing, approvals, and signatures | Your team | The paralegal | Usually your team, unless scoped in | The AI paralegal |
| Where it lives | The tool's own app and add-ins | Your systems | Email and the provider's portal | Your inbox, Slack, and Teams |
| Who maintains it | Your team: prompts, workflows, playbooks | You: hiring, training, management | The provider | The provider |
| Best for | Research, drafting, and analysis by lawyers | Steady volume, with budget for headcount | One-off overflow | Recurring, multi-step work that has to get over the line |
You will also see "AI paralegal software" used for AI legal assistants sold to law firms. Those tools speed up a person's drafting or research. The person still does the work. The test is simple: does the work come back finished, or does it come back as a draft for someone to complete?
What does an AI paralegal do?
Most of what lands on an in-house team isn't legally hard. There's just a lot of it, and much of it happens around the document rather than in it. An AI paralegal takes on that work:
- Takes in requests from wherever they arrive (a Slack thread, a forwarded email, a CRM ticket), and works out what each one needs.
- Gathers the context: the precedent, the schedule, the figures, the right counterparty contact.
- Chases missing pieces until nothing is outstanding: approvals, signatories, details from the business.
- Negotiates within agreed positions: proposes redlines and strategies inside the playbook your lawyers set.
- Runs documents to signature, chases signatories, and files the signed copy.
- Tracks everything, so legal can see where every request stands.
Typical work: NDAs, supplier agreements, MSAs, order forms, renewals, intake, and triage. For real estate teams, it also covers professional appointments and collateral warranties.
How it works day to day
- Scope it like a hire. Agree the playbook, precedents, approval rules, and escalation points before any work starts.
- Delegate the way you already do. Forward an email or message it in Slack or Teams.
- Get it back done. You're pulled in only for approvals, exceptions, and judgment calls.
- It improves over time. Lawyers monitor the work and feed exceptions back in, so it gets better the way a good hire does.
Playbooks and memory
An AI paralegal works from your team's own material: templates, precedents, past negotiations, and the positions you've accepted before. Each key clause gets a standard position and approved fallbacks.
When a lawyer corrects something, the correction is fed back in, so the next matter starts from it.
Where it works
It works in the tools your team already uses, so nobody has to log into anything new. For Alaro, that means:
- Email: intake, replies to counterparties, and calendar invites
- Slack and Microsoft Teams: requests and updates
- Google Drive: exchanging documents through folders you share, with Alaro's files saved beside yours
- Microsoft Word: tracked changes and comments, plus comparison PDFs
- DocuSign: from preparing the envelope to filing the signed copy
Other systems, such as Salesforce and Ironclad, can be connected on request.
"We didn't want to build another app lawyers have to open. Alaro works in the inbox and in Slack, and each team decides what it can do on its own and what needs sign-off first."
Boundaries and escalation
You decide what it does on its own and what needs sign-off first, such as sending an email or a signature request. Anything outside the playbook goes to a lawyer, with the reason it was flagged.
Chasing and deadlines
Much of the work is following up. An AI paralegal chases missing details, approvals, and signatures until nothing is outstanding. It tracks where each signer is, sends reminders, sends calendar invites for key dates, and tells you where every request stands.
What it handles, by team
Requests come from across the business, not only from legal. A typical spread looks like this:
- Sales: NDAs, order forms, and customer contracts
- Procurement and IT: supplier and software agreements
- Real estate and development: professional appointments and collateral warranties
- Investment: fund documents such as LPAs and side letters, and the requests to send back to the fund manager
- People and HR: offer letters and employment letters
- Legal: intake, triage, renewals, and tracking every open request
"Sales teams no longer have to flood their lawyers with questions or wait on them for an update. They chase Alaro instead, so deals keep moving and legal stays focused on the work that needs a lawyer."
What an AI paralegal doesn't do
Knowing the limits is part of choosing well. A good AI paralegal is clear about all of these:
- It doesn't replace your lawyers' judgment. Strategy, risk appetite, and one-off, high-stakes advice stay with your lawyers or outside counsel.
- It doesn't work outside the rules you set. Anything beyond the agreed playbook, approval limits, or escalation points comes back to a person.
- It isn't infallible. AI can be confidently wrong. That is why lawyer oversight and a clear route for exceptions matter more than raw speed.
- It doesn't remove accountability. Your team still owns the decisions. Ask any provider who answers for errors, and how.
Will AI replace paralegals?
AI is replacing tasks, not the role. Clio's 2024 Legal Trends Report estimated that 69% of paralegal billable hours are spent on work that could be automated, against 57% for lawyers. Yet the US Bureau of Labor Statistics projects paralegal employment to stay roughly flat from 2025 to 2035, with about 40,800 openings a year.
Both can be true. The work that disappears is the repetitive, first-pass layer. The work that grows is checking, coordinating, and making calls on what the AI hands back.
Paralegal tasks AI is taking on
- First-pass document review and summaries
- First drafts from precedent and templates
- Chasing approvals, signatures, and missing information
- Routing documents for signature and filing the signed copy
- Intake, triage, and status tracking
What still needs people
- Judgment on risk, strategy, and anything outside the playbook
- Relationships with clients, colleagues, and counterparties
- Complex or unusual matters with no precedent to follow
- Quality control: checking AI output before it matters
For in-house teams, the question often looks different. Many have never had a paralegal. Their lawyers handle the admin work themselves, or it doesn't get done. In those teams, an AI paralegal covers work nobody was doing rather than taking anyone's job.
For paralegals, the skills that hold value are the ones AI can't supply alone. They include supervising AI output, designing processes and playbooks, and spotting when a matter needs a lawyer.
Ethics, supervision, and confidentiality
Using AI doesn't change a lawyer's professional duties. In the US, ABA Formal Opinion 512 (July 2024) sets out how existing rules apply to generative AI. Lawyers must understand the tools they use (Rule 1.1) and protect client information (Rule 1.6). They must keep clients informed (Rule 1.4), charge reasonable fees (Rule 1.5), and supervise AI-assisted work (Rules 5.1 and 5.3). In England and Wales, solicitors' existing duties under the SRA Standards and Regulations apply in the same way.
The risk is real. In Mata v. Avianca (2023), a New York court fined two lawyers and their firm $5,000 after they filed a brief citing cases ChatGPT had invented. Nobody had checked the output.
For an AI paralegal, this means three things in practice:
- Supervision is built in. Lawyers set the playbook, review what needs review, and handle exceptions.
- Output is checked against sources. Every draft, redline, and summary should trace back to the documents it came from.
- Confidential data is protected. Ask how data is encrypted, whether any AI provider retains it, who can access it, and whether you can have it deleted.
What it can access, and what it can't
Before you connect an AI paralegal to your systems, check how access works.
A good one:
- Works only in the inboxes, Slack workspaces, folders, and tools you connect
- Lets you decide what needs sign-off first, such as sending an email or a signature request
- Never sees your passwords: saved logins and keys are encrypted and hidden from the AI
- Doesn't overwrite your files: its versions sit beside yours
Alaro works this way by default.
Signs your team needs an AI paralegal
In-house teams are stretched thinner every year. The median in-house lawyer now supports 367 employees, up 26% in a year (ACC 2026 benchmarking, cited by Legora). Only 32% of legal departments expect to hire more lawyers (CLOC). An AI paralegal fits if several of these sound familiar:
- Every request routes through one or two lawyers, and you've become the bottleneck.
- Headcount isn't growing, and the answer from above is "use more tech."
- You pay law-firm rates, document by document, for routine work.
- Your AI tools produce drafts someone still checks line by line. That's the verification tax.
- Requests arrive everywhere, from Slack threads to forwarded emails, and some get lost.
- Nobody owns the tail of the process: approvals go unanswered and signatures aren't chased.
- Your team has no time to learn another tool.
When it isn't the answer
- One-off, high-stakes advice. That stays with your lawyers or outside counsel.
- Litigation and legal research, where AI legal assistants are the better fit.
- Teams that want lawyers using AI themselves, across drafting and research, with time to build and maintain workflows.
What an AI paralegal costs compared with a hire
Compare an AI paralegal with what the same work costs today, not with the price of a drafting tool.
A paralegal's salary is only part of the cost. The median US paralegal earned $62,890 in May 2025. Benefits make up about 30% of what US private employers spend on pay and benefits combined. Together, that puts one paralegal at roughly $90,000 a year.
That's before recruiting, training, management time, and cover for holidays and leave.
Many in-house teams never had a paralegal. For them, the comparison is outside counsel, billed by the hour or by the document, or lawyer time spent on admin instead of advice.
When you compare quotes, ask what's included, such as playbook setup, lawyer oversight, and the chasing, and what happens as your volume grows.
Your first 30 days with an AI paralegal
Scope it like a hire, then widen its remit as it earns trust. A sensible first month looks like this:
- Week 1: start sending work. Forward your first requests on day one. Share your templates, playbooks, past signed contracts, and approval rules as you go, and agree who can instruct it.
- Weeks 2 and 3: start narrow. Begin with one document type, such as NDAs, from one or two teams.
- Week 4: review and widen. Look at what was escalated and why, update the playbook, and add the next document type or team.
Track a few numbers from the first week:
- Time to first reply
- Time to a reviewed contract or first redline
- How many matters stayed inside the playbook, and how many were escalated
- How many requests were closed: signed, filed, or answered
How to evaluate an AI paralegal
Ask any provider these eight questions:
- Does it finish the work, or hand back a draft? Ask it to show you a request taken all the way to a signed, filed document.
- Who checks the output, and when? Find out which work lawyers review, which is done to pre-agreed positions, and how exceptions reach you.
- Where does it work? If your team has to log into something new, adoption becomes your job.
- What can it do in your systems? Find out which systems it can read and act in (CRM, e-signature, document storage), and what you control.
- Who maintains the playbooks and workflows? If your team does, you've bought a tool, not a paralegal.
- Who is accountable when it's wrong? Ask about insurance, regulation, and how errors are handled.
- How is it priced? Ask what's included and what happens when your volume grows.
- What proof is there? Ask for a named customer, the work it handles, and results you can check.
An example: Alaro
Alaro is the AI paralegal for your whole business. Teams delegate repetitive legal work to it by email, Slack, or Teams, and it comes back done. It drafts documents, proposes redlines, coordinates stakeholders, and chases signatures. Ex-Am Law and Magic Circle lawyers teach, monitor, and improve it.
Lawyers set the playbook with your team, and anything outside it comes back to you. Alaro maintains the playbooks and workflows, so your team doesn't have to. You step in only for the judgment calls.
Dominus Real Estate, a £2.4 billion family-owned London developer and operator of hotels and student housing, began with Alaro in January 2026 with two people and three contracts, one of them a six-figure deal.
By May, 27 people across seven of Dominus's eight departments were instructing Alaro directly, the CEO and CFO among them. Across 49 matters and more than 100 deliverables, opposite firms like Freshfields and Herbert Smith Freehills, Dominus has flagged not one error. Every matter is worked to Dominus's own playbooks and monitored by Alaro's lawyers.
"It's clear this is the start of something special."
Read the Dominus case study or see what Alaro is.
FAQ
What is an AI paralegal?
An AI paralegal is an AI system that does recurring legal work end to end, the way a paralegal would: intake, drafting, chasing, signatures, and tracking. Lawyers set the rules and step in for judgment calls.
What is the difference between an AI paralegal and an AI legal assistant?
An AI legal assistant, such as Harvey, Legora, or Claude, helps a lawyer draft, research, and summarize faster. The lawyer still does the work and checks it. An AI paralegal does the work itself, from request to completion, and brings lawyers in only for judgment calls.
Will AI replace paralegals?
AI is taking on much of the repetitive process work paralegals have traditionally done: chasing, tracking, first drafts, and routing for signature. Judgment, relationships, and complex matters still need people, and US paralegal employment is projected to stay roughly flat to 2035. For in-house teams that never had a paralegal, an AI paralegal often covers work nobody was doing.
Is an AI paralegal the same as a virtual paralegal?
No. A virtual paralegal is a person working remotely, usually through an agency, task by task. An AI paralegal is an AI system that works inside your inbox and chat tools, with lawyers overseeing it.
What tasks can an AI paralegal do?
Intake and triage, first drafts from precedent, redlines within an agreed playbook, chasing approvals and signatories, running documents to signature, filing, and tracking every request's status.
Can an AI paralegal negotiate contracts?
Within limits your lawyers set. A good AI paralegal proposes redlines and negotiation strategies inside your agreed playbook, and escalates anything outside it.
Who is responsible if an AI paralegal makes a mistake?
Your team remains responsible for its legal decisions, as it would be with any paralegal or outside provider. Ask each provider how work is reviewed, how errors are caught and fixed, and what insurance and regulation stand behind it.
Is an AI paralegal safe for confidential documents?
It depends on the provider. Ask how data is encrypted, whether AI providers retain any of it, who can access it, and whether you can have it deleted on request.
How much does an AI paralegal cost?
Pricing models vary, so ask what's included and what happens as your volume grows. Compare it with what the same work costs today in outside-counsel fees or lawyer time, not with the price of a drafting tool. For reference, the median US paralegal salary was $62,890 in May 2025, before benefits and management time.
Is an AI paralegal cheaper than hiring a paralegal?
It depends on your paralegal. A US paralegal costs roughly $90,000 a year once benefits are included, before recruiting and management. Compare that, or your outside-counsel spend on the same work, with the AI paralegal's quote.
What can an AI paralegal do on its own, and what needs approval?
It can take in requests, draft, redline within your playbook, chase people, and track status on its own. You decide what needs sign-off first, such as sending emails and signature requests, and anything outside the playbook goes to a lawyer.
How long does it take to set up an AI paralegal?
You can send the first request on day one by forwarding an email, without connecting any systems. Sharing your templates, playbooks, and approval rules makes the work match your standards. A good start is one document type, such as NDAs.
Does an AI paralegal work in Slack, Teams, and Outlook?
A good one does. Your team sends requests the way it already does, by email, Slack, or Microsoft Teams, and gets the work back in the same place. Alaro works in all three, with nothing new to log into.
Which companies offer an AI paralegal for in-house legal teams?
Alaro is built as an AI paralegal for in-house teams. Tools such as Harvey, Legora, CoCounsel, and Spellbook are AI legal assistants: they help lawyers work faster rather than doing the work end to end.
