Know what AI your organisation is actually running.
Every system, who owns it, what it depends on, and the evidence behind each of those claims.
For organisations of 200–1,500 people, without a governance team to spare.
Many signals. One system. A person who decides.
Nothing a connector finds becomes a fact on its own. Signals correlate into a candidate; Allow suggests what it is, with the evidence behind the suggestion and what it might have wrong; and a person confirms, merges or rejects. Every original source record is kept.
Five layers, one platform, one object model.
Discovery without understanding is a list. Understanding without action is a report.
Discovery and resolution
Find the AI you did not know you had.
Import a spreadsheet, ask your people, or connect Microsoft, Google and GitHub. Everything arrives as a candidate for review.
Connected understanding
One record, and everything around it.
Identity, purpose, architecture, people, relationships, evidence and history on one surface.
Action and governance
Findings that become decisions.
Evidence-backed findings, recommendations, and decisions a named person is answerable for.
Continuous intelligence
A picture that stays current.
Sources are re-read on a schedule, and every run records what it could not see as well as what it found.
Foundation and access
Start without an implementation project.
Sign up, invite your team, get to work. Enterprise SSO is there for the organisations that need it.
Six sources, and an honest account of each.
Nothing a source reports becomes a system on its own. It enters the intake queue as a candidate for a person to confirm, merge or reject.
Manual registration
Guided creation with owner, purpose and evidence.
CSV import
Mapped import with preview, validation and duplicate checks.
Employee declaration
A short survey. Often the only source for shadow AI.
Microsoft Entra ID and 365
Applications, service principals and their permissions.
Google Workspace
Authorised apps, OAuth grants and directory context.
GitHub organisation
Repositories, AI dependencies and provider references. Never source code.
The coverage rule
If Allow can only see part of an organisation’s AI landscape, it says so plainly and explains what would improve the picture. A confident-looking empty result is worse than an honest partial one.
Rules the product is built under, not claims it makes about itself.
These constrain what the engineering is allowed to do.
Evidence before conclusions
No finding or decision without traceable evidence — or an explicit “insufficient evidence”. There is no third option.
One object, one Passport
Source signals may be many. The canonical record is one, and every merge keeps what it merged.
Provenance is never lost
Every fact carries where it came from. Corrections create new states; they never overwrite history.
AI assists, humans decide
AI may draft, classify and explain. It may never approve, accept risk or assign accountability.
Honest uncertainty
Confidence is high, medium or low with its signals visible — never an unexplained percentage.
What Allow is not
- A model platform or agent framework
- An execution or orchestration layer
- A compliance database or GRC replacement
- A generic dashboard product
- An autonomous decision maker
One platform, four ways in.
Entitlements on one codebase and one object model — not four products. Moving between them changes what is switched on, never what your data has to become.
Discover
A generous point-in-time discovery and report.
Lite
A simplified ongoing experience for smaller organisations.
Intelligence
The core mid-market platform.
Enterprise
Advanced security and administration.
Allow Discover
Allow Discover shows every valid finding your own scan produced. No blurring, and no “upgrade to reveal”.
Pricing is not self-serve yet — packages are agreed directly while the product is in its pilot phase.
Start with what you already know.
Create an organisation, import the spreadsheet you are already keeping, and give each system an owner. The ones nobody owns are what Allow surfaces first.