Legaltech · 720-hour cycle
kaedax builds legaltech software — matter management, document automation, AI-assisted review, client intake — in a fixed 30-day (720-hour) cycle. Privilege-aware access controls, defensible audit trails, and evaluation harnesses on any AI output are designed into the spec, because in legal software, 'mostly correct' is a malpractice exposure.
First-pass review
−68%
attorney time
Citation coverage
100%
every claim
Eval accuracy
97.4%
atty-graded
Privilege checks
0
violations
[ §01 ] the problems we're hired for
An AI that summarizes contracts impressively but unverifiably is unusable in practice. Every AI surface we ship for legal work carries an eval harness — graded against attorney-reviewed answers — plus citations back to source text for every claim.
Matter-level access, ethical walls, conflict checks — retrofitting these into a flat data model means rebuilding. We design privilege boundaries into the schema on day one, enforced at the database layer, not just the UI.
Legal users have the lowest tolerance for friction of any professional vertical: every extra click is billable time lost. We design around the matter workflow as practiced, instrumented from day one to find where minutes leak.
Redlines, approvals, signature chains, precedent libraries — legal documents need diffing, history, and audit the way code does. We bring software-grade version discipline to document automation.
[ §02 ] what ships
The shapes of legaltech work that fit a 720-hour cycle. Every build ships with runbooks, monitoring, and 30–60 days of post-launch on-call.
Matters, parties, deadlines, conflicts — with privilege boundaries enforced in the schema.
Template engines, redline workflows, approval chains, signature integration with full history.
Extraction, summarization, and first-pass review — eval-gated against attorney-reviewed answers, every claim cited to source.
Conflict-checked intake, secure document exchange, status surfaces that cut 'any update?' emails.
[ §03 ] proof
Our legaltech posture is built on adjacent regulated work — see all case studies across fintech, healthtech, insurance, and AI products. The engineering disciplines transfer; the scope call establishes the specifics.
Partners don't trust software that touches privilege. kaedax shipped review tooling with ethical walls and an audit trail our GC signed off on — first try.
[ §04 ] questions
Only with a defensibility layer: evaluation harnesses graded against attorney-reviewed answers, citations from every AI claim back to source text, and confidence thresholds that route low-certainty work to humans. That layer is the difference between a legal AI product and a liability — and it's the first thing we build.
Privilege boundaries are enforced at the database layer — matter-level access, ethical walls, conflict-aware permissions — not just hidden in the UI. Mutual NDA precedes any scope conversation, and all data lives in your cloud.
Market quotes for legaltech MVPs run $50,000–$200,000 over 3–6 months. kaedax prices per fixed 720-hour cycle, shared on the scope call — fixed scope, fixed timeline, no overrun billing.
Yes — template engines, redline workflows, approval chains, and signature integrations with software-grade version history. Legal documents are version-control problems, and we treat them with that discipline.
Both: legaltech founders building products, and firms or legal-ops teams building internal tooling. The privilege-aware architecture is the same; the scope call establishes which workflows fit a single cycle.
A 15-minute scope call with a kaedax founder and the engineer who'd lead your build. We either fit your legaltech product into 720 hours, or we tell you why not — same call.
▷ other verticals