Helio AI
Helio is the AI layer across Helios: it investigates devices, heals known problems, triages the inbox and talks to your clients' users. It is designed around one principle: act where safe, ask where not.
Investigations
Point Helio at a misbehaving device and it runs a real diagnostic loop: inspects state, forms a hypothesis, proposes a fix, applies it and then verifies the fix actually worked. Actions are tiered: safe read-and-repair steps run autonomously, while novel or risky actions stop and wait for a technician's explicit approval. Every step is recorded in a transcript you can read afterwards.
When an approved novel fix resolves an investigation, Helio can save it as a reusable playbook, so the library grows from your estate's real problems.
Critical alerts can start an investigation automatically: Helio triages the ticket and begins diagnosing without anyone clicking, with guardrails (one active investigation per device, cooldowns, and the same approval gates). This follows your healing scope in AI Ops: new accounts start with workstations enabled, and if the scope is set to off, Helio only investigates when you ask it to.
Auto-healing playbooks
Alert rules can attach healing scripts: alert fires, script runs, a ticket documents what happened, and if verification passes the ticket closes itself. Over twenty built-in playbooks cover the classics (spooler down, disk cleanup, service restarts), plus your own scripts and the ones Helio synthesises. Every run is audited, and cooldowns prevent a flapping alert from re-running a fix in a loop.
Daily review and network driver faults
Once a morning Helio reads what changed across the estate and lists what needs doing. A problem on one machine stays one finding, whatever it is called on the day. It comes back to your inbox only when it gets clearly worse, or a fix did not help.
For Wi-Fi and network driver faults, Helio reads the installed driver's version and date first. A current driver is never sent for an update. A few drops a week on a current driver are treated as normal background.
When the driver is old, Have Helio update the driver runs the built-in Update Network Drivers fix. It uses Dell Command Update where installed, then Windows Update. It reports the driver before and after and never restarts the machine. A needed restart goes into the client's restart window. A week later Helio checks the fault rate and closes the finding, or names the next step.
AI service desk
On email tickets Helio can triage, categorise and draft or send replies. The autonomy level is a per-tenant choice: off, draft-for-approval, or auto-reply. It can also propose a healing action for the reported problem, queued behind the same approval gates as investigations.
Ask Helio
Inside the app, Ask Helio answers questions with full context of your estate: inventory, health, tickets and its own action history. In the client portal, a client-facing version lets end users describe a problem in plain English; it reads that company's devices and tickets (never anyone else's), walks the user through safe self-serve steps, and raises a properly written ticket with findings attached when a human should take over.
Cost control
AI usage on the platform key is subject to a fair-use cap so a runaway conversation cannot surprise anyone. Prefer full control? Add your own Anthropic API key per tenant and usage is uncapped and billed to you directly.
Guardrails are the feature. Anything that changes a device is either a known-safe playbook or gated behind human approval. Autonomy is earned per action type, not assumed.
