Dotsy support· all mockups

How Dotsy support works. Report it, we check it, people fix it.

Bug reports come first: one click, a screenshot, and the details we need are attached for you. Our support agent (an AI) checks every report before Zare or Luka spend time on it. Questions get an instant answer from the help centre.

1. When someone reports a bug

From "Report a bug" in the help bubble, on dotsy.ai or in the dashboard.

1
A customer reports it

Words, a screenshot they can mark up, and the page details, attached for them.

2
Lands in Slack

One card per problem. Repeat reports are merged.

3
HHermes checks it AI

Tries to reproduce it on a test site. Reads the logs.

?
Is it a real bug?

A verdict, with the evidence.

Yes, confirmedseverity and where to fix it
4
ZLZare or Luka fix it

One click on "Start fix". The customer hears "fix in progress".

5
Everyone hears back

"Fixed, live now", in the chat or by email, with the changelog link.

Not confirmedno person needed yet
4
Asks for more info AI

A short question in the same chat, then it checks again.

5
Or answers it AI

Not a bug? It explains and links the help article.

2. When someone asks a question

From the chat bubble on dotsy.ai or inside the dashboard.

1
A customer asks

"Why is my domain still verifying?"

2
AI assistant AI

Reads the help centre. Checks the account and domain, read-only.

?
Can it answer?

Or is it a confirmed bug, or a question only a person can answer?

Yesabout 9 in 10 chats
3
Answers with sources

Every answer links the help article it came from.

4
Solved

The customer rates it. Low ratings get a human look.

No: a real bug or questionabout 1 in 10
3
Sent to Slack

One message in #dotsy-support: summary, who, which site, what was tried.

4
HHermes digs in AI

Reads logs, finds the cause, drafts a reply. Can't send on its own.

5
ZLZare or Luka

Fixes it, then approves or edits the reply.

6
Customer hears back

In the same chat, or by email if they've left.

3. Every unanswered question grows the help centre

The same articles answer the chat, rank on Google and get quoted by ChatGPT, Claude and Perplexity.

A question with no article

The AI spots it: "5 people asked how to move a site between workspaces this week."

HHermes drafts an article AI

From the real chats: short answer first, steps, FAQ.

ZLA person approves

Approve, edit or dismiss, right in Slack.

Published and found

A public page, in the sitemap, indexed, cited by AI search.

dotsy.ai/help/move-a-site-to-another-workspace
Fewer escalationsEach new article removes a reason to wait for a person.
More search trafficHelp pages target the exact words customers use.
Cited by AI answersShort, factual answers are what AI search quotes.

Later: the same tools for our customers

Customers' own visitors report bugs on their sites. They see the reports in their Dotsy dashboard, and their AI agent can read and answer them over MCP.

See the idea

4. The ground rules

What keeps it safe and honest.

It says it's an AIAlways labelled. A person's replies look different and carry their name.
Read-onlyIt can look at your site and settings. It can't change anything.
People sign offNo AI-drafted reply or article goes out until Zare or Luka approves it.
Shares only what's neededThe team sees the chat and the checks it ran. Never passwords or payment details.
Press Walk me through it to step through both flows.