Example Co. · IT Service Desk · 2026-10-19 to 2026-10-23
Each task is scored on how consistently the team does it, how much time it takes and how much is redone. 2 tasks are ready for an agent now; 2 need a standard procedure first. Together the designs below would give back about 26–49 hours a week ($73K–141K a year).
5 of these designs are validated with the team: decisions, exceptions and what “done” means come from their answers. What the team told us →
| Task | Verdict | Readiness | Team h/week | Back h/week | Build with |
|---|---|---|---|---|---|
| Password resets and account unlocks | Ready for an agent | 78 | 29 | 10–16 | Copilot Studio, Power Automate |
| Provision software and access requests | Ready for an agent | 81 | 18 | 6–10 | Power Automate, Copilot Studio |
| Remote troubleshooting sessions | AI assist (a person acts) | 92 | 20 | 2–4 | Microsoft 365 Copilot, Copilot Studio |
| Status update emails to users | Standardize first | 72 | 25 | 4–11 | Copilot Studio, Power Automate |
| Triage and route incoming tickets | Standardize first | 84 | 20 | 3–9 | Copilot Studio, Power Automate |
Password resets and account unlocks · Ready for an agent · size M, about 3–5 weeks to build
Validated with 6 team members · 3 of 5 decisions described
An agent reads the locked-account ticket, checks the caller's employee ID and manager name against the directory, and prepares the reset or unlock. After the analyst approves, it applies the change and drafts the standard resolution note. It removes the hand-copied username and the knowledge-base lookups.
Provision software and access requests · Ready for an agent · size M, about 3–5 weeks to build
Validated with 6 team members · 8 of 9 decisions described
A Power Automate flow with a Copilot Studio agent for unclear cases prepares each approved ServiceNow access request: it picks the Entra group and license, checks approvals and license stock, then shows the analyst one change to approve. After approval it applies the change, writes the work note and sets the state.
Remote troubleshooting sessions · AI assist (a person acts) · size M, about 2–4 weeks to build
Validated with 6 team members · 2 of 5 decisions described
The company asked to keep remote troubleshooting with people, so this is an assistant, not an autonomous agent. It prepares likely causes and past fixes before the Quick Assist session and drafts the Work notes afterwards. The analyst runs the session and approves everything saved to the ticket.
Status update emails to users · Standardize first · size M, about 3–5 weeks to build
Validated with 6 team members · 4 of 7 decisions described
When a ticket appears or changes in the queue, the agent reads the ServiceNow ticket and drafts a status or resolution email in the team's standard structure. The analyst approves or edits it in one step, then it is sent through Outlook and the note is saved on the ticket.
Triage and route incoming tickets · Standardize first · size M, about 3–5 weeks to build
Validated with 6 team members · 6 of 7 decisions described
Reads each new ServiceNow ticket in the unassigned queue and proposes a category, priority and assignment group using the team's rules. An analyst approves the proposal in one step before the agent writes it to the ticket. P1, security and VIP tickets are flagged for the analyst.
Readiness weighs how many cases follow the same steps, how many team hours are at stake, whether the work moves data between systems, and how often steps are redone. “Standardize first” means the team does the task too many different ways for an agent to follow; its training guide comes first and its savings here are discounted. Designs use the company's own tools and its approval rules; build times are estimates for an experienced builder, including testing and a pilot. After launch, HourBack measures the same task again to show what came back.