Jarvis
Jarvis is the AI assistant that is always one keystroke away — open it with Cmd+J (Mac) or Ctrl+J (Windows) from any The One product without leaving your current context. Unlike the Playground, Jarvis knows where you are and what you are doing. It reads your current product context, remembers your role and preferences, and uses that to give you answers that are relevant right now.
Opening Jarvis
Press Cmd+J (Mac) or Ctrl+J (Windows) from anywhere in any The One product. The Jarvis panel slides in from the right side of the screen without navigating away from your current page.
Alternatively, click the Jarvis icon in the HubBar (the AI sparkle icon at the top right of every product).
To close Jarvis, press Esc or click anywhere outside the panel.
How Jarvis Knows Where You Are
When you open Jarvis, it reads context from the current product:
| You are in... | Jarvis context includes... |
|---|---|
| PSA — Ticket detail | Ticket ID, status, assigned technician, company name, recent notes |
| PSA — Dashboard | Open ticket count, SLA breaches, recent activity |
| CRM — Company detail | Company name, contacts, recent activities, open opportunities |
| Defend — Alert detail | Alert type, severity, device name, detection timeline |
| Defend — Investigation | Process tree context, lateral movement indicators |
| Voice — Call in progress | Caller identity, live transcription, previous call history |
| Books — Invoice detail | Invoice amount, due date, company, line items |
| RMM — Device detail | Device name, OS, patch status, recent alerts |
| CMDB — Asset detail | Asset type, related CIs, recent changes |
This context is injected into Jarvis's system prompt automatically. You do not need to paste in ticket IDs or client names — Jarvis already knows.
What Jarvis Can Do by Product
In PSA
- Summarize a ticket's history in plain English
- Suggest the next troubleshooting step
- Draft a response to the client
- Identify similar past tickets
- Draft a time entry description from your notes
- Check SLA status and flag breach risk
Example prompts:
"What has been done on this ticket so far?" "Draft a client-facing update for this ticket" "Has this client had this issue before?"
In CRM
- Summarize a company's recent activity
- Draft a follow-up email after a meeting
- Identify upsell opportunities based on service history
- Prepare a QBR agenda for a client
- Suggest next best action for an opportunity
In Defend
- Explain a detection in plain English
- Summarize an investigation with a narrative attack chain
- Recommend a remediation plan for a threat
- Draft a security incident notification email
- Assess whether an alert is likely a false positive
In RMM
- Summarize device health across a client site
- Identify which devices are overdue for patches
- Draft a maintenance window notification
- Analyze monitoring alert patterns
In Voice
- Summarize a call in progress (live or post-call)
- Identify action items from a call transcript
- Draft a follow-up email from the call
In Books
- Explain a billing discrepancy to a client
- Summarize outstanding invoices for a company
- Draft a payment reminder email
Anywhere (no product context)
- Answer general MSP questions
- Help write documentation, SOPs, or runbooks
- Generate training material
- Draft marketing copy
- Explain technical concepts
Jarvis Memory
Jarvis remembers you across sessions. Memories are stored in your AI Platform user profile and are automatically included in Jarvis's context with every conversation.
Types of memory:
| Type | Example |
|---|---|
| Preference | "I prefer concise responses under 200 words" |
| Fact | "My primary focus is Microsoft 365 environments" |
| Instruction | "Always include a 'Next steps' section in summaries" |
| Context | "I manage 45 clients in the healthcare and finance verticals" |
Sources of memory:
- Explicit — You tell Jarvis something directly ("Remember that I prefer bullet points")
- Inferred — Jarvis notices patterns in your interactions over time ("You often ask about SLA-related questions")
Managing Your Memories
- Navigate to AI Platform → Settings → Jarvis Memory
- View all current memories with their type, source, and confidence score
- Delete any memory you no longer want
- Edit memories to correct inaccurate inferences
- Click Clear all to reset Jarvis memory entirely (this cannot be undone)
User Preferences
Your preferences shape every Jarvis response:
| Preference | Options |
|---|---|
| Response style | Concise, Detailed, Technical, Balanced |
| Timezone | Used for time-relative answers |
| Role | Owner, Technician, vCISO, Account Manager, etc. |
| Specialties | M365, Networking, Cybersecurity, Cloud, etc. |
| Clients managed | Count and verticals — helps Jarvis personalize scale |
Update preferences at Settings → Jarvis Memory → Preferences.
Conversation History
Jarvis maintains a conversation history within each session. When you close the panel and reopen it, the previous conversation is still there. To start fresh, click New conversation at the top of the panel.
Conversations are private to you — they are not shared with your organization and are not visible to your manager or platform administrators.
Jarvis vs. the Playground
| Feature | Jarvis | Playground |
|---|---|---|
| Access | Cmd+J anywhere | Navigate to AI Platform |
| Context awareness | Yes — reads current product context | No — general chat only |
| Memory | Yes — persistent across sessions | Yes — conversation history only |
| Model selection | Automatic (GPT-4.1 via Gateway) | User-selectable |
| Custom GPT invocation | Yes — can invoke your Custom GPTs | Yes |
| Branching conversations | No | Yes |
| Folder organization | No | Yes |
| Best for | In-context help while working | Deep research, drafting, testing |
Using Jarvis to Invoke Custom GPTs
You can ask Jarvis to hand off to a specific Custom GPT:
"Use the QBR Prep Assistant to prepare for my meeting with Contoso tomorrow"
Jarvis will invoke the QBR Prep Assistant agent, pass the relevant context, and continue the conversation through that agent's lens.
Privacy and Data
- Jarvis conversations are stored in your tenant's isolated database partition
- Conversation data is not shared with other tenants
- Conversation data is not used to train AI models
- Admins cannot read individual user Jarvis conversations
jarvis-assistant feature route, which routes to GPT-4.1 through the AI Gateway. All platform quotas and governance policies apply to Jarvis interactions.