Resolves the 70% of IT tickets that follow predictable patterns — password resets, software access requests, VPN setup, hardware troubleshooting, account unlocks — autonomously, 24/7. Your IT engineers focus on infrastructure and complex problems, not L1 admin.
70%
tickets auto-resolved
L1 without human touch
< 3 min
resolution time
for autonomous tickets
24/7
coverage
no out-of-hours backlog
50%
engineer time freed
for real engineering work
Typical build: 3-week sprint · Fixed price · Production-grade
Auto-resolution
70% of tickets
Response time
< 3 minutes
Coverage
24/7
This is the actual Agentforce configuration Kovil AI builds and deploys — not a diagram. Here is what runs inside every node.
The IT Helpdesk Agent accepts tickets from multiple intake channels: Slack command (employees type /it-help and describe their issue), email to the IT helpdesk address (parsed via email-to-case), the Salesforce employee self-service portal, and direct escalation from the HR Onboarding Agent for new hire provisioning issues. Einstein Case Classification analyses the ticket description to identify: issue type (credential, access, hardware, software, network, account management), urgency tier (blocking work vs inconvenient), affected system (Salesforce, Slack, Jira, laptop, VPN — extracted from ticket text), and whether a known resolution exists in the knowledge base. Classification determines the resolution path within seconds of submission.
For every classified ticket, the agent performs a semantic search of the IT knowledge base — which covers documented resolution procedures for all common L1 issues. Unlike keyword search, semantic search handles the variety of ways employees describe the same problem: 'can't log in', 'password not working', 'locked out of my account', and 'authentication failing' all retrieve the same credential reset procedure. The KB search is filtered by the affected system identified in classification — a Salesforce login issue retrieves Salesforce-specific procedures, not generic login steps. If no procedure exists above the confidence threshold, the ticket routes directly to L2 engineering without attempting autonomous resolution.
For resolvable L1 issues, Atlas Reasoning Engine does not just surface a knowledge article — it executes the fix directly via configured API integrations. Password and credential resets: calls your identity provider (Okta, Azure AD) API to reset credentials and sends the new access link to the employee's verified contact method. Software access requests: calls the relevant application's provisioning API to grant the requested access role. Account unlocks: calls the identity provider to unlock the account. VPN troubleshooting: runs a connectivity check and sends the employee a guided resolution sequence. The agent verifies that the action completed successfully before closing the ticket.
When the resolution action completes, the agent sends a personalised notification to the employee via Slack (or their configured channel): a clear description of what was done, any steps they need to take to complete the fix (e.g. click the password reset link, restart their VPN client), and a confirmation prompt asking them to verify the issue is resolved. If the employee confirms resolution, the ticket is closed automatically and a CSAT prompt is sent. If the employee reports the fix did not work, the ticket is escalated to L2 engineering with full context: the resolution that was attempted, what the employee reported, and the complete ticket history.
Tickets that cannot be resolved autonomously — infrastructure issues, security incidents, complex network problems, hardware failures requiring physical intervention — are routed to the IT engineering queue via OmniChannel. The routing considers: issue type (network vs security vs hardware — different specialist queues), urgency tier (P1 incidents bypass the queue and page the on-call engineer directly), and the affected employee's role (C-suite tickets are escalated immediately regardless of tier). The assigned engineer receives a pre-prepared ticket summary: issue type, affected system, resolution already attempted (if any), employee's exact description, and Einstein's diagnostic assessment. Handle time drops because the engineer does not start cold.
Every resolved ticket — autonomous and human-resolved — is logged with: resolution action taken, time to resolution, employee confirmation, and CSAT score. Tickets resolved by engineers that do not have a matching KB article are flagged as documentation gaps. Engineers are prompted to document the resolution procedure after closing the ticket — a Prompt Builder draft is pre-generated from the ticket details to reduce the documentation burden to a 5-minute review and approve rather than starting from scratch. Over time, new L1 resolution types are added to the knowledge base — expanding autonomous resolution coverage. Weekly gap reports are posted to the IT team's Slack channel.
Resolution execution
Executes L1 resolution actions directly via API — credential resets, access provisioning, account unlocks — rather than just surfacing a knowledge article.
Ticket triage
Classifies every ticket by type, urgency, and affected system within seconds — determining whether it goes to autonomous resolution or specialist engineering.
Resolution procedures
Houses documented resolution procedures for all common L1 issues. Semantic search ensures the right procedure is retrieved regardless of how the employee describes the problem.
Action execution
Okta and Azure AD integrations enable the agent to execute credential resets, account unlocks, and access provisioning directly — without human involvement.
Employee interface
Primary channel for ticket submission (/it-help command), resolution notifications, confirmation prompts, and CSAT surveys.
Engineer routing
Routes complex tickets to the correct specialist queue — with skill-based routing, P1 incident paging, and role-based priority logic.
Security
Critical for an IT context where actions include credential management. All API calls and employee data stay within Salesforce's trust boundary.
Kovil AI scopes, builds, tests and deploys this Agentforce configuration end-to-end. You do not touch Agent Builder until it is live and resolving tickets autonomously.
The autonomous resolution library is configured during implementation based on your IT environment. Common autonomous resolutions include: password and credential resets, account unlocks, MFA reset, basic software access grants for configured applications, VPN troubleshooting (connectivity checks and guided steps), standard onboarding system provisioning, and known configuration issues with documented fixes. Issues requiring physical intervention (hardware failure, network infrastructure), security investigation, or systems without API integration are always routed to engineering.
The agent calls your identity provider's API (Okta or Azure AD) to initiate the reset — it does not handle or transmit the new credential. The reset link is sent directly from the identity provider to the employee's verified secondary contact method (personal email or backup phone number configured in their profile). The agent never sees the credential itself. All API calls are made within the Einstein Trust Layer — no credential data leaves Salesforce's security boundary.
P1 incidents are detected by urgency classification signals: keywords like 'everyone is affected', 'whole team can't access', 'security breach', 'data leak', or 'production is down'. P1 tickets bypass the standard queue and trigger immediate on-call paging via PagerDuty integration (or your configured on-call system). The page includes the full ticket context. Simultaneously, a Slack alert fires to the IT leadership channel. The agent does not attempt autonomous resolution for P1 incidents — it escalates immediately.
Initial autonomous resolution rate depends on the breadth of the configured action library and KB coverage at launch — typically 50–60% in week 1. As engineers document new resolution procedures (prompted by the KB gap detection system) and new API actions are added, the rate improves. Most clients reach 70%+ within 90 days. We track resolution rate, CSAT, and KB coverage as standard post-launch metrics and report weekly.
Book a 30-minute discovery call. We'll audit your current helpdesk ticket distribution, identify your L1 automation opportunity, and scope a 3-week fixed-price Agentforce implementation.
3-week sprint · Fixed-price · Production-grade · Post-launch support included