IT and service desk
A proposed workflow for it and service desk. Revised 17 September 2026. The roles below are design options, not claims that these agents are deployed or that multiple agents are necessary.
The business opportunity
A colleague cannot access an approved application. The service team needs to distinguish a missing entitlement from an outage or a device fault.
Start with the simplest useful solution
Compare the agent with a current knowledge article and a rules-based request form. Use the same representative cases and count human review, errors, integration, and ongoing operation. Add an agent only where choosing what to investigate or handling varied evidence improves the result.
Proposed responsibilities
- Agent
- Triage agent
- What it does
- Classifies, enriches with asset and entitlement context, routes
- Autonomy and learning position
- A3, L2
- Why here
- Highest volume, cheapest failure, cleanest labels. This is the eval-suite factory for everything else
- Agent
- Knowledge agent
- What it does
- Answers from a curated, owned corpus; refuses and escalates on weak evidence
- Autonomy and learning position
- A2 to A3, L1 to L2
- Why here
- The corpus is the product. Refusal behaviour is the quality signal
- Agent
- Provisioning agent
- What it does
- Executes bounded, reversible fulfilment: group membership, licence assignment, software deployment
- Autonomy and learning position
- A3, L1
- Why here
- Deliberately L1. Provisioning is where a learned shortcut becomes an entitlement error
- Agent
- Diagnostic agent
- What it does
- Correlates telemetry, reproduces, proposes a fix with evidence
- Autonomy and learning position
- A2, L2
- Why here
- Proposes only. The human applies
Start with one agent and ordinary tools. Separate a responsibility only when it needs different permissions, independent review, or its own operating schedule. The autonomy and learning positions are starting choices to test, not certification levels.
How the architecture supports the task
- Capability
- Knowledge and memory
- Role
- Direct. Curated per-intent corpora with named owners. The whole quality story
- Capability
- Tool gateway
- Role
- Direct. Tool servers wrapping the ITSM platform's governed APIs and the identity platform's provisioning APIs
- Capability
- Identity and policy
- Role
- Direct. Registry, agent identity, the deterministic gate on provisioning
- Capability
- Evaluation and release
- Role
- Direct. Ticket outcomes are labelled by resolution, which makes this the best-instrumented learning loop in the enterprise
- Capability
- Observability and audit trail
- Role
- Supporting. Standard audit trail baseline
- Capability
- Runtime
- Role
- Supporting. Retrieval-grade isolation is sufficient; no code execution needed
- Capability
- Human oversight
- Role
- Direct. Tier 2 becomes the exception queue, and its shape changes
Controls and human decisions
Never approve your own privilege increase. Check entitlement in the identity system before every change.
Give every tool call a task identity, a limited permission scope, and a recorded result. Before a write, check that the source record has not changed. Stop when required evidence is unavailable, the task budget is reached, or a proposal exceeds the authorised scope.
A useful investigation loop
Observe the exception, identify the missing fact, request that evidence, and check whether it changes the proposed action. Repeat only while there is a concrete unanswered question. End with a reviewed proposal, a confirmed result, or an explicit request for human help.
Economics and measures
Measure successful recovery, reopened tickets, time to human support, and specialist review minutes.
Agree the baseline with the process owner. Count value only after the outcome is confirmed. Compare the value of recovered time and improved outcomes with the full cost of tools, models, evidence preparation, supervision, and correction.
Limits to test
A ticket marked closed does not prove the colleague can work. Diagnose unfamiliar failures with a specialist.
Evaluate ordinary cases, uncommon failures, conflicting evidence, and recovery after an interrupted action before expanding authority.
External reading
The workflow above is the guide's proposal. These sources support the general investigation and risk-management methods; they do not validate the proposed business result.
- ReAct: reasoning and acting with language models, 2022. Research basis for alternating reasoning and tool use.
- NIST AI Risk Management Framework, 2023. General framework for managing AI risks.
Sources reviewed 17 September 2026.
Cite this page
Written by Murali Sid, Founder and AI Systems Architect, last updated 15 October 2026. Licensed CC BY-SA 4.0, so quote it freely with attribution.
Murali Sid (2026). IT and service desk. Agentic Architecture Skills. https://www.agenticarchitectureskills.com/library/blueprints/departments/it-and-service-deskMachine-readable version of this page: /library/blueprints/departments/it-and-service-desk.md
Blueprints
These proposed workflows show how business goals connect to agent responsibilities, controls, and measurable outcomes. Start with one bounded task. A list of responsibilities does not require a team of separate agents.
Customer service
A proposed workflow for customer service. Revised 17 September 2026. The roles below are design options, not claims that these agents are deployed or that multiple agents are necessary.