RegAI by AIGC · For compliance teams in banks and financial institutions
Know what each new regulation asks of you. See exactly what changed when it is revised.
RegAI is an AI product that turns new and revised regulations into proposed risks and controls for your risk and control register. It follows your regulators, reads each new regulation article by article, and compares a revised edition with the one before it. Each proposal quotes the text, and a qualified reviewer approves it before it reaches the register.
How RegAI works
From the regulator’s announcement to your register.
The job todayThe register lags the regulation. Reading, mapping, and typing all queue behind the same few people, so the controls arrive long after the circular.
Watch
Every day, RegAI collects publications from the regulators you follow. Each one links to its source on the regulator’s own site.
Sort
RegAI tags announcements, circulars, licenses, and amendments by regulator, type, and importance. Regulations and revisions move into the analysis queue. Everything else stays readable as context.
Read
RegAI reads a new regulation article by article, and compares a revised one with the edition before it.
Review
A qualified reviewer approves or edits every proposal, and every approval carries the reviewer’s name.
Register
RegAI checks the approved risks and controls against what is already in your register, so nothing is entered twice, and exports them.
Nothing reaches the register without a human.
A new regulation
Every article, read for the duty it creates.
The job todayTwo good analysts can map the same article two different ways. When a supervisor asks which article requires a control, the answer sits in someone’s memory or an old spreadsheet.
For each article, RegAI proposes the risk and the control, with the supporting passage quoted from the text. RegAI removes any quote that does not exist in the article. When a supervisor asks which article requires a control, the passage sits beside it.
When the article creates no duty
Many articles are definitions or permissions. They create no duty, and then the right answer is nothing. RegAI drafts, checks its draft against the article, then answers.
A surface reading
Due diligence is a control, so there must be a risk to log.
Confident, well written, and wrong.
RegAI’s reading
- Draft a due-diligence risk.
- Check the draft against the article.
- The article says “may”: a permission, with no duty attached.
- Remove the risk. Nothing to log.
“May” gives permission. “Must” creates a duty. RegAI checks which one it is reading before anything goes to review.
A revised regulation
Show exactly what changed. Then explain it.
The job todayA revised edition means the whole text is read again. A threshold that moves inside one definition is hard to spot across two long documents.
Give two versions of a regulation to a general-purpose model and it can report a new layout or a scanning error as a change. RegAI keeps the two jobs apart. Fixed rules find the change, and only then does the AI explain it.
- Fixed rules
Line up
Old and new text, matched article by article.
- Fixed rules
Mark the changes
The exact words that changed, apart from layout.
- AI
Explain
The marked text only: a duty added, changed, or removed, and whether it matters.
- People
Review
A qualified reviewer checks each difference before it is released.
Run it twice and you get the same result.
Article 2: a spelling correction, meaning unchanged. Dismissed as a non-change, and shown in the count.
A revision, as your team receives it
Sorted by what it means
New obligations, changed obligations, removed obligations, scope and thresholds, editorial, and unclassified. The material ones come first.
Before and after, quoted
Every explanation sits on the passages behind it, quoted from each version. Both versions stay available beside it.
Nothing disappears silently
Differences dismissed as non-changes still appear, and still count.
Why it reads the way your experts read
Some answers can be looked up. Some have to be learned.
For shared standards
Library-backed
For shared standards with an established control pattern, such as AAOIFI’s Sharia standards and IFSB’s prudential standards. RegAI asks what your most experienced analyst asks: have we seen a clause like this, and what did we decide? It searches a library of known clauses and the controls already approved for them, then quotes only your document. When a standard changes, we update the library and add every approval back.
For supervisory articles
Expert-trained
For supervisory articles where the answer takes a professional’s reading, and where “may” is not “must”. A specialist model learned from published regulations and the risks and controls that compliance professionals set for them, and was tested on articles it had never seen.
If the standard already decided, RegAI looks it up. If it takes judgment, RegAI applies what it learned from the professionals.
Built for groups
One rule, several regulators, several countries.
Valid where you operate
A rule that applies in Jordan may not apply in Iraq or Palestine. RegAI keeps country validity for groups that answer to several regulators.
Compare across countries
Set the same topic side by side in two jurisdictions, and see where the obligations, thresholds, deadlines, and definitions differ, so a group can decide where one policy works and where a local one is needed.
Arabic, through the whole job
The comparison engine reads Arabic regulations in Arabic, and clause matching understands Arabic terms directly. One switch turns the portal to Arabic, with full right-to-left layout.
For Islamic banks
AAOIFI’s Sharia standards and IFSB’s prudential standards take the library path, so every institution gets the same answer, cited from the current edition. On Sharia questions, qualified Sharia experts keep interpretation and final approval.
The regulators you answer to
Add each regulator you report to. Its publications join the same watch, queue, and register.
Getting started
A pilot on one regulation, judged by your numbers.
How accurate is it on your regulations? A pilot answers that on your own articles, before you commit.
Choose one regulation
We choose it with you, and decide which method fits it.
Bring what you have
If you keep a control library or a risk register, we start from it. If not, we start from ours. Either way, nobody starts from a blank sheet.
We build and test
The library or the model, tested on articles held back from the start.
Your reviewers work in RegAI
On real documents, measured against your own team’s decisions.
The rules we score it by
- We hold test articles back before the work starts and never use them to build it.
- A finding is correct when it means the same as your professional’s finding, even in different words.
- “Nothing applies”, when nothing does, counts as correct.
- A quote that is not in the article counts as a miss, even if the idea was right.
- The score weighs how many proposals are right against how many of the right answers it found. Proposing very little, or everything, scores badly.
Where your documents stay
RegAI reads public regulation. Your control library and your decisions stay where you choose: a managed cloud, a private cloud, a managed appliance inside your network, or air-gapped. In an air-gapped site, your security team agrees with us at scoping how regulator updates arrive.
Next step
Bring us your next regulation.
Bring the regulation and the people who sign off, with your control library and risk register if you keep them. We show you which method fits it. We run the pilot on your own articles.