Abel learns about AI4J at a conference. He sees a demo of a bot that can reduce intake time from 2 hours to 15 minutes.

Abel runs a 15-person organization on a shoestring budget. Spends time on fundraising, case oversight, policy advocacy. Has zero AI/ML expertise. Has one IT person who keeps the website running. His agencies serve 2,000 cases per year and turn away 10,000 more.

PEOPLE

Sven is AI4J's Implementation and Partnerships Lead.  Skills: Project management, justice sector knowledge, technical fluency

Abel is director of a legal aid NGO in Ghana. He is trained as a lawyer. Skills: management, fundraising, and government relations.
 

He imagines 'LegalMate Ghana' helping people resolve basic issues and triaging cases for his team.

But when Abel returns to Kumasi the reality of 15 paralegals in cramped offices makes "LegalMate Ghana" seem like a bit of magical thinking.

Undeterred, he goes online and opens up a project with AI4J.

The platform requires more than just registration. It turns out that IT uses an AI bot to handle intake. The AI4J bot leads Abel through an almost hour long initial assessment. What's the idea? What is happening now? What kind of an organization? And on and on.

Abel iterates with the AI4J bot and exchanges emails with AI4J staff. The bot wants to know a lot about their process, their successes and failures, staff capacity, bottlenecks and hurdles they encounter in their work. It asks Abel to describe the app he has in mind and offers pushback that helps him refine the concept.

The AI4J bot synthesizes a work plan and advises Abel about the kinds of permissions and such that he'll need to secure in order to get the project started.

The bot suggests that version 0 of the app will use these plug and play AI4J modules

• APPFrame
• UserReg
• ListenLearn
• Report2Case
• Language (Asante Twi)

These will arrive with basic customization for language and branding. Bot generates concept materials for feedback

Abel runs workshop with concept materials and records feedback from his team. The feedback is uploaded so the bot can make a few last minute adjustments before releasing version 0.0 and creating a build and test plan and schedule.

Step one of the build is to upload Ghanaian legal codes and local rules and document standards and legal aid eligibility requirements.

Since this is the first AI4J project in Ghana Abel uses the GetNationalLaw module to carry out the process.

He scans the document their office uses on local rules.

Version 0.1 is ready. It speaks the local language, knows Ghanaian Law, and is familiar with local courts and government agency document and form rules.

It's time for evaluation.

The FormFillerOuter eval generates narratives based on intake forms created by paralegals and feeds these to our intake bot and scores it on how well it reproduces originals.

The AI4J platform facilitates Abel's testing of version 0.1 with his coworkers.

Abel and his team spend a lot of time filling out ethics paperwork to allow for testing of the new app with actual legal aid users. The AI4J bot is available for consultation during this process. It's been through this many times before in other jurisdictions.

They meet with officials at the ministry of justice.

Finally, they get approval to set up a test version of the Ghana Justice Node.

Extensive testing with in-office clients leads to many user-experience tweaks.

One year later, LegalMate Ghana has increased access to legal aid by almost 180%

They meet with officials at the ministry of justice.

Finally, they get approval to set up a test version of the Ghana Justice Node.

Extensive testing with in-office clients leads to many user-experience tweaks.

One year later, LegalMate Ghana has increased access to legal aid by almost 180%

Abel learns about AI4J at a conference. He sees a demo of a bot that can reduce intake time from 2 hours to 15 minutes.

Abel runs a 15-person organization on a shoestring budget. Spends time on fundraising, case oversight, policy advocacy. Has zero AI/ML expertise. Has one IT person who keeps the website running. His agencies serve 2,000 cases per year and turn away 10,000 more.

PEOPLE

Sven is AI4J's Implementation and Partnerships Lead.  Skills: Project management, justice sector knowledge, technical fluency

Abel is director of a legal aid NGO in Ghana. He is trained as a lawyer. Skills: management, fundraising, and government relations.
 

He imagines 'LegalMate Ghana' helping people resolve basic issues and triaging cases for his team.

But when Abel returns to Kumasi the reality of 15 paralegals in cramped offices makes "LegalMate Ghana" seem like a bit of magical thinking.

Undeterred, he goes online and opens up a project with AI4J.

The platform requires more than just registration. It turns out that IT uses an AI bot to handle intake. The AI4J bot leads Abel through an almost hour long initial assessment. What's the idea? What is happening now? What kind of an organization? And on and on.

At AI4J

The bot flags this as a strong fit and connects Abel with Maya, an AI4J Implementation Lead who's worked with 3 other legal aid orgs in East Africa.

Partnership Formation (2 months)
 

xxx

Barriers.

Abel's idea will never happen because...

...he can't even locate the idea in the landscape of the possible

What can AI realistically do? Should one build something or buy something? Is this an AI project?

...person-power needed to begin to research does not exist

Everyone in the organization works too much already. Any new idea that takes more than 4 hours for a first look, won't happen.

...his board and ministry won't approve something this undefined

Past experience tells Abel that getting ministry buy-in for anything new is challenging at best. The organization does not have the staff or expertise or resources to mount a lobbying effort.

...the organization lacks regulatory/ethical confidence to proceed

The buzz around AI is all about privacy and regulatory compliance. Abel's organization has no experience or expertise. The impulse is to steer clear of these dangers.

...minimum necessary human capital is nowhere to be found

He has one technical person who manages the website.

...he can't even locate the idea in the landscape of the possible

...person-power needed to begin to research does not exist

...his board and ministry won't approve something this undefined

...the organization lacks regulatory/ ethical confidence to proceed

...minimum necessary human capital is nowhere to be found

At AI4J

The bot flags this as a strong fit and connects Abel with Maya, an AI4J Implementation Lead who's worked with 3 other legal aid orgs in East Africa.

Partnership Formation (2 months)

Over the course of two months AI for Justice Implementation Lead meets Legal Aid Ghana Director multiple times in person and over Zoom.
Platform explains what's possible, what AI4J can contribute to the process, what's required on the partner side.

Legal Aid Ghana commits their Director (0.3 FTE), their head paralegal (0.2 FTE), and a domain expert (0.2 FTE) for two month co-design process.

They sign MOU, agree to pilot scope. 

The AI4J Platform digests the preliminary conversations,

Finally, they get approval to set up a test version of the Ghana Justice Node.

Extensive testing with in-office clients leads to many user-experience tweaks.

One year later, LegalMate Ghana has increased access to legal aid by almost 180%

They cannot "fine-tune models" or "set up RAG infrastructure." 

Let's take a design approach...

Think Like a Designer

Let's think like designers for a moment.

What is the justice problem? What are stakeholders trying to do?

Let's consider two stakeholders: people and SMEs with legal needs and governments and institutions that want to meet these needs.

What are their struggles?

The Status Quo

Before we address that, perhaps we should ask "how does the status quo justice production system work?"

Should we add a middle
column for the NGO, justice innovation  organization or can that be subsumed by category on the right for now?

Legal Aid NGOs

Personal networks

Self-help Guides

Informal Mediation

Digital Document Filing

Pro bono Legal Services

Fee-based Legal Services

e-Gov Digital Forms, etc.

Non-localized Internet Information

What status quo justice infrastructure is available to meet these needs?

Legal Tech Apps (including chatbots)

Generic LLMs

Digital Document Filing

Static Government Web Sites

These don't scale.

Informal Mediation

What if there were tools...

Really clever time/place bespoke legal tech.

Vision 1: A few giant corporations and well funded entrepreneurs vacuum up all the justice wisdom and monopolize global justice service provisions. Countries have limited control and become platform dependent.

Vision 2: The governments and institutions and justice NGOs build tools that fulfill their access to justice aspirations to reduce the time and cost of dispute resolution; strengthen small-business economies; rebuild trust in justice institutions; and reduce conflict.

But vision 2 only happens
if we can build an infrastructure that
empowers justice innovators to build on one
another's efforts to that innovation, learning,
and quality improvements compound across
countries, without centralising control or
creating platform dependency

PEOPLE

Have I been wronged?

What can I do?

How do I proceed?

How do I find affordable help?

How do I efficiently and effectively provide help?

How do institutions make themselves accessible?

How do institutions make themselves legible?

How do institutions make themselves usable?

What are the odds? How do I decide if it is worth the effort?

STRUGGLE TO

that want to provide justice

disseminate information

GOVERNMENTS and INSTITUTIONS

innovate processes

handle caseloads

supplement legal services

be accessible

Sven

AI4J's Implementa-
tion and Partnerships Lead.

Skills: project management, justice sector knowledge, technical fluency.

Chioma

Business Development Officer, Lagos State Chamber of Commerce. Trained as a lawyer. Skills: SME advocacy, government relations, business formalization. Has spent years trying to improve legal access for Chamber members. Has one part-time IT contractor. No AI/ML expertise.

Abel Mensah posts on LinkedIn about LegalMate Ghana a tool his NGO created with AI4J. "In one year we increased access to justice by 180% with AI. #scalingJustice-AI4J.

Chioma reads about LegalMate Ghana on LinkedIn — an AI intake tool built by a Ghanaian legal aid NGO. She messages Abel. He replies the same day and connects her with Sven at AI4J.

In a call with Sven, Chioma explains her frustration. The Chamber serves 4,000 member businesses. Most stay informal because CAC registration, tax IDs, and sector permits feel impossibly complex and expensive. Many also fear hiring staff because they don't understand their obligations under Nigerian labour law.

Sven gets Chioma started with AI4J. Then the onboarding bot asks about the Chamber's current services, member profile, biggest obstacles, and what success would look like in two years. Chioma finds herself energised.

Chioma iterates with the AI4J bot. It elicits the user journey.  Where do businesses stall? What do they ask? What do staff spend hours explaining? It pushes back on her initial vision of a full legal platform and recommends starting with business formalization.

The AI4J platform generates a work plan. It advises Chioma on the data permissions, bar association notifications, and government data-sharing agreements she'll need to get the project started.

The platform generates plans for Version 0.0 - the basic user experience - using these plug-and-play AI4J modules:
  o AppFrame
  o UserReg
  o ListenLearn
  o Guide2Form
  o Language (Yoruba, Igbo, Pidgin)

AI4J generates Chamber branded concept materials to help the team generate feedback.

Chioma runs a workshop with Chamber staff using the concept materials. Feedback is uploaded to help the platform refine the design before creating Version 0.0 and generating a build and test plan.

With Version 0.0 they assess functionality and user experience to ensure the product they are building will achieve the desired outcomes.

                                      The AI4J platform identifies the legal content
                                          the system will require from the Nigeria National
                                          Justice Hub set up when AI4J started working
                                          with Nigeria.

                                          The build plan shows the team how to access the Hub.
                                          They find: Nigerian
                                          commercial law, CAC
                                          registration procedures,
                                          FIRS tax ID require-
                                          ments, and Lagos State
sector permit schedules.  Chioma uses the
GetNationalLaw module to help locate and structure
additional documents.
 

She also uploads informal guides her staff have built
up over years - institutional knowledge that has never
been consolidated before.

Version 0.1 - it combines the basic user experience with real laws and rules - is ready. It's time for evaluation. The FormCheck eval generates synthetic SME narratives, runs them through BizReady Nigeria, and scores the tool on whether it correctly identifies the right registrations and generates accurate guidance documents.

The AI4J platform facilitates Chioma's testing of Version 0.1.

Crucially, AI4J identifies three local business lawyers to invite into the project as backstop reviewers — they contribute to testing and see how the tool will amplify their capacity, not replace them.

The three partner lawyers review the tool's guidance documents for accuracy and flag anything that needs professional clarification — building a review and escalation pathway into the product from the start.

Chioma and her team spend considerable time working through ethics documentation — data privacy under NDPR, bias review, escalation protocols, and formal notification to the Nigerian Bar Association. The AI4J bot guides this process. It has navigated bar association relationships in multiple other jurisdictions and knows the standard concerns and how to address them constructively.

The Chamber meets with officials at the Corporate Affairs Commission and the Ministry of Trade. They present BizReady Nigeria not as a replacement for official processes but as a guided onramp that prepares businesses to engage with those processes correctly the first time — reducing the burden on government offices as well as on businesses.

They receive approval to launch a pilot as the Nigeria Business Formalization Node. Sven is present for the signing. Abel sends a congratulations message from Accra. The approval acknowledges that the tool will be monitored, with lawyer review built into every guidance document generated.

Extensive testing with real Chamber members leads to dozens of user-experience refinements. Business owners who had given up on formalization come back and complete the process. The labour law guidance module is added in Version 0.2 — helping owners understand exactly what hiring one employee actually means for them, removing the fear that had kept many businesses deliberately small.

One year later: the Lagos Chamber has tripled its annual business formalization rate. Member businesses report growing with more confidence. Now that the Nigeria Commercial Justice Node is built — on Nigerian law, CAC procedures, and local business realities — other state Chambers can deploy their own version of BizReady without starting from scratch.

Each participating country operates its own Justice Node, a nationally governed instance of the AI for Justice platform that embeds domestic law, institutional procedures, language, and data protection requirements.

Setting up a Justice Node

The first project in a country requires the creation of a national justice node.

Justice Node in a Box

AI4J provides a work plan for creating a justice node. This includes:

• Preliminary web presence temporarily hosted by AI4J
• Admin site: local participants work with AI4J to begin to populate system with domestic law, institutional procedures, language, and data-protection requirements (no data per se - case data, etc. does not cross borders).
• Framework for preliminary governance and development staffing

Infrastructure 0.0

AI4J and local champion collaborate to set up Justice Node infrastructure. This includes:

• Multisectoral governance
• Project Management
• Technical team
• Hosting capacity

Prototype 0.0

In conjunction with AI4J local champion selects demonstration use case such as:

• Information and referral chatbot
• Smart case tracking
• DIY dispute resolution in a specific area (e.g., land rights)

Using preliminary customization data and AI4J near-plug-and-play components the system generates a demonstration prototype and ancillary information materials.

Setting up a Justice Node

The first project in a country requires the creation of a national justice node.

Justice Node in a Box

AI4J provides a work plan for creating a justice node. This includes:

• Preliminary web presence temporarily hosted by AI4J
• Admin site: local participants work with AI4J to begin to populate system with domestic law, institutional procedures, language, and data-protection requirements (no data per se - case data, etc. does not cross borders).
• Framework for preliminary governance and development staffing

Infrastructure 0.0

AI4J and local champion collaborate to set up Justice Node infrastructure. This includes:

• Multisectoral governance
• Project Management
• Technical team
• Hosting capacity

Prototype 0.0

In conjunction with AI4J local champion selects demonstration use case such as:

• Information and referral chatbot
• Smart case tracking
• DIY dispute resolution in a specific area (e.g., land rights)

Using preliminary customization data and AI4J near-plug-and-play components the system generates a demonstration prototype and ancillary information materials.

BLAH

blah

BLAH

blah

BLAH

blah

Project 0.0

In parallel with creation of a Justice Node AI for Justice works with local partner on a pilot app that serves as proof-of-concept for the low-overhead production of a localized justice application.

Action: Select jurisdiction="NY state housing court" and useCase="Pro Se litigant assistant"

Platform: loads "Standards Pack" (SP) for this Justice Node. The SP knows about reqired disclaimer language, thresholds for advice vs. info, etc.

Output: Custom test protocol.

Dashboard: "0/15 standards passed"

Action: Bombard app with semantic testing.

Hallucination test: Queries reference non-existent laws, etc.
Adversarial red teaming. Attacker tries to get bot to ignore its safety rules.
Reading level audit to ensure responses meet 5th grade std required for accessibility.

Output: "Friction log" detailing prompts that break the app.

Action: Run 500+ gold standard legal scenarios based on real world cases from this jurisdiction. System response compared to "gold" answer stored in Justice Node.

Output: "Legal accuracy log" flagging errors of different severities in different categories (e.g., Disability Rights, Liability Assignment).

AI4J validation does not just rely on AI to police AI
Action: Sample of bot conversations routed to verified human reviewers.

Output: Qualitative badges for empathy, risk escalation, etc.

Value Add: AI4J protocol makes organizing human app review low overhead, efficient, and useful.

Action: App builder receives a "Gap analysis report." Not "app is biased" but "app fails 40% non-English queries; response weakness on 2025 tenants' rights updates. Recommendations: revise system prompt to prioritize Simple English libraries; audit and update statue content.

Milestone: After revisions and retesting Certificate of Submission Readiness.

THE UPL Minefield

A week before launch the bar association delivers a cease and desist letter.

How AI4J Helps. Platform generates "Supervision Memo" proving no "advice" given without human review (log of every attorney review this month!) and/or bot meets "self help" criteria. "Jurisdiction Toggle" tailors these to different local requirements.

Bias Backlash from Watchdogs

Tenant advocacy group posts on social media that the tool is biased and hallucinates.

AI4J produced apps include a public/auditor facing dashboard documenting actual rates of, say, proper handling of Spanish queries and case law hallucination. Turn the critics into auditors.

To get government contract we need liability insurance. Insurers don't know how to price the risk.

AI4J SaaS includes "Generate Insurer Pack" that compiles stress-test data in language tailored for insurance company mindset.

Each jurisdiction has different sandbox or registration requirements. Understanding how to file is daunting.

AI4J Justice Nodes maintain database of regulatory requirements. Choose "Deploy to Utah" and receive forms already 80% filled in for completion, review, and submission.  
Compliance feels more like a software dependency.

Problem: rules and laws change; yesterday's disclaimer might be insufficient today.  

AI4J Justice Nodes push "Disclaimer Updates" to apps. If changes in actual app behavior are needed, the deploying organization is notified.

BLAH

blah

BLAH

blah

BLAH

blah

Blah

blah

What if we used
ChatGPT app as a tool that can draw on a number of legal/justice tools.

Companies like Bubble.io 
has nocode app development at least partly figured out. Why not imitate or partner
with them?  

We need to think
about the 'last mile' problem. Médecins Sans Frontières solved this for healthcare with community health workers. Could we create a similar paralegal network that uses AI as a force multiplier?

We should look at how
Khan Academy used AI tutoring - couldn't we adapt that conversational scaffolding approach for legal guidance? People need step-by-step support, not just answers.

App framework would have to come with localization suite, evals, country-specific compliance, and...

What's Missing?

What would it take to deploy

Version 0 to include 2-3 app types

Articulates Need

Imagines Solution

Inspires

Iterative Sketching

IDEA 0.0

MockUp 0.9

Product Testing Suite

Plan & Build

MVP 1.0

Certified components, vetted content

Regulatory Landscape Scan

Deployment

Assurance Choreography

Regulatory Landscape Advice, Assistance, Chaperoning

Deep Supply of "if only there were a way to..."

Crushed by unmet needs

Hears about AI4J at conference

Visits Website

Generates "Design Conversation Guide" and works through with colleagues

Preliminary "Use Case Discovery" session with AI4J Bot

Bot brokers use case discussion with AI4J staff.

AI4J "Make It Happen" Project Manager assigned

BOT+PM+Abel create testing and registration plan

Platform generates workshop plan for org

Abel uses workshop results to create initial app MVP

Platform evaluates app and advises on necessary approvals

Platform+PM advise on funding and technology

System needs change

Can't push staff any harder

Lots of ideas

Abel is ED of legal aid organization in Ghana. Lawyer and entrepreneur and activist

What Abel Sees

What Abel Feels

What Abel Has

What Abel Lacks

Time to develop or explore ideas.

IT capacity

Technology Background

Knowledge of local system

network of local partners