38 Hours a Week ofManual Support Work,Automated in 8 Weeks.
A £2.6M DTC ecommerce brand was losing its operations team to repetitive support tickets and manual data entry. We deployed an AI support agent with clear escalation rules, connected the systems it needed to answer from, and automated the supplier workflows behind it — cutting first response times from 12+ hours to under two minutes.
Free, no obligation — we reply within 48 hours.
Client DTC ecommerce · £2.6M annual revenue · small in-house ops team
A Team Buried in Work That Did Not Need a Person
The brand is a DTC ecommerce business generating £2.6M in annual revenue, with a small operations team handling customer support tickets, order status enquiries, return requests and inventory-related supplier emails by hand.
As order volume grew, that work grew with it. The team spent its days on repetitive, low-complexity tasks instead of the higher-value work only they could do — and every new order added to the pile rather than to the margin.
Three Places Manual Work Was Creating a Bottleneck
Discovery mapped where the hours were actually going. The pattern was consistent: the volume was high, the complexity was low, and nothing separated the two.
Support stalled outside business hours
Roughly 60% of support tickets arrived outside the team's working hours, so customers routinely waited 12+ hours for answers to simple questions like order status or return eligibility.
Manual data entry across disconnected tools
Orders, returns and supplier communications lived across Shopify, a helpdesk and email. Staff copied information between systems by hand for routine tasks, several times a day.
No triage, so every ticket cost the same
Simple and complex tickets were handled identically, with nothing routing what an automated system could resolve away from what genuinely needed a person.
Built and Deployed in Three Phases Over Eight Weeks
Integration first, then the agent, then the internal workflows behind it — in that order, because an agent that cannot read live order data is a chatbot.
- Weeks 1–2
Process mapping and systems integration
We mapped every recurring manual task across support, order management and supplier communication, then connected Shopify, the helpdesk and internal inventory data into a single automation layer so information moved between systems without anyone copying it.
one automation layer - Weeks 3–5
AI support agent deployment
We built and deployed a custom agent trained on the brand's policies, product catalogue and order data to handle order status, return and exchange eligibility, and common product questions directly — with escalation rules routing anything ambiguous or high-value straight to a human. It ran across email and live chat, covering the hours the team was not staffed.
escalation rules first - Weeks 6–8
Internal workflow automation
We automated the supplier and inventory workflows that had previously required manual entry — low-stock alerts, reorder point notifications and standard supplier order confirmations — taking the remaining routine admin off the operations team entirely.
supplier + inventory
Eight Weeks, Measured Against the Starting Point
| Metric | Before | After (8 weeks) |
|---|---|---|
| Manual hours per week on support and admin | ~46 hrs | ~8 hrs |
| Average first response time | 12+ hrs | Under 2 minutes |
| Tickets resolved without human involvement | 0% | 61% |
| Monthly operational cost (support and admin) | £5,400 | £2,300 |
The team went from being consumed by repetitive tickets and manual data entry to spending most of its time on the handful of genuinely complex cases the system escalated to them.
Automation as Triage, Not Replacement
The impact came from treating automation as a triage system rather than a full replacement. The agent handled the high-volume, low-complexity work instantly and reliably, while clear escalation rules made sure anything nuanced still reached a person — which is what cut cost and response time without sacrificing the customer experience on the cases that actually needed judgment.
What Would This Look Like in Your Operation?
Most teams are sitting on the same pattern: high-volume, low-complexity work that nobody has separated from the work that needs judgment. A free call is enough to find out which of yours is which.