MOBILE
ENTERPRISE
SERVICE MANAGEMENT
BUSINESS POC
The Intelligent Service Desk
THE INTELLIGENT
SERVICE DESK
A mobile-first service management redesign with AI embedded at the point of action.
Role: Sole Designer
|
Type: Business POC
|
Timeline: 6 weeks
|
Year: 2026
Overview
A Business POC reimagining the mobile experience for IT service management agents. The existing app reproduced the desktop product in a smaller frame, dense, form-heavy, and rarely used. This project reframed mobile around the moments of work where it actually matters: brief, interrupted, high-pressure decisions, and field workflows where mobile is faster than desktop. AI was treated as a supporting layer, deployed in two distinct modes depending on what the agent needed: embedded at the point of action for in-flow decisions, and conversational on demand for broader queries.
Context and Problem
Enterprise IT service agents work in fragmented contexts, triaging tickets, resolving incidents, and receiving assets, often away from a desk and often in flows where mobile is genuinely faster. But the existing mobile app reproduced the desktop experience in a smaller frame: dense fields at equal weight, no SLA visibility, no priority signals, and a 12+ field asset receipt form with validation only on submission.
Testing with internal IT professionals confirmed what the brief implied. Agents weren't using the app, not because mobile didn't fit their workflow, but because it didn't add anything the web portal didn't already offer. When they needed a ticket summary, they were manually copying content into external AI tools to get one. Asset receipt wasn't easier on mobile than on desktop, so they didn't bother.
The brief was open: explore what a mobile-first, AI-embedded service management experience could look like.
Four moments of work mobile needs to serve
01
Interrupted check-in
Between meetings.4 min to triagewhat's urgent
→ Attention Now
02
On-the-move resolution
In transit. Critical ticket needs action now.
→ AI Summary
03
Returning agent
Days away. Needs a fast queue overview.
→ Aviator Chat
04
Field receipt
On-site. Receiving hardware faster than web portal.
→ receive assets
Roles and Constraints
Sole designer working alongside product managers and engineering leads. POC scope: three core flows. AI capabilities were aspirational but technically feasible (on-device ML, existing data infrastructure). Decisions were grounded in workflow analysis, stakeholder interviews, and testing with internal IT professionals - the core user group for this tool.
Approach
I reframed the problem. Instead of "redesign the mobile app," I asked: what does mobile design need to look like for people who can't sit at a desk and for people who could, but where mobile is faster? While competitors in the ESM space were embedding AI into operational workflows, the existing mobile experience lagged in surfacing intelligence at the point of action.
This reframing surfaced four moments of work mobile actually needs to serve, each mapping to a screen or AI mode:
The interrupted check-in: between meetings, 4 minutes to triage what's urgent. Served by the Attention Now carousel and embedded queue signals on Home.
The on-the-move resolution: in transit, a critical ticket needs action now. Served by the Ticket Detail with embedded AI Summary card.
The returning agent: picking up context after days away, needing a fast queue overview. Served by Aviator's conversational summary across the full queue.
The field receipt: on-site or at the desk, receiving hardware faster than the web portal allows. Served by Receive Assets with photo-driven AI pre-fill.
Three screens - Home, Ticket Detail, Receive Assets, and two AI modes were designed to serve them.
Key Decisions
DECISION 01
Designing for interruption, with AI in two postures
Decision: Restructure Home around AI-prioritised urgency, with AI in two distinct postures, a dedicated Attention Now section showing AI's strongest opinion (2–3 highest-priority tickets), and subtle AI signals woven into existing queue views (flagged counts, SLA-at-risk indicators).
Why:
The existing screen surfaced metric counts but no signal about what actually needed action, wasting mobile's most valuable real estate
Trust in AI is variable, not binary, a single dominant surface forces an all-or-nothing relationship with it
Dual placement lets agents engage at the level of trust they're ready for
Attention Now is curated and prominent; queue signals are subtle nudges, distinct roles, no confusion
Risk: Visual complexity. Mitigated by distinguishing the two roles clearly, Attention Now is gradient-styled and prominent; queue signals are understated.
Outcome: Agents see what needs action in 2 seconds while retaining full queue control.
Top of screen reclaimed for AI-prioritised urgency. Queues retain agent control with subtle AI signals layered in.
Top of screen reclaimed for AI-prioritised urgency. Queues retain agent control with subtle AI signals layered in.
Home BEFORE
home AFTER
DECISION 02
Two modes of AI: embedded at the point of action and conversational on demand
Decision: Deploy AI in two distinct modes. Embedded AI lives inside the ticket detail, a Summary card surfacing 1-line context, SLA countdown, and a suggested action grounded in past resolutions. Aviator, a dedicated conversational tab, handles broader queries: summarising multiple tickets, drafting resolution comments, answering cross-queue questions.
Why:
An agent triaging between meetings needs context without leaving the ticket, a chat assistant alone breaks that flow
An agent with two minutes to think wants to ask a question across their entire queue, embedded AI alone can't serve that
The workaround: agents were manually copying ticket content into external AI tools to get summaries. Embedding AI inside the ticket eliminates that detour
Aviator is available when the agent chooses, never surfaced mid-flow without intent, no fragmentation
Risk: Two AI surfaces could confuse users about which to use. Mitigated by clear separation, embedded AI is a card within the ticket, Aviator is a dedicated bottom nav tab.
Outcome: Agents engage with AI at the level of trust and time they have. Embedded AI compresses context into a 5-second glance. Aviator handles the broader thinking agents previously did manually.
AI Summary card
AI at the point of action
AI at the point of action
Aviator - AI on demand
Aviator - AI on demand
SLA countdown
Suggested action
Natural language query
Queue summary
Quick action chips
Resolution draft
DECISION 03
Photo-driven asset receipt
Decision: Replace the form-first flow with photo-scan as primary entry, using AI to pre-fill device details from the captured image. Required fields drop from 12+ to two.
Why:
IT professionals flagged asset receipt as a pain point, the mobile experience offered no advantage over the web portal, so receipts went undocumented until agents were back at their desks
Asset receipt is a flow where mobile genuinely beats desktop, the web portal requires device-switching to capture a photo
Photo-first design eliminates that friction and turns a form into a confirmation step
Risk: AI photo detection has accuracy limits. Mitigated through editable confirmations, confidence indicators, and a clear "Enter details manually" fallback.
Outcome: Asset receipt moves from data entry to confirmation. Mobile beats desktop for the first time in this flow.
Asset receipt becomes confirmation, not data entry. Mobile beats desktop web portal in this flow.
BEFORE
AFTER
Asset receipt becomes confirmation, not data entry. Mobile beats desktop web portal in this flow.
Projected Outcome
Projected impact based on workflow analysis, internal IT professional testing, and broader ITSM industry data. As an unshipped POC, directional rather than validated at scale.
15 - 30% reduction in average handling time - AI Summary compresses context agents currently reconstruct manually
10 - 20% reduction in SLA breaches - visible countdowns and AI-driven prioritisation replace manual urgency discovery
40 - 60% faster asset receipt vs the existing web portal - photo-driven AI pre-fill eliminates device-switching
20 - 35% increase in sustained mobile adoption among field and at-desk agents
Dual AI adoption pathway - two-mode placement removes the binary trust barrier; agents who distrust embedded suggestions can engage with Aviator first
Reflection
This project was a study in restraint. The easy version was an AI chat assistant; the harder version was treating AI as ambient infrastructure - two modes serving different cognitive states, not competing for the same attention.
The workaround agents had already built, manually pasting tickets into external AI tools, made the design direction feel less like a bet and more like a formalisation of something already happening.
What I'd do differently: extend validation to field agents across different operator contexts, and explore the end-user side, where AI could help requesters describe issues at submission, reducing triage burden downstream.
b2b saas case study
Digital Marketplace - 6x faster storefront setup
→
HCI Case study
AuDHD Tinted Love - Inclusive design framework
→
Designed with love by Swati ✦
© 2026 Swati Bhat
Designed with love by Swati ✦
© 2026 Swati Bhat
Best viewed on desktop
This portfolio is designed for desktop and tablet. Open it on a larger screen for the full experience.
Best viewed on desktop
This portfolio is designed for desktop and tablet. Open it on a larger screen for the full experience.
Swati Bhat · Senior UX Designer
Swati Bhat · Senior UX Designer