Skip to main content

Companies and Autonomy

SmartTech follows the directory companies you activate. For each one, the Client agent page sets how the agent behaves: instructions, tools, autonomy and memory.

Activating a Company​

The Companies page lists the directory companies. Activating a company requires an active PSA integration and a SmartDoc subscription.

StateDescription
Not activatedSmartTech does not handle this client's tickets
ObservationSmartTech triages and diagnoses, but nothing is published to the PSA
ActiveLive service: notes may be posted to the PSA and fixes run according to the command rules
DisabledSmartTech stops handling this client's tickets and its autonomy is revoked

An enabled client always starts in observation. "End observation" takes it live.

Time Window​

Diagnoses only run during the hours chosen for the client (tenant time). A card outside the window waits in the "Waiting" column.

When the client is linked to SmartDoc, the diagnosis reads its documentation and machine records. Without a link, it runs without them.

Import Open Tickets​

"Import open tickets" brings in the open tickets created in the last days (200 at most) on watched boards. They follow the normal flow.

Client Agent Page​

Every change to the page creates a version, visible in its history.

TabContent
ProfileTone and stance of the technician for this client, and client-specific rules to follow on every ticket
ProceduresProcedures by category (global, organization or client)
Tools and scriptsTools the agent may use for this client; mandatory tools can never be turned off
AutonomySigned policy, thresholds and automatic closure
MemoryWhat the agent learned about this client

The client's instructions are added to the technician's global personality and to the category procedure, and always stay below the safety rules.

Example instructions: "always quote the machine number"; "never restart a server during business hours".

Autonomy​

The signed autonomy policy is the authorization that lets the agent run commands without human approval for this client, under the thresholds frozen when it was signed. Without a policy (missing, revoked or expired), every command goes through an approval.

ThresholdEffect
Automatic execution thresholdBelow this score, the command runs without approval
Block thresholdFrom this score, the command is refused, even without autonomy

In between, the command waits for an approval. The block threshold always stays above the automatic threshold. The page shows the client's real commands and their zone under the thresholds.

  • Signature — Enabling, renewing or saving autonomy creates a policy signed in your name. A changed threshold creates a new signed version.
  • Expiry — An expired policy cuts autonomy: renew it before the deadline.
  • Revocation — The agent no longer runs any fix without approval for this client; the reason is kept in the log.
  • Restrictions — Commands that always need an approval, or are always refused, for this client.

A client in observation has no autonomy: end observation first. Default thresholds for clients without their own setting are set in Settings → Agent.

Automatic Closure​

A card whose every change was verified as fixed by its own check can move to "Done" on its own.

ModeEffect
OffNo card closes on its own: the agent works, a technician validates
SimulationSmartTech records the cards it would have closed, closing none: enough to measure before turning it on
OnA card that meets every condition moves to "Done" on its own, within the daily cap

Conditions:

  • Every applied change was verified as fixed by its own check.
  • A 15-minute hold follows the report: time to step in before the closure ("Take the card" keeps it open).
  • Never a server, a critical priority or the "Other" category.
  • Never after a human step-in, a security flag, a ticket changed in ConnectWise or a repeat on the same machine.

Automatic closure is first allowed in the organization settings (main switch, off by default), then client by client in the Autonomy tab, which requires active autonomy.

Memory​

The Memory page shows what the agent learned about your clients and their machines during its diagnoses: facts, lessons and negative lessons. Each item has a scope (company, machine, person or organization lesson), a confidence and its source.

Each item can be fixed, removed, or reviewed in its history. A manually added fact stays data: it is presented to the agent like any other fact, never as an instruction. Text that looks like an instruction is refused.