AI Chatbot
A next-generation trading dashboard for retail investors
Problem Statement
Enterprise IT support teams handle thousands of ServiceNow tickets daily. L1 engineers manually hunt for SOPs, diagnose issues, and resolve tickets — a slow, repetitive process. When they can't resolve, escalation to L2 creates bottlenecks, with no intelligent system to surface relevant knowledge or assist in real time.
Solution
Initially conceived as an automatic ticket-resolving system pulling from ServiceNow, the project pivoted when the client restricted direct resolution access. The final solution is a ChatGPT-like AI assistant — an interactive chatbot that surfaces ticket context, relevant SOPs, suggested solutions, and L1 engineer details, enabling real-time conversational support. It marks Wipro Intelligence's expansion into AI-assisted IT operations across domains.
Key Results
SOP retrieval vs manual search
Down from manual searchReal-time ticket assistance
Context-rich L1 → L2 handoffs
Foundation for Wipro Intelligence across domains
Discovery & Research
We started by shadowing L1 and L2 support engineers across the ServiceNow workflow — observing how tickets arrived, how SOPs were hunted down, and where time bled out of the resolution cycle.
Voice of the Engineer
- ▪Engineers manually searched across disconnected knowledge bases for relevant SOPs.
- ▪Ticket context was scattered — related incidents, prior resolutions, and owner details lived in separate tabs.
- ▪L1 engineers escalated too early or too late, with no intelligent signal on when to involve L2.
- ▪Resolution knowledge stayed in people's heads rather than a reusable, searchable system.
Key Insights
Resolution speed was gated by knowledge discovery, not engineer skill.
Engineers didn't need automation that acted for them — they needed intelligence that worked with them.
A conversational layer could collapse search, context-gathering, and SOP retrieval into one interaction.
Define
How might we give every L1 engineer instant access to ticket context, relevant SOPs, and resolution guidance — so they resolve faster and escalate smarter, without replacing the human in the loop?
Ideation & The Pivot
An automatic ticket-resolving system that fetched tickets from ServiceNow, resolved them directly, and served SOPs for each issue — escalating to L2 only when the L1 engineer could not resolve.
The ServiceNow client was not willing to allow tickets to be resolved directly from Wipro's end — the autonomy we'd designed for was off the table.
We shifted from resolving for engineers to assisting with them — a ChatGPT-like AI chatbot that surfaces everything around a ticket: ticket details, related SOPs, suggested solutions, and the assigned L1 engineer's context. It works as a conversational model — users ask, it answers, they go deeper. Wipro Intelligence is now entering domain after domain with this same assistance-first model.
Solution Architecture
A conversational layer sits above the ServiceNow data and the SOP knowledge base. Every query is grounded in live ticket context and retrieved procedures — the assistant never hallucinates outside the enterprise's own knowledge.
Chat Interface Design
The interface mirrors a ChatGPT-style conversation — but every response is grounded in the enterprise's live ticket data and SOP knowledge. Engineers ask in natural language and receive structured, actionable answers they can immediately act on.
Show me details for ticket INC0123456.
Here's the ticket summary: • Issue: Email delivery failure on SMTP relay • Priority: High • Assigned L1: Priya Menon • Related SOP: SOP-NW-204 (SMTP Relay Troubleshooting)
What's the solution for this issue?
Based on SOP-NW-204: 1. Verify relay connector config 2. Check DNS MX records 3. Validate outbound queue health 4. Restart the SMTP service Would you like me to walk through any step?
Explain step 3.
Step 3 — Outbound queue health: Inspect the queue for stuck messages. If backlog exceeds 500, flush the queue and monitor retry rate. The L1 engineer can action this from the admin console.
Design Challenges
Grounding responses in real data
Designing the assistant so it never invents outside the enterprise knowledge base — every answer traces back to a real ticket or SOP.
From automation to assistance
Reframing the product after the ServiceNow constraint — shifting the mental model from "resolves for you" to "works with you."
Dense information, light interface
Ticket metadata, SOPs, and solutions are heavy. The chat surface keeps it conversational without overwhelming the engineer.
Trust in AI-assisted support
Engineers needed to trust the assistant enough to rely on it — transparency on sources and confidence was built into the response design.
Mockup Gallery
A side-by-side look at where the interface started and where it landed — the early auto-resolve concepts, then the conversational assistant the project became.
Initial Concept 01
Initial Concept 02
Initial Concept 03
Initial Concept 04
Initial Concept 05
Initial Concept 06
Initial Concept 07
Latest Design 01
Latest Design 02
Latest Design 03
Latest Design 04
Latest Design 05
Outcomes & Wipro Intelligence
The chatbot became more than a support tool — it became the template for Wipro Intelligence's assistance-first approach. The same conversational model is now being adapted to enter new domains, bringing AI assistance wherever knowledge discovery slows work down.
Prototype
A walkthrough of the conversational assistant in action — from ticket intake to guided resolution, showing how knowledge retrieval and escalation intelligence work together in a single flow.