AI Chatbot Assistant (MVP) that merges many AI Assistants (2022)
Bot of Bots - A Concierge Chatbot - Discovery & Design
Bot of Bots - A Concierge Chatbot - Discovery & Design
There are about 10 services and chatbots internally that offer different capabilities and experiences to have a conversation with users. I designed for the goal to create a concierge chatbot to provide a single, seamless and personalized user experience to the visitors and drive efficiency and effectiveness and improve the overall experience. Concept design, design exploration on conversation design to improve user engagement.
My Role
Our Team
My Deliverables
Intro
Role: Lead Conversation Designer & AI Strategist Timeline: Dec 2020 – May 2021 Platforms: MyPepsiCo portal, Microsoft Teams Partners: ServiceNow, OneReach.ai, HR Ops, IT, GBS
Overview PepsiCo had 20+ internal chatbots across HR, IT, Finance, Legal, and Operations. The experience was fragmented, intent coverage was inconsistent, and help desks were overloaded with repeat questions. Objective: Design Chester, a unified concierge assistant (“bot-of-bots”) that provides one trusted entry point, understands employee intent, and routes actions across systems. Promise: One voice, one entry point, enterprise-wide services. Insert image: Hero mock of Chester interface (from PPT)
Business Problem & Solution
Problem & Constraints Fragmentation: Employees had to remember which bot handled which request. Inconsistency: Different tone, flows, and error handling across bots. Throughput: High volume of repetitive tickets driving up support costs. Technical constraints: Must integrate with ServiceNow (tickets, KB), myidM, Google Enterprise Search, and myHR (Alight/Lisa). Insert image: Ecosystem map of existing chatbots (PPT slide)
Research & Discovery
Research & Discovery Methods 12 stakeholder interviews (IT, HR, Finance, Legal) Audit of 20+ chatbots (overlap, gaps, fallbacks) Conversation journey mapping for top 10 employee tasks Integration workshops (ServiceNow NLU, OneReach.ai) Key Insights Employees want one reliable assistant with context memory. Interop > new UI: value comes from orchestration, not replacement. Governance (tone, flows, testing) is essential for scale. Insert images: Conversation journey map Findings summary / insights slide
Solution
Solution Chester = Concierge + Orchestrator + Context Core capabilities Multilingual support and tone consistency Intent taxonomy spanning HR/IT/Finance + robust fallbacks Context carryover across web & Teams ServiceNow: ticket creation, status, KB surfacing Smart handoff to domain bots or humans Insert image: Architecture diagram showing Chester orchestrating child bots
What I designed
Conversation Design (What I Designed) Voice & Tone system for HR, IT, Finance (empathetic, efficient, precise) Intent & entity model (150+ utterances → 20 top-level intents) Dialog flows using a 5-step pattern: Greeting → Recognition → Clarification → Action → Confirmation Error recovery patterns and escalation standards Reusable dialogue components (confirmations, disambiguations, status updates) Insert images: Tone of Voice matrix Intent taxonomy Representative flow diagram
Interaction & UI
Interaction & UI Compact, enterprise-friendly layout for dense info (tickets, KB cards) Clear system status & provenance (e.g., “Answer sourced from ServiceNow KB”) Accessibility: WCAG 2.1 AA color & typography, keyboard nav, screen-reader labels Insert image: UI screens / component library
Validation & Iteration
Validation & Iteration Methods: Wizard-of-Oz sessions, telemetry on fallback and completion, CSAT pulse. Results (pilot) Fallback rate: 32% → 7% after utterance/intent refinements Task completion: 58% → 86% for top HR/IT flows CSAT: 3.2/5 → 4.4/5 Insert image: Before/after metrics chart
Business Impact
Business Impact (6 months post-MVP) Average resolution time: 3.2 days → 1.1 days Help-desk load: −40% automated via Chester Pilot adoption: 68% active users “Chester didn’t just save time — it changed how we think about digital support.” — Brion Goudreau, MyPepsiCo Sponsor Insert image: KPI dashboard / sponsor quote slide
My contributions
My Contributions Led conversation strategy, tone, and standards Designed intent taxonomy, entities, and 300+ dialogue variants Authored flow specs and error-recovery patterns Drove ServiceNow and OneReach.ai alignment with design Set up testing protocol and analytics tracking
Conducted internal workshops to align HR & IT around a shared conversation design language • Mentored 3 designers transitioning into AI-first product design • Collaborated with ServiceNow engineers to define reusability standards for future bots
Organisation Impact
fter Chester’s success, PepsiCo adopted the Conversational Governance Framework across 5 departments. It became the reference model for all future bots — influencing over 30 design and IT stakeholders.
Whats next
What’s Next Voice & mobile expansion (field operations) Adaptive personalization (preferences, recent tasks) Conversational AI Center of Excellence for enterprise reuse
My learnings
“Chester taught me that designing an AI assistant is not just about interaction quality — it’s about designing the organizational muscle to sustain intelligence at scale.”
Design Process
Customer Journey Touchpoints
Customer journey for registration process in various regions with sms feature and Store owner registrtaion high level user journey were updated to the digital ordering touchpoints. Painpoints and opportunities for storeowners were discussed with stakeholders and users - Online and offline touch points.
UI UX Heuristics Review
Overall UI UX heuristics reviews and also focusing on Product card - units per case Product cards, Basket, Pricing - all prices with VAT, Campaigns were analysed to come up with recommendations.
Multiple design concepts
With the insights from the review and feedback from stakeholders and users the design was conceptualised leveraging the existing flow and pattern. Various iterataion and concepts were analysed and received feedback to update them.
Digital Ordering Touchpoints
Analysed and updated the flow for the painpoints and opportunities for storeowners - Online and offline touch points.
