Email automation for the inbox your customers write to

Two hundred messages wait in info@. Someone reads them top to bottom, forwards a third of them, and answers the easy ones with a paragraph they have typed four hundred times.

An urgent complaint sits at position ninety because it arrived at eleven on Saturday night. Nobody knows it is urgent until it is a phone call.

Two colleagues answer the same customer within an hour, saying slightly different things.

The inbox stays where it is, on Gmail or Microsoft 365. Make connects it to the rest, and an AI model reads each message before a human opens it.

• Classified on arrival. Topic, urgency and customer identified, then the message is tagged and assigned to whoever owns that kind of work.
• Answers drafted from your own knowledge. Your help docs, past replies and order data produce a response ready to send in one click.
• Tickets where they belong. Anything needing a trail becomes a ticket in Zendesk with an owner and a due time, visible to the whole team.
• Escalation with context. Angry or high value messages reach the right manager in minutes, with the customer history attached.

Attachments get read too, so an invoice or a purchase order becomes structured data instead of a file somebody has to open.

Bad news goes out written by a person. A delay, a price increase or a mistake on your side deserves someone who can take responsibility in the next sentence.

For the first weeks every draft waits for approval, which also trains the system on what your team actually sends. Routine categories move to automatic send later, one at a time and only with your sign off.

An online retailer rebuilt its support this way in the Pepper case.

Classification and routing are usually live in two weeks, because they change nothing the customer sees. Drafted replies take another two or three, since that part needs your knowledge base in order.

We test on a month of real historic email in a controlled environment, then compare what the system would have sent against what your team actually sent. The reasoning behind that is in the process section.

Start with the AI and operations audit. We look at volume, categories and how long an answer really takes today, then hand you a ranked plan in two working days with its fee credited to the first project.

Export a month of email and whatever help documentation exists, even if it is out of date. When the answers live only in the head of one person, writing them down is the first job and process and SOP design covers it.

These builds often run on Make, described in the Make page.