OmniDesk

Enterprise Bot

Experimental AI Contact Center Platform

0→1 product strategy for an AI intelligence layer for contact centers.

An experimental 0→1 initiative exploring how AI could evolve the contact center from disconnected channels and workflows into a continuously improving system — combining omnichannel ticketing, AI agents, human agents, routing, integrations and operational intelligence with a learning-in-the-loop mechanism.

Role: UX Architect · Product Strategy · 0→1 Product Design
Team: CEO · COO · Partnerships · Sales & Delivery · CSMs · Engineering
Domain: Enterprise AI · Contact Center

00 — Key highlights

Where I contributed

UX Research

Discovery into contact-center workflows, operational friction, existing infrastructure and where AI creates real value.

AI & Emerging Tech

Agentic AI patterns, AI-assisted workflows, human-in-the-loop systems, RAG and tool calling.

Agentic Pipeline

Mapped intent → context → knowledge → tools → reasoning → action → response → human → outcome.

AI / ML Collaboration

Translated model behaviour, pipelines and constraints into usable, configurable experiences.

RAG & Knowledge

Made retrieval-augmented knowledge behaviour understandable to enterprise users.

API & Tool Integration

Let AI agents move beyond conversation and perform real work in enterprise systems.

Genesys Native

Complemented established infrastructure through native Genesys integration — no rip-and-replace.

Cross-functional

Connected vision, customer needs, feasibility and go-to-market across leadership and engineering.

Strategic Product

Positioning, solution definition, demos, presales strategy, roadmap and market exploration.

01 / 09

01 — Opportunity

From conversational AI to an intelligent contact center

The product direction centred on conversational AI. Instead of treating conversations as isolated interactions, OmniDesk explored how conversations, tickets, AI agents, humans, routing, integrations and outcomes could form one connected system — working with existing infrastructure, not replacing it.

Unified agent console with inbox, conversation thread and AI suggested replies

What if AI could become an intelligent layer across the entire contact-center operation?

02 — Core question

How might we create a contact-center system where AI doesn’t just automate interactions, but learns from what happens after every interaction?

Connect the entire interaction

Conversations, tickets, channels, agents and workflows in one operational experience.

Keep humans in the loop

Agents intervene, correct, approve and improve AI — automation is never a black box.

Every interaction teaches

Outcomes, feedback and corrections continuously improve future AI behaviour.

03 — Personas

Designing for an ecosystem, not a single agent

Customer

Resolve an issue quickly through the channel they already use.

Needs

Fast response
Consistent information
Easy escalation
Context across channels

Pain points

Repeating information
Being transferred
Slow resolution

AI Agent

Resolve predictable requests and assist human agents.

Needs

Relevant knowledge
Tools & integrations
Clear guardrails
Human feedback

Pain points

Missing context
Incorrect knowledge
Ambiguous intent

Human Agent

Resolve complex cases while staying in control of AI-assisted interactions.

Needs

Complete context
AI recommendations
Ticket history
Ability to correct AI

Pain points

Switching systems
Repetitive tasks
No visibility into AI decisions

Operations Manager

Improve service quality, efficiency and operational performance.

Needs

Cross-channel visibility
Routing & workload
AI performance
Resolution trends

Pain points

Fragmented data
Manual reporting
Hidden recurring issues

01 / 04

04 — User flow

End-to-end experience flow

Intake

CustomerChannelCapture interactionIntent & contextAI Agent

Decide

Can AI resolve?

YesAI action / response

NoHuman agent

Ticket / caseResolution

Learn

Customer & agent feedbackLearning loop

KnowledgePromptsTools & workflowsAI behaviour

Back to AI Agent

User flow — the complete lifecycle of a customer interaction. Resolution isn’t the end; it feeds the next interaction.

05 — UX architecture

An intelligence layer, not another isolated chatbot

Channels, AI services and Genesys sit underneath one experience layer that agents, supervisors and admins all work from.

Channels

VoiceChatEmailOther

Omnichannel

Conversation & ticket contextIntent understanding

AI intelligence layer

AI AgentRAG / KnowledgeTools & APIsAgentic pipelineGuardrails

Operations

Ticket / caseRoutingResolution

Two-way via Tools & APIs

Existing infrastructure

GenesysEnterprise systemsExternal APIs

Escalation & takeover

Human-in-the-loop

Human agentAgent assistTakeover

Learning loop

Interaction dataHuman feedbackOutcomesAI performanceContinuous improvement

Improves knowledge, agent & pipeline

UX architecture — OmniDesk as an AI intelligence layer connecting channels, AI, humans and existing contact-center infrastructure.

06 — Agentic pipeline

Bridging AI/ML architecture and UX architecture

What does the AI know?
What can it do?
What did it decide?
When does a human take over?
What happens after?

Messages from multiple channels routed through an AI node into organised queues

RequestIntentContextRetrieve knowledge · RAGAgent reasoning

Tool / action required?

YesTool / APISystem response

Confidence & policy check

PassGenerate responseHelpHuman agent

Customer outcomeFeedback & learningRefines retrieval & reasoning

Agentic pipeline — technical AI behaviour translated into concepts users can understand.

07 — Research

Dual-track discovery

Connecting what users needed with what the technology could realistically enable.

Research workshop table with sticky notes and workflow diagrams

User & workflow

Contact-center workflows
Interaction patterns
Agent responsibilities
Ticket lifecycle
Escalation & routing
Operational systems

AI & technology

Agentic workflows
RAG
Tool calling
AI orchestration
Human-in-the-loop
Learning loops

08 — User journey

From request to continuous learning

01Customer initiates interaction
02AI understands request
03AI retrieves knowledge & context
04AI uses tools / performs action
05Resolution possible?
06Human agent intervenes
07Ticket / case resolved
08Customer & agent feedback
09Outcome analysed
10AI, knowledge & workflow improved
11Applied to future interactions

User journey — from customer request to continuous learning. Step 11 loops back to step 02.

09 — Learning in the loop

AI that is observed, corrected and improved

Each improvement lands in knowledge, prompts, tools, workflows or AI behaviour — and applies to the next interaction.

Learningin the loop

Observe
Intervene
Learn
Improve
Apply

Abstract continuous feedback loop illustration

10 — Designing for complexity

It wasn’t the screens. It was the relationships.

CustomerChannelIntentAI AgentKnowledgeToolWorkflowHuman AgentTicketResolutionFeedbackLearning

Progressive complexity

The right level of information at the right moment.

Context over configuration

Explain why something happens, not just what is configured.

Connected objects

Customers, tickets, agents, tools and outcomes stay linked.

Human control

AI intervention and escalation are explicit, never hidden.

11 — Integration strategy

Work with Genesys, not against it

Existing contact centerGenesysOmniDesk intelligence layer

AI agentsKnowledge / RAGTools & APIsHuman-in-the-loopTicketing

AutomationOperational outcome

Integration strategy — an intelligence layer on top of the infrastructure organisations already invested in.

12 — Collaboration

Aligning business, customer and technical teams

CEO & COO

Product vision, strategic direction and business opportunity.

Head of Partnerships

Integration possibilities, ecosystem and strategic partnerships.

Head of Sales & Delivery

Customer requirements, demos, implementation and presales feedback.

Head of CSMs

Customer pain points, adoption challenges and operational feedback.

Engineering & AI/ML

Feasibility, AI pipelines, APIs, RAG and system architecture.

01 / 05

13 — Presales

A complete product story

Before

“Here are our AI features.”

After

“Here is how AI can participate across the contact-center lifecycle.”

ProblemChannelAI understandingAI actionHumanResolutionLearningBetter experience

14 — Outcomes

A broader proposition, faster to demonstrate

Operations dashboard showing SLA, queue volume and agent performance

Faster integration exploration

An integration-led architecture made it easy to show OmniDesk alongside existing infrastructure, including Genesys — no replacement conversation needed.

Faster presales activity

Sales, Partnerships and Delivery gained one connected narrative: Customer → AI → Human → Ticket → Resolution → Learning.

Stronger strategic direction

Expanded the proposition from conversational AI toward an AI intelligence layer for contact centers.

Technical-product alignment

RAG, agentic pipelines, APIs and learning mechanisms translated into coherent product experiences.

01 / 04

15 — What I learned

The UX of AI isn’t just the interface where AI responds.

It includes everything around the model — context, knowledge, tools, actions, human intervention, outcomes, feedback and learning. For enterprise AI, the designer’s role becomes architecting how people, AI, data and workflows interact as one system.