Hours Saved. Proof It Works.

AI-generated product content needed statistical validation across 41,000+ products before rollout

• −No experimentation framework existed to measure AI content impact
• −Manual review of product descriptions at scale was impossible
• −Stakeholders needed data-backed evidence, not assumptions
• −Risk of deploying AI content that could increase return rates

Built a full experimentation framework with A/B testing, MDE calculations, and statistical validation

• +Designed experimentation methodology with minimum detectable effect sizing
• +Built SQL analytics layer processing 14.2M+ records for measurement
• +Created AI-generated product summaries with Claude API and Databricks ML pipelines
• +Implemented A/B testing via Optimizely with rigorous statistical controls

Statistically validated +2 percentage point improvement with £5-10M potential annual impact

Onboarding new clients takes 45+ minutes of manual admin work every time

• −Copying details from emails to spreadsheets by hand
• −Sending the same welcome email over and over
• −Creating folders and scheduling calls manually
• −Dropping the ball on follow-ups when busy

End-to-end automation from enquiry to first meeting — no copy-paste needed

• +Automatic CRM entry when enquiry arrives
• +Personalised welcome email sent instantly
• +Client folder structure created automatically
• +Calendar link sent with availability

From 45 minutes down to 2 minutes per new client

Processing invoices manually takes 60+ minutes per batch with frequent errors

• −Extracting details from PDFs by eye
• −Cross-checking against purchase orders manually
• −Missing discrepancies until it's too late
• −Updating spreadsheets by hand

AI extracts, matches, and flags — you just review what matters

• +Automatic PDF data extraction
• +Smart matching against purchase orders
• +Discrepancy detection and flagging
• +Auto-update to tracking system

From 62 minutes to 10 minutes per batch, with fewer errors

Support tickets take 25 minutes each, with repetitive research every time

• −Reading and understanding each message from scratch
• −Searching knowledge base for answers
• −Writing similar responses over and over
• −Logging everything manually

AI drafts responses from your knowledge base — you review and personalise

• +Automatic categorisation of issue type
• +Knowledge base search and synthesis
• +Draft response with relevant info
• +Human review for personalisation

From 25 minutes to 3 minutes per ticket, happier customers

Creating one piece of content takes 2-3 hours of research and writing

• −Hours spent researching each topic
• −Staring at blank pages trying to start
• −Editing takes longer than writing
• −Formatting and scheduling is tedious

AI handles research and first draft — you add your voice and publish

• +Automated topic research and synthesis
• +AI-generated outline based on your style
• +First draft in your brand voice
• +Human refinement and personal touches

From 160 minutes to 48 minutes per piece, 3x more output

Managing email takes 60+ minutes daily and still things slip through

• −Checking inbox constantly throughout the day
• −Sorting and categorising manually
• −Writing similar responses repeatedly
• −Missing important emails in the noise

AI sorts, drafts, and flags — you just review and send

• +Automatic email categorisation
• +Priority flagging for urgent items
• +Draft responses for common queries
• +Archive processed emails automatically

From 60 minutes to 12 minutes daily, nothing gets missed