Nothing runs on its own
Suggestions do nothing until a person clicks Apply. You can edit a suggestion and apply your version, or dismiss it with a reason.
Intelligence
Intelligence is three things working together: triage that reads a ticket with everything the PSA knows and proposes a plan, Scout that investigates a device under supervision, and memory that keeps what your team learns about each customer. All three record every step as evidence on the ticket.
AI triage
Press Run triage on any ticket. The run is shown live, phase by phase, so you can see what it gathered before you read what it concluded.
What you get
| Suggestion | What applying it does |
|---|---|
| Update ticket | Sets the classification and priority |
| Escalate | Escalates with the reason filled in |
| Internal note | Posts the assessment for whoever picks it up next |
| Customer reply | Adds the drafted reply to the ticket, after your edits |
| Tenant action | Prepares a lookup or change in the customer's tenant |
| Scout diagnostic | Starts a Scout session with the objective written |
| Save to memory | Records a fact about the customer or user |
| Link KB article | Attaches the relevant article to the ticket |
Transparency
Suggestions do nothing until a person clicks Apply. You can edit a suggestion and apply your version, or dismiss it with a reason.
Each finding carries chips that open the ticket note, memory entry or article it came from. A claim with no source is labelled uncited.
Watch each phase with its timing as it runs, and review the trace afterwards.
Every result lists the gaps: what was missing, what needs care, what a person should confirm.
See how much of each kind of context went to the model, so there is no doubt about what it knew.
If the model returns something unusable, the run is marked failed. It is never quietly replaced with a guess.
Scout
Scout investigates a Windows device toward an objective you write, such as "find out why the D: drive is full and whether the logs can be archived". It works within a time budget and a command budget, and stops when the objective is met.
Evidence
Every command, approval and output goes into a hash-chained ledger, and the console shows whether the chain is intact. When the session ends, the report is written to the ticket with a table of the commands behind each conclusion.
Customer memory
Memory keeps notes and facts per company and per contact. Open a ticket and the card for the caller is there: how they prefer to be reached, the device they use, the problem that keeps coming back.
Straight answers
Triage talks to an OpenAI-compatible AI gateway that you configure, so the model and where it runs are your choice. Until a gateway is configured, triage runs on a built-in stand-in that is clearly labelled as a mock, which is useful for training and evaluation.
No. Triage only proposes. A person applies each suggestion, and each one that is applied leaves a note on the ticket and an entry in the audit log. Scout pauses for approval before any command classed as destructive, broad impact or a sensitive read.
Windows devices. A technician starts the Scout agent on the device with a one-line launcher that the PSA generates for that session.
In DPMem, Dual Path's memory service, with a separate space per customer. The PSA shows clearly when a company has not been mapped to a memory space yet.
Tell us how your service desk runs today and we will walk you through Frontline PSA with a setup that looks like yours.