VAT and GST change monitoring, upstream of any tax engine.
The agent does the drudgery. Your reviewers hold the judgement.
Kehai watches official tax sources for the markets you sell in, flags only the changes that touch your footprint, and gives each one an owner, a deadline and a cited source. Independent of any tax engine: it sits upstream of all of them.
You pick the markets; Kehai monitors them. Not a fixed list: the same primary-source-first method on every jurisdiction you choose, added as we scope your footprint.
A tax engine absorbs a rule change into its calculations, so the change never reaches you. Kehai has no engine to absorb anything. Every change that touches your footprint comes to you: owned, dated, and backed by its source.
Kehai drafts. Your reviewers decide.
Built by an indirect tax practitioner who has run multi-country VAT and GST compliance in-house.
Acme France SASHighAI-generated« Le portail public de facturation publie la nouvelle version des spécifications externes. Les opérateurs concernés doivent adapter leurs flux de facturation et de e-reporting avant l’échéance applicable. »
Kehai is built by a global indirect-tax leader who has run multi-country VAT/GST compliance in-house, and who also ships software.
Operator who builds · production AI tooling on the same stack (Next.js, React, Claude)
Panellist, 'VAT Risk in the Instant Payments Era', Uniglobal 8th Annual VAT Management Summit, London, June 2026.
Why I built Kehai →What Kehai does
Scoped monitoring
Track updates across your configured jurisdictions, primary-source prioritised; coverage measured against that scope.
Business-model impact
Assess relevance against entities, products, ERP, tax engine and invoice flows.
AI assistant
Ask questions from stored tracker data, not generic internet guesses.
Audit & workflow
Confidence scores, review status, action owners, deadlines and evidence.
How Kehai beats what you're using today
Before any tax engine enters the picture, your real alternatives are a Big-4 feed, a chatbot or a spreadsheet. Here's the gap they leave.
| Kehai | Big-4 feed or newsletter | Generic AI | Spreadsheet | |
|---|---|---|---|---|
| A change happened | Detected from official sources | Yes, eventually: everything, unfiltered | Only if you ask the right question | Only if someone goes looking |
| Filtered to your footprintImpact layer | Assessed against your entities, products and flows | No. You sift it yourself | Guesses, no real context | Manual |
| Tells you what to do | A business-specific action on each item | No | Generic, sometimes wrong | No |
| Owner and deadline | Owner + deadline on every item | No | No | Only what you type |
| Cited source and a record you acted | Cited source + append-only evidence trail | Links, no trail | Can hallucinate citations | Manual, easily lost |
Kehai sits upstream of any tax engine: it does not calculate or file, and needs no integration to work.
Not another VAT newsletter. The product thesis is that the value sits in the impact layer: what changed, whether it affects your business, why, who owns the action, by when, and what source evidence supports the assessment.
Kehai doesn't replace tax judgement. It gives that judgement an operating system.
Illustrative board with sample data. Sign in to see your configured jurisdictions and live items.
Explore the live demoSee it work, not just described
The read-only demo runs the real thing on a fictional business. Explore it before you sign up. No card, no signup.
Official sources watched across your configured footprint.
Assessments drafted per legal entity, with impact and confidence.
Actions routed with an owner and a deadline.
Signed off by your team.
Audit-ready evidence packs
Every item exports a self-contained pack an auditor can read end-to-end.
- Claim, official source, and the distinct reachability and claim-support signals.
- Original-language excerpt and a content hash, captured when the source was checked.
- An append-only sign-off lineage: who reviewed, who approved, and when.
Your specialists already know the rules. What they don't have is time to refresh government websites. The agent drafts the per-entity assessments; your reviewers approve, reject or turn them into owned actions, and every sign-off lands in the audit trail.
AI assistant over your data
Ask questions answered only from your stored tracker items and sources.
- Retrieval and summary over your stored items, sources and business profile.
- It never browses the web, and it is quota-limited by design.
- Labelled as a triage aid, not advice, with its limitations stated.
Someone forwards you a scary headline about a new invoicing law. Instead of falling down a 2-hour Google hole, you check the tracker: it already logged it, scored whether it touches your setup, and linked the official source.
Ready-made boards
Reform waves arrive as a ready-made view, not a scramble.
- Built-in presets derived from your configured topics: high-impact triage, e-invoicing readiness, official sources only.
- Admin-defined presets shared with the whole workspace, every change audited.
- A per-jurisdiction board plus your own saved views alongside.
You own the risk for a business selling across eight countries, and tax is one line of a very long job description. Instead of reading six newsletters, you open one board: high-impact items at the top, each with an owner and a deadline you can actually delegate.
What makes the difference trustworthy
Honest by design: here's how the monitoring actually works.
Primary sources first
Tax authorities, official gazettes, courts and e-invoicing platforms are watched before any newsletter or commentary.
How this worksCited, scored, reviewed
Each item carries its source, a confidence score and a human-review status, kept distinct.
How this worksLocal-language, handled honestly
Every assessment links to the original-language source excerpt.
How this worksHardened source verification
Source checks record whether a URL is reachable separately from whether it supports the claim.
How this worksPoint-in-time provenance
Each verified source is captured with a content hash, so past evidence cannot be silently rewritten.
How this worksAssessed per legal entity
Relevance is assessed for each of your legal entities, not as one blanket verdict.
How this worksData & trust
No badges we haven't earned: just what is actually true today.
EU data residency
Your workspace and data are stored on infrastructure in the European Union (Vercel and Neon).
Named subprocessors
Vercel (hosting and cookieless analytics), Neon (Postgres) and email/Slack enquiry notifications. Nothing else.
Method runs server-side
Your business context is private to your workspace; the assessment method runs server-side, not in the browser.
What we don't claim yet
No SOC 2 or ISO certification, and no measured precision or recall figure: assessments are AI-generated and human-reviewed, not yet benchmarked against a labelled gold set.
Who it's for
Kehai is built for cross-border businesses selling in a handful of markets, upstream of any tax engine. The rules might be watched by a lean team with no tax department, or by in-house specialists who are tired of refreshing government websites.
Already have a tax team or adviser? Kehai takes the manual monitoring off their plate, so they spend their time on judgement, not on refreshing government websites.
You sell software subscriptions into Germany, France and the UAE. Each one keeps tweaking its VAT and e-invoicing rules. You don't have a tax team. You have a to-do list that's already on fire. The tracker quietly flags the two changes this quarter that affect you, and ignores the forty that don't.
You are the tax function: one person, a dozen jurisdictions, and a calendar full of everything except reading gazettes. The agent does the monitoring overnight; you spend your time on the three judgement calls that actually need you.
Your specialists already know the rules. What they don't have is time to refresh government websites. The agent drafts the per-entity assessments; your reviewers approve, reject or turn them into owned actions, and every sign-off lands in the audit trail.
You own the risk for a business selling across eight countries, and tax is one line of a very long job description. Instead of reading six newsletters, you open one board: high-impact items at the top, each with an owner and a deadline you can actually delegate.
Someone forwards you a scary headline about a new invoicing law. Instead of falling down a 2-hour Google hole, you check the tracker: it already logged it, scored whether it touches your setup, and linked the official source.
How the design partnership works
No public price yet. Kehai is in a design-partner phase: you shape the product and we set pricing together.
Kehai is onboarding a small number of design partners at no cost. You get it set up for your markets; in return, your feedback and, if you're happy, a reference. Pricing is set with partners and published when the programme closes.
Limited to the first 5 design partners, free for 90 days, in exchange for structured feedback and permission to be named as a reference.
Starter
Lean teams, 1-3 jurisdictions
- Scoped monitoring, primary-source first
- Business-specific impact triage
- Owners, deadlines and source evidence
Growth
Scaling businesses, 3-10 jurisdictions
- Everything in Starter
- AI assistant over your tracker data
- Exportable audit evidence packs
Scale
Multi-entity, 10+ jurisdictions, audit-heavy
- Everything in Growth
- Priority onboarding and support
- Scoped indirect-tax healthcheck to map exposure
Questions, answered straight
Is this tax advice?
No. Kehai is a triage and change-control aid, not legal or tax advice. Every item links the official source so you (or your adviser) can confirm it before acting.
How is this different from a tax engine or an all-in-one platform?
Those calculate and file your tax and absorb a rule change into their engine, so you never see or assess it. Kehai sits upstream of any engine: it surfaces the change that hits your footprint and gives you the decision, the owner, the deadline and the audit trail, needing no engine to do it, and still useful if you adopt one later.
How complete is the coverage?
Coverage is measured against the official sources you configure, not the whole world. Gaps are surfaced, not hidden, and the freshest changes can take a refresh cycle to appear.
How does the human review work?
Items are AI-generated and labelled as such until someone in your workspace reviews them. Source, confidence and review status are shown separately, so you can see exactly what has been checked, and you decide what's reviewed before it's relied on.
How is my data handled?
Your business context stays private to your workspace and the method runs server-side. We don't claim enterprise-grade security before those controls exist and are evidenced. See the methodology page.
It's founder-led. What about continuity and my data?
A fair question for any single-operator product. The monitoring runs automatically on a schedule rather than by hand each day; your workspace and data are stored on infrastructure in the European Union (Vercel and Neon); and every item exports as a self-contained evidence pack, so your register and its audit trail stay portable and usable independently of any one person.
In plain English
It's a smoke alarm, not a library: (1) tell it about your business once, (2) it watches the official sources, (3) you get a short, honest list.
Become a design partner
Share a few details and I'll set up a short call to scope your markets and the pilot. Takes a minute, no account needed yet.