Questions, answered.

PROVEAiBLE is AI client intake for law firms — it works whether or not the client has a document to upload. Clients with documents get redaction-first analysis. Clients with nothing to upload still produce a structured matter file from the conversation alone — classification, parties, timeline, gaps, and what to bring to the first meeting. Either way, the lawyer gets a usable handoff before the first call.

A structured matter file — not a chat dump. It contains the redacted document, key facts with sources, a constructed timeline, flagged inconsistencies, missing evidence, and timing concerns. Every claim points back to where it came from — so your lawyer can start from an organised file instead of a blank page.

Forms collect whatever the client types. Generic chatbots produce long transcripts. PROVEAiBLE starts with the issue, requests the key document if relevant, identifies what is missing, and prepares a handoff your team can act on. When there's nothing to upload, the conversation alone produces the same structured matter file.

It's for law firms that want better-prepared first calls, less time chasing documents, and a cleaner way to review new matters before replying.

Yes — that's part of what the founder rate buys. About 2–3 weeks after you go live, we book a 30-minute founders' shaping call: what's missing, what would you add, what's broken. Round 1 firms have direct input into the next features (practice-area routing, CRM integrations, additional jurisdictions, additional languages). The founder rate is locked for life partly because you're co-creating the product with us.

Many real intakes start as pure conversation — the client just got fired this morning, the accident was yesterday, the deportation notice was verbal. PROVEAiBLE handles this. Without any upload, the conversation alone produces the same structured handoff: classification, parties, timeline, client goal, procedural posture, plus a clear list of what the client should bring to the first meeting. Documents enrich the file but aren't required.

No — that's deliberate. PROVEAiBLE gathers facts, redacts identifying details, extracts what your client said, and surfaces timing risks from the documents themselves (long gaps, fast-approaching dates, unusual delays). It does not verify statutes against an authoritative law library, run procedural-compliance checks, or produce legal opinions. Legal analysis is your lawyer's job — we hand you a clean, organised file so the lawyer's analysis can start immediately.

It follows a simple sequence: understand the issue, request key documents if relevant, review what's been provided (verbal account or upload), identify what is missing, and ask only the next questions that matter.

Yes. At launch the system is calibrated for the most common matter types — employment, family, personal injury, immigration — and we tailor the document requests, follow-up questions, and handoff structure to your firm's specific practice areas during setup.

Centralised inbox delivery is available — every completed matter lands in your firm's intake address. Practice-area routing (a different lawyer per matter type) ships as a no-charge upgrade for founding members.

Yes. The system is designed to deliver case files where lawyers already work, especially email and Google Drive. The goal is to avoid giving your team another dashboard to manage.

Sessions persist on the backend so the client can resume via their original session link. Your firm only receives the structured handoff once the intake reaches a useful, review-ready stage. Incomplete sessions are purged after 90 days.

PDFs (native and scanned), JPEGs, PNGs, and DOCX files all work end-to-end. Image-only documents with no extractable text get flagged — the client is asked to retake or upload a different version. Audio and video transcription aren't yet supported (rolling out via Whisper integration in a later phase).

The handoff is meant to be read, not blindly trusted. Every fact carries a source basis tag (verified by document, claimed by party, expert assessment, etc.) so you can quickly see what's grounded vs what's claimed. Lawyer review before billing is the safety net — the system surfaces facts; you decide what to act on. We tune the AI's skill prompts based on real handoffs you flag.

Yes. The system has built-in prompt scaffolds for seven matter types — criminal, family, employment, personal injury, immigration, housing, contract. Per-firm tuning is done by us during setup (we configure your firm's matter types, question sets and redaction directly). Ask any time and we adjust.

Soft cap of ~50MB per upload. No hard page-count limit, but only the first ~12,000 words of any single document are sent to the analysis model — long documents get truncated for analysis, with the full original preserved in the handoff package.

Today the widget accepts PDF, JPEG, PNG, and DOCX. Per-firm restriction (e.g. "PDF only") isn't exposed as a config flag yet — it's part of the admin UI rolling out shortly. If you need a specific restriction at install, we can wire it manually in your firm's config.

No. Client data is processed to build the case file, then delivered to your firm. Completed intakes are deleted on delivery; incomplete sessions are purged after 90 days. PROVEAiBLE does not operate as a long-term system of record for ongoing case files.

Yes — and that's the central design choice. The client's identifying details are redacted first, and the client reviews what's hidden. Only the protected version is ever sent to the case-analysis model. Names, addresses, phone numbers, identifiers — whatever the client chooses to hide — never reach the case-analysis model. That answers the number-one objection cautious lawyers raise about AI: "I can't send client data to an LLM." With PROVEAiBLE, you're not — the model only sees the version your client approved. (See the next question for the OCR-stage detail — we're transparent about it.)

Honest answer: yes, the OCR layer does process the raw document text. We use Google Cloud Vision API for text extraction (industry-standard, GDPR data-processor agreement, used by major financial and healthcare institutions). That's a separate layer from the AI that performs the case analysis.

Two AI layers, two privacy boundaries:
1. OCR (Google Cloud Vision): extracts text from images. An industry-standard utility, not a reasoning model. Subject to standard data-processor agreements.
2. The case-analysis model: only ever sees the privacy-redacted version. Names, addresses and identifiers replaced with role tags (CLIENT, OTHER_PARTY, WITNESS) before it reads a word.

The redaction happens between layer 1 and layer 2, before the case-analysis model reads anything. We're explicit about this distinction so you can answer your bar association honestly: no client identifier ever reaches the case-analysis model. The OCR layer is the same kind of infrastructure your firm probably already trusts when scanning court filings.

Redaction is keyed off the names and terms the client lists in the redaction step — those are removed at OCR-word-level with substring matching, so "Maria", "Silva", and "Maria Silva" all get blacked out together. If the client doesn't list a name, mentions of it in the document text will pass through. The lawyer view shows the redacted PDF plus a count of redaction terms applied, so you can see exactly what was hidden. Auto-detection of unlisted PII is rolling out as an additional safety layer.

The intake processes the client's messages, uploaded documents, extracted text, and the structured case state needed to request documents, ask follow-up questions, and prepare the handoff.

No. PROVEAiBLE is a processing and delivery layer for intake, not a case-management database. The case file is prepared and handed off to the firm for review and ongoing work.

It covers the intake system, workflow tailoring, support, maintenance, and keeping your firm's installation running.

No. AI usage is separate. Your firm connects its own AI key, and usage is billed directly by your provider. The PROVEAiBLE monthly fee does not include model usage charges.

Your firm provides an API key from Anthropic, OpenAI, or OpenRouter. You set up the account once during install — it takes about ten minutes. Token usage is billed directly by your provider on your existing card, never through us. OpenRouter is the cheaper option if you want to economise on flagship-model costs while still getting solid quality on most matters.

You stay on the rate you join at as long as you remain an active customer — we don't raise pricing on existing firms. Your plan is scoped to your firm at setup, with no per-lawyer multiplier.

Token cost depends on document length and which provider you use. A typical single-document intake — say a 5-page termination letter and three follow-up questions — runs roughly $0.10–$0.80 with Anthropic Claude Sonnet, or $0.02–$0.10 with OpenRouter's cheaper models (Gemini Flash, Haiku-class). A heavy multi-document matter — several uploads plus a full case file — can reach around $1. Larger documents scale proportionally. Spend is visible in your provider's own dashboard — never marked up by us.

Yes — you set spending limits directly in your Anthropic, OpenAI, or OpenRouter account. Most providers let you set a hard monthly cap or alert thresholds. Because the key is yours, the controls are yours. We don't sit between your firm and your token spend.

The AI converses in 60+ languages. Your client tells their story in their own language, and your firm opens the finished case file in your language. The widget interface itself is fully localised in 13 languages. If you work in a particular language pair regularly, tell us during setup so we can tune the intake and handoff for it.

Yes. Your client can tell their story in their own language, and your practitioner opens the finished case file in English — the account, the documents they hold, and the deadline that matters. This is exactly what the immigration and asylum intake is built for.

The OCR layer (Google Cloud Vision) supports a wide range of languages for text extraction, and the AI converses in 60+ languages on top of it. If you regularly handle documents in a specific language, flag it during setup so we can confirm extraction quality and tune the handoff with you.

The handoff is delivered as an email + structured PDF + ZIP. Most CRMs (Clio, MyCase, PracticePanther, Smokeball, etc.) support email-to-case forwarding rules, so the matter file lands in the right case automatically. Direct API integrations are on the roadmap but not yet built — if you need deep CRM integration on day one, talk to us before signing so we can be honest about what's possible.

Yes. The email subject template, the PDF cover page, and the handoff email header are all swappable per firm — we configure these manually during onboarding. A small "Generated by PROVEAiBLE" footer on the PDF stays as required attribution on the standard tier; that's removable on a future enterprise tier.

Cancel any time — access continues to the end of the current paid period. Existing intake sessions stay live until the end of that period so any in-flight client finishes properly. Before the cutoff date you receive a complete export of every completed intake (PDF + JSON + ZIP). After cutoff the widget stops accepting new sessions; everything you've already exported is yours to keep. To cancel, email support and we handle the rest within one business day.

Yes. Intake can be deployed as a hosted link, through an email signature, with a QR code, or through another direct contact path. A full website is not required.

We install it with you — tailoring the intake flow, document requests, routing and delivery to your workflow before launch.

The next step is a founder-led walkthrough of your workflow, followed by setup, testing, and installation. Once the intake is calibrated for your firm, it goes live and starts delivering structured case files.

Redaction runs server-side before the case-analysis model sees a word, so the model only ever receives the protected version — names replaced with role tags (CLIENT, OTHER_PARTY, ADDRESS, ID, PHONE). Your client confirms what's hidden, and each choice is recorded as a consent record (GDPR Article 7) during the intake. The redacted documents your team receives are themselves the evidence the model never saw identifying data — defensible by architecture, not by promise.

Yes. The AI cross-references the client's chat answers against the uploaded documents and flags inconsistencies in a dedicated "Disputed facts" + "Contradictions noted" block in the matter file. Example: the client says "performance-based termination" in chat; the termination letter cites "company restructure". You see the mismatch before you sign the retainer — when you can still decline, not after billing twenty hours.

Yes. The handoff reads like a junior associate prepared it — every fact-ledger entry shows the source document, page number, and line from the client's own materials. Traceable from minute one. No "trust the AI." If your team wants to verify a single claim, the source quote is right there in the brief; if you need the original document, it's attached.

Yes. The AI extracts every date in the client's uploads and flags suspicious patterns: long gaps, fast-approaching response windows, unusual delays, statute-of-limitations proximity. A 14-month gap between incident and witness statement, a 9-day filing window, a missing severance schedule — all surfaced in the matter file's "URGENT" banner or "Timing risks" section. Final verification stays with you; the AI surfaces, the lawyer judges.

Only what the client's own documents contain — and only as unverified references. If a client uploads a document that mentions a statute, article, or case number, that reference is preserved in the file alongside the source quote and flagged for your verification. PROVEAiBLE does not run legal research, maintain a law library, or check any citation itself. Confirming what the law says, and whether it applies, stays your lawyer's job.