DTC FOOD · UK

60% of CS tickets resolved without a human — CS team scaled sub-linearly with revenue.

A $14M DTC food brand had 3 CS reps and an 8,400-ticket month. The COO had spent 36+ hours setting up Gorgias macros and rules. Half were broken on Shopify order edits and variant changes. The Gorgias bill came in at $1,400 for the month and the team was still drowning. The founder's choice was hire two more CS reps (£90k/yr combined) or rebuild the workflow. We built a tool that reads each customer's full context — Shopify orders, tracking, past tickets, warranty — drafts a reply in the brand's voice grounded in the actual refund policy, and for clear-cut cases issues refunds and address changes directly. The team did NOT need to hire the two reps.

Brand name withheld. CS deflection measured against a baseline of 6 weeks pre-pilot at the same ticket volume.

60%

AUTO-RESOLVED

−68%

FIRST-RESPONSE TIME

1.5×

TICKETS/CSR THROUGHPUT

4 wk

PILOT

DTC FOOD · UK · 2026 · DETAILS REDACTED

THE PROBLEM

Three CS reps doing the same conversation 200 times a week.

The brand's CS team was excellent at the hard tickets — allergen questions, recipe issues, real complaints. They were getting buried in 'where is my order' and 'I bought the wrong flavour, can I exchange'.

Gorgias auto-macros caught some. Most still needed a human, because each customer's context was slightly different and the brand cared about voice.

WHAT WE BUILT

A tool that reads context, then acts within policy.

Step 1: load the brand's actual policies — refunds, exchanges, address changes, warranty — as the source of truth. The tool can quote them when answering.

Step 2: learn the brand's voice from 200 hand-picked past replies. In blind tests, reviewers couldn't tell drafts from real ones.

Step 3: set a confidence threshold. Above it, the tool issues the refund or address change automatically. Below it, a human reviewer gets a draft with two suggested edits.

It doesn't try to be clever. It quotes the policy, references the order, and gives the customer back five minutes. My team writes the hard ones now.

THE OUTCOME

Same team, 1.5× volume, customers happier.

CS tickets auto-resolved0% → 60%+60 pts
First-response time (median)1h 50m → 35m−68%
Tickets per CSR per week650 → 9751.5×
Headcount avoided (2 CS hires)~£90k/yr saved+£90k
Gorgias plan tierPro → Starter$2.6k/mo cut

WHAT WAS MESSY

Two things we got wrong before we got them right.

First version was too cautious — confidence threshold set too high, only 25% auto-resolved. We calibrated against a labelled set of 400 real tickets to find the right cutoff.

Policy was scattered across Notion + a PDF + tribal knowledge. The audit week was spent consolidating it into one source. Worth it: that doc is now the operating manual for the whole CS team.

WHAT'S NEXT

The same copilot is being extended to subscription change requests.

Subscription change requests (skip a month, swap flavour, pause) are now flowing through the same engine. The brand expects another 15–20% deflection on those by Q3.

More field notes

All field notes →

DTC SKINCARE · EU