2.36M
cases pending in Pakistan's courts (Dec 2024) — 83% in district judiciary
§1The Problem
A two-sided failure of the justice system
Pakistan runs a legal failure on both sides of the market. Citizens cannot afford, read, or navigate the law; lawyers work without the tooling to move faster. TVL is built to attack both at once.
Demand side · citizens
Justice is unaffordable, English-only, opaque
The median citizen can't fund a first consultation, can't read a statute in legal English, and doesn't know which of 5,325 statutes governs their tenancy, inheritance, FIR, cheque-dishonour or divorce problem. Their realistic options: ask a relative, Google it and get US/Indian law, or paste it into ChatGPT and get a fluent, confident, often-wrong answer with no Pakistani citation.
Supply side · lawyers
~250k advocates, researching by hand
Mostly solo or small-firm, they search case-law manually in a market where a single missed or fabricated citation is malpractice. There is no trusted, Urdu-and-English, citation-grounded research layer built for Pakistani law — which is exactly why a hallucinating general LLM is a liability, not a tool.
Rs 3T
tied up in tax disputes stuck in litigation
Rs 1.1T
non-performing loans frozen in the courts
83%
of the 2.36M backlog sits in the district judiciary — closest to ordinary people
§2The Product
A legal super-app with a trustworthy engine at its core
The core is the Lex-Engine — retrieval-and-reasoning over real Pakistani law. Around it sits an integrated platform the database already supports: find a lawyer, consult, learn, and manage your matter.
Lex-Engine — the defensible core
Grounded
Cites the section, or declines
Every verified answer carries the exact § 302 PPC-style section + statute + leading case. When unsure it abstains and routes to a human. In law, that is a liability control, not just a feature.
Urdu-first
Native Urdu, Roman-Urdu & English
Built for how Pakistanis actually ask — میری زمین پر قبضہ — not just legal English. The underserved, uncontested lane the field ignores.
Local & private
Runs on-box, near-zero cost
A local open model (qwen2.5:7b + bge-m3), not a metered API. Sensitive queries never leave the country; marginal cost per answer is effectively electricity.
Trust, measured
Held-out routing accuracy — % of unseen citizen questions hitting a correct verified answer
Built · 2,537 verified facts, 97% corpus-grounded Funded milestone · 12,000+ facts, published audit
The super-app around it
AAsk & Researchfree citizen Q&A · lawyer research with citations & draftingengine
BVerified Marketplacefind, book & pay a domain-matched, KYC-verified lawyernetwork
CConsult & Hirethe AI pre-scopes the problem; the lawyer picks up a qualified leadrevenue
DLearnstudy content & bar-prep for 50k+ law students — the future supplyfunnel
EWorkspacesaved matters, notes, tasks, tracked cases & documentsretention
§3Market & Opportunity
A large reach, a focused revenue pool
We size bottom-up — value pools per segment, not a top-down slice. FX ~PKR 285–295 / USD (2026). The "200M underserved" is a reach number; bankable near-term revenue is lawyers, enterprise and episodic B2C.
Sizing · bottom-up, annual
TAM → SAM → SOM
Segments — ranked by size × willingness-to-pay × reachability × strategic value
| Segment | Size (Pakistan) | Core need | Pays? | Role |
|---|---|---|---|---|
| Practising lawyers & small firms | ~250–300k advocates | Fast, cited research + case mgmt | Moderate–High | Profit engine · seeds supply |
| Corporate legal / banks / insurers | ~1–3k institutions | Bulk, private, grounded research | High | Anchor contracts |
| Individual citizens | ~140M smartphones | "What's my right / what do I do?" | Very low | Moat + funnel |
| Overseas Pakistanis | ~9M diaspora | Property, family, inheritance "back home" | Moderate–High (USD) | High-margin niche |
| Government / judiciary / legal-aid | Federal + 4 provinces | Backlog relief, citizen literacy | Grant / contract | Legitimacy · corpus |
| Law students / bar prep | ~50–80k | Study, exam prep | Low | Cheap future-lawyer funnel |
Why now
Local models crossed the line
Open models are finally good enough for Urdu-first, citation-grounded retrieval — and cheap enough to run locally, so a free national tier is economically survivable. Every API-dependent competitor bleeds per-token.
The rails are live
140M smartphones, 162M broadband subs, 84% of retail transactions digital, JazzCash (40–54M) / EasyPaisa (55M+) / Raast ready to collect. The infra-vs-usage gap is runway, not ceiling.
The pain is acute and worsening
A 2.36M-case backlog makes triage and lawyer efficiency structurally valuable to citizens, courts and the state — at once.
The Urdu lane is open
~38 legaltech startups, only ~4 funded, and almost all target lawyers in English. Nobody owns citizen-facing, Urdu-first, verified.
§4Business Model
Monetise the professional, subsidise the citizen
Near-zero marginal cost inverts the normal AI-startup constraint: TVL gives citizen Q&A away at national scale where API-native rivals cannot. The model is deliberately cross-subsidised — free where the crowd is, priced where the budget is.
01Free citizen Q&AFreemium acquisition — the funnel & the missionRs 0 · ~0 cost
02Pay-per-documentrent agreement, legal notice, affidavit, contract — instantRs 149–999 / doc
03TVL Plus (consumer sub)saved matters, unlimited docs, deeper research — the repeat 2–4%Rs 299–499 / mo
04Marketplace take-rateon pre-qualified, AI-scoped consultations12–18% of fee
05TVL Pro — lawyer SaaScited research + drafting; the anti-hallucination guaranteeRs 2,500–4,000 / seat
06TVL Firmseats + case-management + workspace lock-inRs 15k–60k / mo
07Enterprise / API / white-labelbanks, insurers, legaltechs — private, on-box, groundedRs 1.5M–12M / yr
08Institutional licensinglaw schools, bar councils, legal-aid, judiciaryRs 0.5M–8M / yr
09Education / bar-prepsticky student subscription — the supply funnelRs 199–399 / mo
Wedge · 0–9 mo
Win the professional, seed the marketplace
TVL Pro first (budget, acute pain, low CAC) + pay-per-document (fastest B2C cash) + launch free Q&A loudly for growth & the demand side.
Scale · 9–24 mo
Monetise the flywheel
Marketplace take-rate once both sides exist; TVL Plus for revealed power-users; Firm tier + bar-prep.
Expand · Yr 2+
High-ACV lines
Enterprise / API, institutional licensing — each needing the corpus, trust and scale that only accrue after the wedge lands.
§4.1Pricing
Priced for a low-WTP, cash-first market
B2C anchors inside a mobile-top-up mental budget (Rs 200–500/mo). Lawyer and enterprise tiers anchor to the value of an hour saved and a liability avoided.
CitizenFreeRs 0
- Unlimited grounded Q&A in Urdu / Roman / English
- Cited answers or a safe decline
- Directory of verified lawyers
Citizen · episodicPay-per-DocumentRs 149–999 / doc
- Rent agreement, legal notice, affidavit, contract
- 90–95% cheaper than a lawyer, instant
- No subscription commitment — fits the cash market
Advocate · the anchorTVL ProRs 2,500–4,000 / seat·mo
- Cited case-law & statute search, Urdu + English
- AI drafting with grounded citations
- Declines-when-unsure = no malpractice from a fake cite
- Workspace & saved matters
FirmTVL FirmRs 15k–60k / mo
- Multiple seats + case management
- Workflow lock-in, lower churn
Bank · insurer · corporateEnterprise / APIRs 1.5M–12M / yr
- Private, on-box deployment — data never leaves
- Grounded research over their own matter set
- White-label & API
School · bar · legal-aid · govInstitutionalRs 0.5M–8M / yr
- Bar-prep & study licences
- Legal-aid triage; judiciary backlog tools
- Turns the mission into recurring revenue
§5Unit Economics
The structural edge: near-zero marginal inference
The engine runs a local open model, not a metered API. Inference cost scales with fixed compute steps, not per query — which makes a free national tier survivable while API-native rivals bleed.
Cost per 10M answers / month
The moat competitors can't price against
But the honest cost base is people, not compute. What actually scales is human trust-work: lawyer curation, corpus re-verification as laws amend, forum moderation, marketplace dispute ops, Urdu support, and CAC. We model those as linear — and don't hide them behind the cheap inference number.
CAC / LTV by segment (assumption-driven)
| Segment | CAC | Monetisation | LTV / CAC |
|---|---|---|---|
| B2C citizen (free) | Rs 15–60 | Indirect — funnels to docs & marketplace | funnel value |
| B2C episodic (pay-per-doc) | Rs 60–200 | Rs 149–999 / doc, repeatable | strong |
| Lawyer seat (Pro) | Rs 3–8k | Rs 2.5–4k / seat / mo, 24–30 mo life | ~9–12× |
| Marketplace | shared w/ B2C | 12–18% take on GMV | per-transaction |
| Enterprise / B2G | Rs 200k–1M | Rs 3–15M / yr, multi-year | high-margin |
Blended contribution margin is unusually high (~75–86%) — the dominant revenue lines carry almost no incremental serving cost.
5-year projection — base case
Corrected for the obvious critique: break-even is carried by lawyer SaaS + pay-per-document + marketplace + institutional. Consumer subscriptions are modelled as upside, not base.
Revenue (PKR M) & EBITDA
Break-even ≈ Q3–Q4 Year 3 · ~Rs 300M ARR
Y1 EBITDA−61
Y2−94
Y3−25
Y4+220
Y5+780
Year-5 revenue mix
No single-line dependency — and no reliance on B2C subscriptions
Lawyer Pro & Firm
Marketplace take-rate
Enterprise & institutional
Episodic B2C
Cumulative burn to break-even ≈ Rs 180–220M ($650–800k).
§6Go-to-Market
One wedge, one flywheel
Beachhead: the Urdu-speaking citizen with an acute, self-diagnosable legal problem — tenancy, inheritance (وراثت), FIR & bail, cheque dishonour, khula, wage & consumer disputes. The one lane where TVL's edges are decisive and incumbents can't follow on price. Lawyer Pro is the paying anchor; pay-per-document is the first B2C cash.
Phase 10–6 months
Prove the loop. 50k users, 15k WAU. Organic Urdu SEO (2–3k cited answer pages — a compounding moat nobody ranks for), 60-second legal explainers on TikTok/Reels/Shorts (CAC < Rs 20), a WhatsApp entry point, and hand-seed 100–300 verified lawyers in 3 cities.
Phase 26–18 months
Monetise & densify. 500k–1M users, 1,500+ lawyers. Bar-council & 20–30 law-school MOUs, NGO / legal-aid / CSR donor-funded seats, and turn on paid rails: marketplace take-rate, lawyer Pro, JazzCash/EasyPaisa/Raast micro-payments.
Phase 318–36 months
Distribution at scale. 3–5M users. Telco/fintech bundles (Jazz / Easypaisa super-apps — "free legal help" as a VAS), government / donor programs as national digital legal-aid (data stays in-country), and SME/tax-litigation adjacency.
The growth flywheel — spins on electricity, not venture subsidy
01
Free Urdu answers
SEO + short-video acquire citizens at ~$0 marginal cost
02
Some convert
to a paid document or a marketplace consultation
03
Lawyers join
for the demand + Pro tools; they vet answers
04
Engine improves
every query, decline & correction grows the verified layer → routes better → more users
↻ demand ↔ supply · data ↔ accuracy — two reinforcing loops
North-star: Weekly Grounded Resolutions — unique users/week who get a citation-grounded answer or complete a consultation. Couples reach with the core promise and predicts every downstream revenue line.
§7Competition & Moat
The uncontested quadrant
~38 startups (only ~4 funded) cluster in one corner: English, lawyer-facing, API-dependent (Digital Wakeel, Wakeel.ai, Pakistan Law Bot, CauseList). The most dangerous competitor is the substitute — a citizen asking ChatGPT and getting a confident, uncited, wrong answer. That's the liability TVL replaces.
| English / generic-LLM | Urdu-first / verified-citation | |
|---|---|---|
| Lawyer-facing | Wakeel.ai · Digital Wakeel · Pakistan Law Bot · CauseList | — largely empty — |
| Citizen-facing | LawGPT · raw ChatGPT/Google · Wukla / QanoonOnline (marketplace only) | TVL — uncontested |
The moat — ranked by durability, honestly
1 · Curation velocity + trust brand durable
The corpus is public law — scrapeable. The real moat is the lawyer-vetted verified layer + an independent, published accuracy audit. Today ~2,500 rows — a head start; the raise takes it to 15k+ fast. "Benchmarked & audited" is the only defensible trust story.
2 · Local-cost structure durable
Free-at-scale that an API-cost rival cannot match on price — and privacy (queries never leave the box) that a foreign-API competitor cannot offer regulated clients.
3 · Network effects building
Demand↔supply marketplace + data↔accuracy loops compound with usage — but only once liquidity is real. Treated as a liability today, not an asset.
4 · Urdu-first NLP eroding
A genuine lead in intent-understanding — but Big Tech could add Urdu. Bank the lead now and convert it into the brand and corpus, which don't erode.
§8Risk & Mitigation
The register, and how each is contained
| Risk | Priority | Mitigation |
|---|---|---|
| Accuracy / liability / unauthorised practice of law | Critical | Framed as information + lawyer connection, never advice; persistent Urdu+English disclaimer; decline-when-unsure as a hard invariant; citation-grounding = show your work; human-in-the-loop escalation to the marketplace; LLC + ToS liability cap + E&O insurance. |
| Bar-council / regulatory stance & solicitation rules | Critical | Engage a province's bar now; a named regulatory advisor; marketplace structured as directory/subscription, not per-lead touting; propose the self-regulatory AI standard. |
| Low B2C willingness-to-pay | Critical | Don't bet the model on citizen subs. Monetise B2B lawyer SaaS, marketplace, institutional; free B2C is funnel & mission, subsidised by ~zero compute cost. Base case break-evens without consumer subscriptions. |
| Marketplace leakage (off-platform after intro) | High | Escrow, milestone payments, on-platform-only ratings, and a flat lawyer subscription alongside the take-rate — measure leakage from day one. |
| Trust/adoption in a conservative profession | High | Position as force-multiplier, not replacement; paid advocate "verifier" advisory panel; bar-association CPD pilots. |
| Content drift as laws amend | High | Versioned corpus + amendment tracking + re-verification cron + in-product lawyer-flagging loop; a part-time legal editor from first funding. |
| Competition / big-tech entry | High | Compounding curated-answer moat + local-cost price advantage; move fast on the audit & corpus depth. |
| Data privacy & security | Managed | Largely de-risked by the local model — sensitive queries never leave the country. Residual: app-layer hygiene, pen-testing before scale. |
Positioned honestly, the access-to-justice mission is TVL's cheapest customer-acquisition channel, its regulatory shield, and its path to non-dilutive capital.
The mission unlocks funding B2C revenue never will: impact investors & blended finance (Acumen, Karandaaz, i2i/Katalyst), access-to-justice grants (UNDP, World Justice Project, EU rule-of-law, World Bank/IFC), and government/judiciary MoUs where local hosting and data-sovereignty make TVL politically palatable.
§9The Sharp Investor's View
What a skeptical VC will push on — and our answer
We war-gamed this model against a hostile Series-A investor. The verdict: a strong, milestone-gated seed bet — not an A on paper projections. The real holes, and how we resolve them. Candour is the point.
?
"Your differentiation is in citizen/Urdu — the unpaid lane. Strip B2C subscriptions and does the business still break even?"
Yes. The base case is rebuilt on lawyer SaaS + pay-per-document + marketplace + institutional; consumer subs are upside, not base. Break-even lands in Y3 without a heroic conversion assumption.
?
"Legal help is an episodic distress purchase. Who pays Rs 399/mo for something they need once every three years?"
Nobody — which is why the B2C wedge is pay-per-document (episodic, no commitment), not a subscription. The sub is only for the repeat 2–4%. First pilot proves document conversion and repeat rate before scaling it.
?
"What stops your matched lawyer and client transacting off-platform and killing your take-rate?"
Escrow, milestone payments, on-platform-only ratings, and a flat lawyer subscription alongside the take-rate — so revenue survives leakage. We instrument leakage from the first transaction.
?
"63% routing is a C-minus for a trust brand, and your 'moat' is 2,500 rows over a public corpus."
Correct — so the moat is reframed as curation velocity + a published accuracy audit, not the scrapeable corpus. The raise takes verified facts to 15k+ and funds an independent audit scoring applicability, not just existence, with a near-zero confident-wrong rate in the fallback path.
?
"A solo, pre-PMF founder building a ten-stream super-app. If I fund one wedge and forbid the other nine, which do you run?"
Without hesitation: free citizen Q&A → pay-per-document, with lawyer Pro as the paying anchor. One metric — Weekly Grounded Resolutions. Everything else deferred; a technical co-founder is the first hire.
§10The Ask
Seed round
$750k–1.0M · staged, milestone-gated · ~20–24 mo runway to break-even
Plus a $150–300k blended grant target (access-to-justice / AI-for-good) to de-risk the free tier and reduce dilution.
40%
Engineering + verified-facts expansion — the data moat
25%
GPU / hosting scale-out
20%
B2B / B2G sales + lawyer-network ops
15%
G&A / legal / compliance
Verified facts 2,500 → 12,000+Routing 63% → 80%+2M+ free MAU
3,500+ lawyer seatsMarketplace GMV run-rate Rs 250M+3 signed B2G / enterprise pilots
1 bar-council relationshipPublished accuracy audit
A real, under-served access-to-justice gap; a genuine structural cost edge that lets TVL run free-at-scale where API-native rivals bleed; and a defensible position in the one lane — citizen, Urdu-first, citation-grounded — that nobody owns. A real company in embryo, with a plan built to prove the paying wedge, not to promise a super-app on day one.