cpdeol
← Back to Projects
AI/Agentic Systems

AI-Assisted Customer Onboarding Platform

87% of customers onboard without any human touch — cutting time from 3 days to 8 hours and achieving 95% satisfaction.

87%

fully autonomous completions

2-8h

onboarding time (was 2-3 days)

95%

customer satisfaction

5k+

onboardings processed autonomously

Role

Lead Technical Business Analyst — AI Implementation

Timeline

Q4 2023–Q1 2024 · 4 months

Delivery context

AI ImplementationProcess MappingUATChange ManagementJIRA

The Problem

Customer onboarding required multiple manual steps with an average time of 2-3 days. High abandonment rate due to friction. Inconsistent experience across customers. The operations team needed a clear definition of what the AI agent could decide autonomously versus what required human judgment.

My Contribution

I led the requirements analysis and process mapping for this AI-assisted onboarding initiative. I documented every step of the existing human agent workflow using UML process models, identified which steps were strong automation candidates versus edge cases requiring human judgment, and defined the functional specifications for the AI decision boundaries. I designed the UAT shadowing protocol used to calibrate autonomous completion thresholds and ran structured change management sessions with the onboarding operations team to prepare them for the new workflow. I also defined the monitoring requirements and KPI framework for tracking autonomous completion rates post-launch.

Process

  1. Workflow Mapping

    Documented every step a human agent performs using UML process models, identifying automation candidates versus judgment-required edge cases.

  2. Requirements

    Defined functional specifications for AI decision boundaries — what the system could auto-approve, re-ask for, or escalate to a human agent.

  3. UAT Design

    Built a shadowing protocol to run 500 test conversations and calibrate confidence thresholds for autonomous versus escalation paths.

  4. Change Management

    Ran structured sessions with the onboarding operations team to prepare them for the new workflow and establish feedback loops for continuous improvement.

The Solution

Requirements-led AI implementation: end-to-end human workflow documentation, automation boundary definition, UAT shadowing design, change management programme, and monitoring KPI framework — ensuring the AI agent was built to the right specifications from the start.

Results

  • 2–3× faster onboarding: 2-3 days → 2-8 hours
  • 87% of customers complete without human intervention
  • Less than 1% escalation rate
  • 95% customer satisfaction score
  • 5,000+ onboardings processed autonomously
Key learning
The hardest requirements problem was defining 'confidence' — what threshold of AI certainty justified autonomous action versus human escalation. Getting compliance, operations, and engineering aligned on that definition in a workshop, before any model was trained, was what made the production system trustworthy enough to run at 87% autonomy.

Tech Stack

AI Platforms

Claude APIAI/LLM Integration

Process

UMLProcess ModelingVisio

Tools

JIRAConfluenceSharePoint

Methodology

Agile/ScrumUATChange ManagementBRD

Related

How this project connects to the rest of my work.