← All case studies
Customer emailDrafted reply

Inbox full of questions. Drafted replies out.

Your team answers the same handful of questions every week, while customers sit waiting. We draft the reply the moment the email lands, so answers go out faster and people feel looked after.

▶ Skip to see it run

The problem

The same questions, week after week.

Most of a support inbox is the same handful of questions, asked again and again, that the business's own docs already answer. Someone reads each one, finds the answer, and retypes it. It eats 2 to 4 hours a day, and customers wait in the meantime.

Customer email
The same FAQs, all day
THE BOTTLENECK
A person, re-answering FAQs
Looking it up, retyping it
Reply sent
Hours later, if at all
0hrs
a week spent re-answering the same questions
2–4hrs
a day the support inbox swallows
Hours
customers wait for answers your docs already have

Why it hurt

Slow support costs more than time.

Skilled time on autopilot

Your team spends hours typing answers that already sit, word for word, in your own help docs.

Customers left waiting

Every email joins a queue. The busier you get, the longer people wait, and the less looked-after they feel.

Quality drifts

Rushed, copy-pasted replies vary in tone and accuracy depending on who happens to be clearing the inbox.

The solution

One agent, reading every message.

We built an agent that reads every new message, works out what it's about, searches the knowledge base, and drafts a reply in the business's own voice, ready for a one-click send. Anything it's unsure about is escalated to a human, with the sources attached.

Customer email
Read the moment it lands
AUTOMATIC
The triage agent
Classifies, searches & drafts
Drafted reply
In your voice, one click to send

Customers get an accurate answer in minutes, not days. The team just reviews and sends, and the inbox stops being a chore.

Live demo

See it run for yourself.

Watch a support thread come in and a grounded reply come back: the agent classifies the intent, searches the knowledge base, and drafts the response, flagging anything it's unsure about.

Live · how it decides
Pick an email
jordan@example.co.nznew message
Do you deliver to rural addresses?
Hi, quick one — do you deliver to rural delivery addresses, and is it the same price as standard?
or pick an email above
1
Email arrives
2
A real customer message?
YescontinueNoIgnore (spam / auto-reply)
3
Classify intent, urgency & sentiment
4
Complaint, refund or upset customer?
YesEscalate to a humanNocontinue
5
Search the knowledge base
6
Grounded, confident answer?
YesDraft reply → one-click sendNoEscalate (with draft + sources)
Outcome

Every escalation and every human edit teaches the system, so the knowledge base grows and it answers more on its own over time.

Sample data. We use a fabricated support thread here to show the pipeline; real customer emails are never published.

Before & after

Same questions. A happier inbox.

Before
  • 2 to 4 hours a day clearing the inbox
  • Customers wait hours, sometimes days
  • The same questions answered over and over
  • Tone and accuracy vary by who replies
  • Support can't scale without more hires
After
  • A reply drafted the moment an email lands
  • Customers answered in minutes, not days
  • Repeat questions handled automatically
  • On-brand, grounded in your own docs
  • Volume climbs, the hours don't

The headline

20 hours a week

of repeat replies handed back to the team, and customers answered in minutes instead of days. Faster support, happier customers, the same size team.

Agency · Paid media
Google Ads reporting, automated

Every client's weekly reporting built straight into Asana, handing 30 to 40 hours a week back to the team.

Read the case study →
Sales · Proposals & decks
Proposals written while the deal's warm

Branded proposals and decks generated from a call transcript or outline, instead of built slide by slide.

Read the case study →

Faster answers, happier customers. Let's talk.

An introduction to learn about your business and see how much of your inbox could answer itself, well.

← See the other automations we've built