Categories
AI + Access to Justice Current Projects

New Tech-Powered Ways to Get People Legal Help

AI changes what tools legal teams can build — and it unlocks new ways to deliver services that are accessible and scalable

Most of the conversation about AI in access to justice is about new tools. A new eviction answer generator, a new expungement eligibility checker, a new intake chatbot. It makes sense to start with an idea for a discrete new thing that tech could do for a consumer or a legal team.

But the more interesting change is one level zoomed out. If we had tech that could accurately, safely do different tasks (reviewing files, conducting interviews, spotting mismatches, looking up databases, etc.), what are the new ways that legal teams can operate services? Can we rethink how we get help to more people, at higher quality and consistency, with more impact?

AI makes new service models possible — new shapes of how legal help reaches people. This essay inventories new service models that are not the typical two main models: full-service representation and brief counsel/advice sessions.

Our status quo of legal help service models

These 2 dominant service models are not going away, but we can imagine other ways of delivering help that might be used in addition, that can reach a greater scale of need — and that can get to better legal and life outcomes.

A little more about the dominant status quo service models:

  • Hotline triage handles a five-to-ten minute call or chat session, mostly screening and referral. Sometimes this extends a little further with brief answers and explanations of a case, or pointing out the right form, document, or next step.
  • Brief services with limited counsel & advice are 45 minutes — 2 hours of time with a legal team, often producing a discrete, templated work product like a letter or a filing to move the case forward.
  • Limited scope goes a little further, with the team helping the person with a specific, more complex task — one motion, one hearing, one filing — over 3–10 hours.
  • Full representation is the whole case, 40–200 hours. Here is what most people think about ‘having a lawyer’ — the legal team is doing all the work product creation, back-and-forths with the other side and court, presentations and negotiations, and strategy.

This piece puts an array of new models forward. They all depend on some amount of technology being used (whether that is rules-based, AI-driven, or more often a hybrid of both) by legal teams.

Some of these service models are already taking shape in pilots. Others are more ideas than reality. All of them are within reach in the next 2 years if the field decides to build them.

Each new model is paired with a fictional story of how it might run in practice. The organizations and counties are fictional, meant to help envision this new way of providing help.

1. The Clinic at Scale: Mass Intake Before, Deep Consults During

A person calls or texts the legal aid line. Over the next 24–72 hours, a tech system conducts an extended intake. It interviews them across multiple sessions and modalities, bringing in their descriptions, texted-in photos, and forwarded documents. The system goes into databases to pull their court records and other key data about their situation. It screens these files, pulls structured data fields, flags issues, and populates a case management system with all the info. It identifies their landlord or creditor. It classifies the issue. Using all the information it’s gathered, the system drafts the initial answer or motion.

The person arrives at a scheduled clinic appointment a week later. Their drafted legal documents are ready. The attorney spends 20 minutes reviewing and customizing the documents with them, instead of two hours drafting from scratch. The person leaves with a filed pleading and an understanding of what the process will look like from here on out.

Clinic throughput goes from 10 people in 4 hours to 50–100 served. The constraint shifts from drafting capacity to attorney review capacity, which is where attorney judgment actually matters. The tech does all the document and preparatory work, and serves it on a platter for the legal team and person to build from.

This could be a good fit for eviction answers, motions to set aside default, expungement applications, fines and fees motions, habitability complaint packages, overpayment waiver requests.

What this might look like. Pine County Legal Aid (fictional) in central Minnesota serves a region of 80,000 people across five rural counties. Their team has 2 attorneys and 1 paralegal. This year they run a monthly eviction defense clinic that helps 12–15 tenants. Most months, they turn away another 50who can’t get scheduled before their answer deadline.

Starting in their new service model, Pine County’s intake line routes incoming eviction calls to a text-based intake system. Over the next 2 days, the system interviews the tenant over text message or a phone call if the person prefers, asks for the eviction notice, pulls the public court record from the court’s online portal, identifies the landlord’s pattern of filings, and drafts a tenant answer with defenses the system has detected — improper notice, habitability issues, retaliation, source-of-income discrimination.

When the tenant arrives at the Friday clinic, their drafted answer document is ready. The paralegal walks the person through it and confirms the information, corrects any issues. The attorney reviews and signs off. The tenant files before they leave, using an efiling connection that the team helps them with. The clinic now serves 70 tenants on a Friday instead of 15.

2. The courthouse intercept clinic

A volunteer legal team member sits at a table in the courthouse hallway near eviction court, traffic court, or family court. A person walks by who looks lost. The volunteer asks if they’re here for a case, confirms that they are. The volunteer pulls the case up on a tablet, and sees that the tech system has already (last night or the day before, when the docket was posted) drafted a case plan and draft documents for this person.

The volunteer calls over the attorney and asks the person if they would like legal assistance. The attorney gets the case review and plan from the system, and sits down with the person. The attorney goes over what’s happening in the case and what is likely to happen at the hearing today. The person can ask questions and correct any facts. The attorney can then work with the person on tailoring the document and case plan, to get them ready for what to present to the judge or other side. The attorney can sub in to represent the person during the hearing, and explain to them what happens afterward.

This model can be good for situations where people are showing up to a high-stakes, rapid-fire hearing like in evictions, court costs, warrant recall, motions to set aside, or traffic tickets — often without any prep or defense.

The intercept model catches them at the moment of highest leverage — when they’re physically at the courthouse and there’s still time to file or strategize. This model differs from other “Lawyer for A Day” or “Hallway Intercept” services, because it equips the lawyer to quickly get up to speed with the full case situation, if not also the pre-populated filing/script beforehand. It can help the team reach people who never sought help, at the moment when help still matters.

What this might look like. Mercer County (fictional), a midsize urban county in the Mid-Atlantic, sees roughly 400 eviction hearings a month. Most tenants show up unrepresented. The county’s legal aid office partners with the housing court to set up a table outside Courtroom 2A on hearing days.

A trained volunteer staffs the table. Tenants check in. The system had already run over the docket yesterday, and queued up case summaries and draft filings for each of them. The volunteer waits for people to approach, or goes up to people waiting in the hallway, to ask if they would like assistance. If they say yes, the volunteer pulls up the case files, asks four or five questions about the notice and the landlord, connects them with the attorney for a consult, and gets them set up with the documents. The tenant might bring the filing to the court window or walk into the hearing with the attorney prepped and ready to represent them in the courtroom.

In the first year, the program reaches 1800 tenants. Two-thirds successfully file an answer or raise defenses before the judge. Dismissals or negotiated settlement rates rise. \The legal aid office’s existing eviction clinic continues running for the more complex cases that need detailed follow-through.

3. The centralized expert hub-and-spoke model

What if there were 20 SSDI specialists in the country instead of needing one in every legal aid office? Or 5 Motion to Set Aside experts in the state of California, rather than 1 in every local group?

A frontline paralegal at a rural Nebraska legal aid office takes intake from a person denied benefits. They enter the case context into a system co-developed by the SSDI expert network. The tech system produces a draft hearing brief, a witness list, and a strategy memo aligned with the expert’s approach. The frontline lawyer reviews and executes. When something exceeds the system’s confidence threshold, the human expert (sitting not in Nebraska but in Wisconsin) is pulled in for live consultation.

One of the legal help field’s chronic problem is thin specialty expertise, especially for legal issues that are not as emergency or high priority. No single legal aid office can carry deep capability in federal benefits appeals, complex tenant protections, VA disability, immigration consequences of criminal records, or child welfare. Can we use field-wide expert pooling, especially when tech can help do some of the more mundane, templated parts of the work.

This could be a good fit for national or regional issues — that have standard rules and strategies across an entire region or country. This might include for SSI and SSDI ALJ prep, VA disability, subsidized housing termination appeals, post-conviction immigration consequences, special education due process.

What this might look like. The National SSDI Practice Hub (fictional) is a 6-person specialist team based at a national legal aid organization. They serve roughly 300 hundred frontline legal aid offices across 40 states.

A paralegal at Cedar Valley Legal Services (fictional) in southern Iowa takes intake from a forty-year-old former meatpacker whose SSDI was denied. The paralegal enters the case details into their regular case management system, which flags that the best track for the client is to bring the case into the National SSDI Practice Hub. The system leads the local Iowa legal team through doing what the expert would do — gathering medical records, work history, denial letter, the SSA’s stated reasons. Over the next 48 hours, the system produces a draft hearing brief, a list of medical evidence to request, a witness list, and a strategy memo. The Iowa attorney spends an hour reviewing and customizing the drafts, and can chat with the Hub’s expert chatbot or live chat with any questions. The hearing is set for 6 weeks out.

When the case has an unusual wrinkle — a prior denial under a different impairment, an evidentiary question the system flags — the Iowa attorney requests a Hub specialist to join a 15-minute video call to advise. The Iowa attorney handles the hearing. The Hub specialist never did detailed work with the client, but makes sure the local team is empowered to do this work. The case wins on the medical evidence the system identified.

Across the Hub’s network, 4,000 SSDI cases are supported in the first year. Six specialists support teams at 300 frontline offices. They acheive a win rate 15 points higher than the national average for unrepresented SSDI appeals.

4. The asynchronous case management model

In this new service model, the person who’s experiencing a legal problem enrolls in a legal team’s case management program. Over the next 6 weeks, the tech system texts them with questions, requests for documents, case status updates, and reminders about upcoming deadlines or events. The system presents them with the documents they need to fill in, and helps them draft each filing as the case progresses — the initial motion, the response to discovery, the trial brief. It also lets the person request the legal team to do an expert human review before filing (or this might be the default). The attorney spends an hour a week per case instead of ten. The person gets continuous support instead of a single 45 minute brief services call.

The current binary between brief services and full representation leaves out the people who need ongoing engagement but not the most intensive intervention. This middle layer doesn’t exist today, but tech might make it economically viable.

This could be the missing middle of legal help services: the cases that need someone walking alongside with them for weeks, not the one-shot motion and not the all-in litigation defense. Tech might make this shape sustainable for the first time.

The automated, walk-along case management service might be fit for ongoing justice journeys like family law modifications, debt collection defense, complex expungement, ongoing benefits cases with multiple appeals, reentry navigation across many simultaneous issues.

What this might look like. Lake County Family Law Project (fictional) serves a midsize county in northern California. They’ve always had a gap, between their brief services that helps a person one time with a single motion, and then full representation is reserved for the most complex domestic violence cases. The bulk of their callers — parents trying to modify custody, respond to a contempt motion, get an order enforced — fall in between and don’t get the help they need.

The Project launches a 6-week, tech-powered case management program. A parent calls. They want to modify custody because their work schedule changed. The Project enrolls them into the system. Over the next 6 weeks, the system texts the parent twice a week. It asks for documentation about the situation. It drafts the modification motion for them to review. It drafts the income and expense declaration. It tracks the deadline for the other parent’s response. It drafts the reply when that response comes in.

The supervising attorney spends an hour a week on the case, especially focused on reviewing each filing before it goes out. The parent has continuous contact through text. They feel supported, not abandoned. The case resolves in 2 months with a stipulated modification. The attorney spent 6 hours total, not 60.

Subscribe to Pro Writer Tools to remove upgrade prompts from your unlocked stories.

In year one, the Project serves 400 modification cases this way, while their full-representation caseload is unchanged. Their brief services line is unchanged. The new middle layer is pure addition for this group of clients.

5. The point-of-service embedded model

A person applies for SNAP and the tech system in the benefits portal spots that they were recently denied SSI based on what looks like an overpayment dispute. It flags this and offers to help them appeal.

A person checks in at a community health clinic and the tech in the clinic’s social determinants screen picks up that they got an eviction notice last week, by calling over to the court database. It offers to help draft an answer right there.

A person applies for an apartment and the tech in the housing assistance application detects a sealed criminal record from a state where the lookback period has passed. It offers to help the landlord understand why the record shouldn’t disqualify them.

Most people with legal problems don’t know they have a legal problem, or they don’t know of the rights and relief that are available to them. That said, they may interact with social services, healthcare, schools, and benefits portals regularly. Putting tech-powered issue-spotting and tool generation at those touchpoints can reach people where they actually are — instead of waiting for them to come to a legal aid office.

This service model might be a best fit for SNAP and Medicaid issue detection, eviction notice detection in housing applications, expungement eligibility screening at workforce centers, or benefits appeal triggers in healthcare settings.

What this might look like. Riverside Community Health Network (fictional) operates 21 community health clinics in a major metropolitan area. Every patient who checks in completes a social determinants of health screen that asks about housing, food security, employment, and legal needs.

The Network partners with the regional legal aid organization to embed a tech-powered legal triage layer in the screen. When a patient indicates they got an eviction notice in the last month, the screen offers help. The patient consents. Over the next ten minutes in the waiting room, the system pulls from databases, and asks the person about the notice, the landlord, and the dates. It drafts a tenant answer, and offers a potential lawyer review before the person files it. It calculates the filing deadline. It offers to text the patient a reminder.

By the time the patient sees their doctor, they have a draft answer and a filing reminder. After the appointment, the patient files the document with the court (or the system might do it for them). The legal aid organization gets a notification that a new case is potentially queued up for their pipeline, with the materials already prepared.

In the first year, the Network’s 21 clinics flag 6000 legal issues across housing, benefits, and employment. Forty percent of those people would never have contacted a legal aid office on their own.

6. The community navigator augmented model

A 211 navigator takes a call and learns that the caller has been denied SNAP. The navigator opens a tech tool that walks them through the local SNAP appeal process, helps them draft the appeal letter, and produces a filing the caller can submit. The navigator never gave legal advice. The tech’s built-in expertise handled the legal substance. The navigator handled the human relationship and the system navigation. The caller gets a filed appeal in the same call.

There are many community workers in this country who already interact with people in moments of legal need — domestic violence advocates, social workers, community health workers, housing counselors, 211 staff. They’ve been blocked from providing legal help because of Unauthorized Practice of Law (UPL) rules and the limits of their own training. Legal tools designed for navigator use, with appropriate guardrails and clear scope, expand the workforce of legal help providers by an order of magnitude — without changing who is licensed.

This service model could fit very templated, formulaic cases, like benefits appeals, SNAP recertification, simple expungement applications, basic housing rights letters, fee waiver applications, simple form-driven matters.

What this might look like. Hilltop Community Action (fictional), a community action agency in rural Tennessee, has 6 legal navigators serving a region of 3 counties. They’ve always done warm referrals to legal aid when a person comes in asking about benefits problems, but the nearest legal aid office is 90 miles away and has a 6-month waitlist.

Hilltop partners with their state legal aid network to deploy a benefits appeal tool designed for the navigators to use. When the navigator encounters a caller with a benefits problem, the navigator opens the Benefits Appeals Tool. It walks both the navigator and the caller through a structured intake, getting key info from the caller and databases. It produces an appeal letter using an approved template, that the caller can review and sign. The system tells the navigator when the case is beyond the tool’s scope and needs escalation to legal aid.

The navigators don’t give legal advice to the callers. They guide people through a process that the legal aid organization designed and supervises. The legal aid attorneys review a sample of cases each month for quality.

In the first year, Hilltop’s 6 navigators file 1200 benefits appeals. Most resolve favorably without escalation. The legal aid network gets to focus its attorney time on the cases that actually require legal judgment.

7. The negotiation-and-resolution model

A person comes to the legal aid org with a debt collection summons and complaint, asking for help with debt defense. The tech system sends an initial response to the collection lawsuit and proposes a settlement based on the person’s ability to pay. It negotiates with opposing counsel over a few weeks. When the deal is acceptable, it’s brought to the human lawyer for review and explanation, and the person for sign-off. The case resolves without a hearing. The attorney spent 2 hours total.

Most civil cases settle. Settlement work can be time-consuming part of an attorney’s day, and it’s often where they can produce the most value for a client. Tech-assisted negotiation can compress what currently takes an attorney 10 hours into something more like a few hours of review and consultation.

These settlement service models could be a good fit for debt collection settlement, wage theft demand letters, security deposit recovery, small claims settlement, housing repair negotiations, eviction settlement or cash-for-keys agreements.

What this might look like. Sunbelt Consumer Justice Center (fictional) is a small nonprofit serving consumer debt cases across a sprawling Sun Belt state. There are 2 attorneys and 2 paralegals on the team. The state has a high volume of debt collection lawsuits. The Center has always been overwhelmed by the volume relative to capacity. It can’t help many people with settlement.

The Center deploys a negotiation tool that handles the routine back-and-forth with debt collectors. A client comes in with a collection lawsuit. The tool sends an initial answer raising statute of limitations and other defenses. It then opens a settlement conversation with opposing counsel through their established channel, proposing terms and explaining why the client can only afford this payment plan.

Over 6 weeks, the tool exchanges offers and counter-offers, calibrated to what the client can pay. Each significant move is queued for the supervising attorney’s review before going out. When the collector finally accepts a payment plan the client can manage, the tool produces the settlement agreement. The attorney reviews it, sits with the client to ensure they understand it, the client signs, and the case is dismissed.

In the first year, the Center resolves 900 debt cases this way. The attorneys went from spending 10 hours per settlement to 2. The settlements are, on average, lower-dollar than what attorneys negotiated under time pressure in prior years.

8. The court-led screening and compliance model

A property management company files an eviction against a tenant for non-payment of rent. Before the case is docketed, tech screening in the court’s case management system checks the notice for legal sufficiency. The notice has the wrong number of days. The case doesn’t proceed; the system flags it for rejection. The clerk sends the landlord a notice that the case requires a corrected filing.

A debt collector files a lawsuit against a person for an alleged medical debt. Before service of process, AI screening verifies the chain of assignment, checks the statute of limitations, and confirms that the alleged debt amount matches the documents in the complaint. The chain of assignment is broken. The case is flagged for the judge’s review. The judge dismisses it before the defendant has to do anything.

A foreclosure is filed against a homeowner. The tech system checks the documents for the indicia of robosigning, irregular assignments, and standing problems. Flagged cases get judicial review before going forward.

This model takes a different shape entirely. Instead of waiting for an unrepresented defendant to raise defenses they will never know to raise, the court does the work of screening for case legitimacy upfront. The court is enforcing the basic procedural rules that already exist but currently go unenforced because of a lack of human staff to do this thorough, intensive review. Tech gives courts the capacity to do the review that courts are supposed to do.

This model could fit eviction case screening, debt collection screening, foreclosure screening, family law procedural compliance, restraining order petitions, and small claims case screening.

What this might look like. Magnolia Circuit Court (fictional) is a midsize urban court system in the Southeast. The court hears roughly 12,000 eviction cases a year. Eighty percent of tenants are unrepresented. Eighty-five percent of cases result in default judgments. The court leadership has grown uneasy about what proportion of those defaults involve procedurally defective notices that no one ever raised.

The court partners with the state’s office of court administration and a public-interest technology partner to embed pre-docketing screening in the eviction case management system. Every landlord-filed eviction now passes through screening before the case is assigned a hearing date. The screening checks the notice for the right number of days, the right language, proper service, and statutory cause where required.

In the first year, roughly 15 percent of filed evictions are flagged as procedurally defective. The court rejects those filings and notifies the landlord to refile with a corrected notice. About a third never refile. The cases that do move forward are procedurally clean.

The court is doing the review and compliance work the court is already supposed to do. The judge is not making advocacy decisions. The screening helps the judicial team to enforce the rules that already exist. Default judgments fall. Procedural irregularities fall. Public confidence in the eviction process rises because the legal process is being followed.

9. The high-volume application factory

A statewide push starts to enroll eligible families in ACA, PSLF, EITC, SNAP, or VA disability. Can there be a statewide expungement clean-up to help with workforce development and second chances? A tech system handles much of the pipeline — eligibility screening, document gathering, application drafting, follow-up, status tracking. Human experts handle the edge cases and review of the system. The state moves from helping a few thousand people a year to a few hundred thousand.

Federal programs with uniform applications and high underclaim rates are a low-hanging fruit in the justice space. PSLF has millions of eligible borrowers and a tiny fraction enrolled. The ACA Marketplace has predictable enrollment gaps every year. EITC underclaiming costs eligible families billions. Most legal aid offices can’t run application drives at scale because of the per-person attorney time required to make this happen. Tech could make statewide enrollment campaigns feasible.

This could be a fit for highly templated statewide or national administrative proceedings, like PSLF, ACA enrollment, EITC outreach, SNAP enrollment, VA disability initial claims, state expungement clean slate initiatives, driver’s license reinstatement clean-up.

What this might look like. The state of Centralia (fictional) passes a Clean Slate law. Six hundred thousand residents become eligible for automatic expungement of certain misdemeanor records. The state department of corrections is mandated to process applications, but they have no infrastructure to do it at scale.

The state’s legal aid foundation funds a coalition of legal aid organizations to run the application pipeline. The coalition deploys a unified system. It pulls eligibility data from the state’s record systems. It identifies eligible individuals. It contacts them by mail, text, and email. It collects the documentation each person needs to confirm identity and current status, from database lookups and communications with the person. It drafts the petitions. It flags difficult cases to go to the expert team. It ensures the person can review the filing for correctness. Then it helps the person file.

Frontline attorneys handle the cases that the system flags as complex, like out-of-state offenses, identity questions, and contested records. Everything else can move through the pipeline with the person, supported by the tech and navigators.

In two years, the coalition processes 400,000 expungement petitions. The state’s record system is cleared for a population the field would have taken decades to reach one by one. Employment outcomes for the affected population improve measurably in the years that follow.

Which of these already exist, how can we see which new models can work?

Each type of legal case & user-type should get a service model that makes sense for them. Our field needs a better array of service models that make sense for:

  • the type of work that needs to happen for the average process needed to get people from problem to solution. This is different in different case types, and different users in these case types.
  • The regional resources, in terms of staffing, volunteers, agency partnerships, community partners, data connections, and funding

Some versions of these models already exist around the country — whether that’s centralized state support for some issue areas, or smarter point-of-service spotting of issues and helping with relief. One of the big changes with tech, though, is the ability to think about safe, strategic delegation more.

For a given workflow, where can expert attorneys better delegate to tech, to junior attorneys, to paralegals, to justice partners, or to people themselves? Once we start to rethink ‘who can do a given task to a sufficient quality standard’ — we can rearrange how expert, senior attorneys are spending their time.

How do we take more mundane, repetitive, templated tasks off their plate, and broaden the number of people who can do these? Also, how do we move away from the user/client having to find and manage all the information about their legal situation — or expect legal teams to spend enormous time researching the case, property, protections, situations, etc. Can we pull more of this data from authoritative records, without having to do lengthy back-and-forth interviews and searching of databases to get the key information needed to fill in forms, check off defenses, and prep strategies?

If you are interested in thinking through how any of these service models might be implemented, or if you’re already trying out new ways to get services to more people, more impactfully, please be in touch with us at the Legal Design Lab! As we envision a future set of technology tools for legal teams, we believe that the service design will be just as important as the tech tool’s performance in making a positive impact on people’s legal and life outcomes.

Categories
Resources

User Persona Template

Is your team working on a legal innovation project, using a human-centered design approach? Then you are likely focused on different kinds of ‘users’, ‘stakeholders’, or ‘audience members’ as you plan out your innovation.

Our Legal Design Lab team has a free Canva template to make your own user personas easily.

This Canva template gives blank & example user persona templates that your team can fill in with your interviewees’ and stakeholders’ details. We recommend making multiple personas, and then possibly also making a schematic comparing different kinds of users.

The template document includes example images drawn by Margaret Hagan, that you can use in making your persona.

Have fun & let us know how your user persona creation goes!

Categories
Ideabook Wayfinding and Space Design

Service Help Centers in airports, based in JFK Delta airport terminal

I have been scouting out service design inspirations, particularly from airports, that courts could use. This one is from JFK airport, in the Delta terminal. I was very impressed with their service design. They had taken over an entire gate with a help center that had all kinds of touchpoints: paper, phone terminals, people, kiosks. There were multiple entry points and lots of capacity, so that people could self-sort depending on the magnitude of their issue, the speed they needed it resolved, and their comfort with tech.

Of particular interest: the Need Help card. Can we make these for all courts?