Practical AI · Customer service
The customer who messaged you on Sunday doesn't show up in any report.
They sent a message at 9pm, waited, and by Monday morning they'd already bought from someone else. Your business didn't lose that sale. As far as it knows, that sale never existed.
The problem
The maths almost nobody does
Count how many messages come in after hours in a normal week. Multiply by 52. Then multiply by the share you usually close and by your average sale. The number you get isn't in any spreadsheet, because customers who give up don't complain. They just disappear.
Ring any bells?
- While one person is being helped, three conversations pile up in the queue
- Someone asks for a price, goes quiet, and no one ever follows up
- On weekends and holidays, messages sit and age until someone opens them
- Every channel has a different conversation and no one knows what was agreed where
- The person answering messages is the same person doing the work, so one of the two always stops
What changes
What your customer service starts doing
- Replies in minutes, on every channel where customers already talk to you
- Checks real data before confirming availability, price or timing
- Brings in a person when the topic calls for it, and that person steps in knowing what's already been said
- Logs every conversation, including the ones that didn't close
What's included
What we set up
-
An agent trained on your business
Your tone, your rules and your data. Not a bot with canned replies.
-
Real-time lookups
Stock, catalogue and calendar checked on the spot, instead of copied into the agent.
-
Handoff to a person
When the conversation calls for it, with the full history attached.
-
CRM logging
Who got in touch, what they asked for and what happened next.
-
Follow-up for drop-offs
One well-timed reminder for people who stopped halfway.
How we approach it
A fast answer is worthless if it's wrong
The customer service we build answers from what your business publishes and records. What isn't recorded, it won't claim, and it hands the conversation to a person instead of inventing a deadline, a policy or availability.
Items, prices and quantities aren't stored inside the agent. They're looked up in the system when the question comes in, because two sources of the same truth eventually disagree.
The path of a message
-
Message arrives
WhatsApp, Instagram or your website.
-
Understanding
What the topic is and what the person wants.
-
Lookup
Your real business data.
-
Reply
Or a handoff to a person.
-
Follow-up
The conversation is logged and anyone who stopped halfway gets a reminder.
Tools we've already used on this sort of project. We pick them once we understand the problem, not before.
- Instagram (DMs)
- Chatwoot
- Claude (Anthropic)
- Notion
- Telegram
Where we've used it
Real work, and we tell you what stage each project is at.
Client projects and our own technology. The label tells you which is which.
Virtual fashion consultant
An agent connected to a shop's real stock that checks availability before answering and notes the customer's name for the team. Currently in testing.
Read in Portuguese Our own technologyInstagram lead qualification
An automated conversation that qualifies people who ask for an assessment in DMs and moves qualified enquiries to WhatsApp.
Find out moreWhen it makes sense
This isn't the answer to every problem.
Telling you when it isn't right for you is part of the job. Sometimes the answer is something else, and sometimes it's leaving well alone.
It's worth it when
- More messages come in than your team can answer in the time customers expect
- Many of the questions repeat
- There's organised information for the AI to check, like a catalogue, stock or policies
- Messages arrive after hours and go unanswered until someone opens them
It isn't worth it when
- Message volume is small and one person handles it well
- Every enquiry needs an individual technical assessment
- Prices, timelines and policies aren't defined anywhere yet. In that case, getting organised comes first
Frequently asked questions
Will customers notice they're talking to a machine?
The goal isn't to hide it, it's to serve them well. The agent guides the conversation naturally, in your business's voice, with no generic canned replies. When the topic needs a person, it brings one in.
Does this replace my customer service team?
No. Your team stops spending the day on repeat questions and focuses on the people who really need a human. Small teams usually feel it as breathing room, not as a cut.
Do I need to learn a new tool?
No. We build it, set it up and hand it over working, with the conversation history in a place your team can already use.
Does it work on WhatsApp?
Yes. The right channel depends on where your customers already talk to you.
You might also look at
Software, automation and AI
Get in touch
How many messages did your business answer late this week?
Tell us how your customer service works today. We'll see whether the problem is volume, timing or information.
Rather use email? dreamharboragency@gmail.com