AI Approval
What is it?
AI Approval (more accurately, AI Approval scoring) is an auto-approval mechanism in Admin By Request (ABR) that uses a continuously updated global reputation database to evaluate application elevation requests - scoring each application and its publisher on a 0-100 scale, and automatically approving requests that clear a configured threshold.
Unlike Pre-Approval rules (which require an administrator to explicitly list trusted applications) and Machine Learning auto-approval (which requires an organization to accumulate its own approval history), AI Approval scoring draws on elevation data from ABR tenants worldwide and requires no per-application setup.
A brand-new ABR deployment with no approval history at all can enable AI Approval scoring on day one and immediately benefit from the global reputation data behind it. The database covers more than 12 million applications and is updated continuously as new elevation events are processed across the platform.
What problem does it solve?
When ABR is deployed with the approval workflow enabled, every elevation request from every user lands in the portal queue. For common, well-known applications - a widely used PDF reader, a major development IDE, a mainstream compression tool - the approval decision is essentially automatic: anyone looking at the request would approve it immediately. The challenge is that someone still has to look at it, and in large organizations with many users, those routine approvals add up to a substantial administrative burden.
Machine Learning auto-approval reduces that burden by learning your organization's own approval patterns over time. But Machine Learning has a bootstrapping problem: a brand-new deployment has no history to learn from, so every application has to go through at least one round of manual approval before ML can take over. In environments transitioning from standing local admin rights to managed elevation, the first few days or weeks can generate a wave of approval requests that overwhelms the help desk.
AI Approval scoring bypasses that bootstrapping problem entirely. Instead of waiting for your organization to accumulate approval history, it draws on a global database built from elevation events across all ABR tenants worldwide. If hundreds of organizations have been safely elevating the same application for years, ABR already knows it is trustworthy - and your users do not have to wait while your team builds that knowledge from scratch.
The other gap AI Approval scoring fills is the "exotic applications" problem. Pre-Approval lists and Machine Learning both work well for software that is common in your environment. But organizations with specialized or creative workforces regularly use software that is only popular within a narrow industry vertical. A graphics studio might use obscure rendering tools; a law firm might rely on legacy document utilities. AI Approval scoring handles the broad common-software case automatically, freeing administrators to focus their attention on the unusual applications that genuinely require individual judgment.
AI Approval scoring and Machine Learning auto-approval are complementary, not competing. AI Approval scoring handles globally reputable applications from day one. Machine Learning handles applications that are common in your specific environment but may not score highly in a global database - for example, internal tools or industry-specific software that no one outside your sector uses. Using both together typically delivers the most substantial reduction in manual approval workload over time.
Applications that score poorly on both dimensions - rare software from unknown vendors, newly released tools with no elevation history - always reach the manual approval queue. This is the intended behavior, since low-scoring software is exactly the category where a human review adds the most security value.
A global reputation database
Every elevation event from every ABR tenant feeds a single 0–100 score per application and publisher, enabling day-one auto-approval.
