After each evening's data refresh, the agent reasons across four signal sources at once - known-exploited vulnerabilities and the CE+ 14-day clock, assessment failures, licence waste with real pounds attached, and device health. One prioritised list, not four dashboards.
Every finding arrives as a proposal: what to do, why it matters, the predicted impact and a confidence score. Approve and Assure dispatches the fix through the automation rails; reject and the exception is logged - either way you have a change record.
Each proposal carries a plain-English explanation for the engineer and a client-friendly summary for the service report - why this, why now, what it clears. The reasoning is written from your estate's data, never from a template.
The agent can only ever choose from Assure's curated, integrity-verified automation catalogue. It never generates code and never runs anything unvetted - a manipulated signal can at worst propose a catalogue action that still needs your approval.
Every client starts advisory-only: the agent proposes, a human approves everything. Autonomy is earned per client and per action risk - and a global plus per-client kill switch stops the agent instantly, surfaced in admin.
For clients who opt in: out-of-date apps get packaged, deployed to a pilot ring inside your maintenance window, verified actually fixed, then promoted - browsers and low-risk apps first, every step audited, rollback defined before anything runs.
See the Proactive AI inbox running against a real estate - and where auto patch and deploy takes it next.
Book a demo