Guide · AI
Automate customer inquiry replies without sending anything unchecked
An inquiry comes in through your website or by email, and then it sits: the right person is in a meeting, the reply needs one more piece of information, and by evening the inquiry is three hours old. Speed matters here. In research published by Harvard Business Review in 2011, 23 percent of 2,241 US companies never responded to an inquiry submitted through their website. A second analysis in the same article showed that companies that reached out within an hour were far more likely to have a real conversation with a decision maker. Even so, if you automate customer inquiry replies with no review at all, the reply may promise something nobody has checked. This guide shows a middle path: AI sorts the inquiry and drafts a reply, and a person approves it in one click.
What a good first reply does
The first reply doesn't have to settle everything. It should show that the inquiry was understood and name the next step. A good first reply:
• responds to the actual request instead of just confirming receipt,
• asks the questions you can't move forward without, such as quantity, timing or location,
• names the next step, such as a suggested meeting time or when a proposal will follow,
• promises nothing that still needs checking, above all prices, dates or feasibility.
Which inquiries are a good fit
Good candidates are inquiries that come in often and in a similar form: questions about process and availability, requests for documents, meeting requests, and inquiries that usually lack the same details. Don't prepare automatic drafts for complaints, price negotiations, legal questions, or anything that puts an existing customer relationship at risk. The workflow recognizes these and forwards them straight to the right person, without a draft.
Automate customer inquiry replies in five steps
Step 1: Route everything to one place
Inquiries from your contact form, from a shared mailbox such as info@, and from forwarded emails all land in one place. That is usually a shared mailbox in Outlook or Gmail.
Step 2: Sort the inquiry
The AI reads the inquiry and notes what it is about, how urgent it is, which details are missing, and whether it belongs to the cases that always go to a person. The result is a short summary with a category.
Step 3: Answer from your own material
So the AI doesn't make things up, it only gets your own material: a page describing your services, frequently asked questions with answers, reply templates, and the link to book a meeting. The instruction says, in essence: answer only with information from these documents. If something is missing, ask a follow-up question instead of guessing.
Step 4: Send the draft for approval
The draft reply goes to the responsible person, together with the original inquiry and the summary. They read it, change a sentence if needed, and approve it. For a good-fit inquiry, that often takes about a minute.
Step 5: Send and follow up
After approval, the reply goes out from the right mailbox. If there is no response after a few days, the workflow reminds you to follow up. Our guide on automatic follow-up reminders in Outlook shows how that works.
How to build it with Microsoft 365
With Power Automate, the automation service in Microsoft 365, the workflow looks like this:
1. Trigger: "When a new email arrives (V3)" for your mailbox. For shared mailboxes, there is a separate trigger, "When a new email arrives in a shared mailbox (V2)".
2. Sort and draft: The "Run a prompt" action runs an instruction that you create in the Power Automate prompt editor. It receives the email text and your material and returns the summary and the draft.
3. Approval: The "Start and wait for an approval of text" action sends the draft to the responsible person. They can edit the text in the approval before approving it.
4. Reply: Once approved, "Reply to email (V3)" sends the approved text as a reply to the original email. For a shared mailbox, enter its address in the "Original Mailbox Address" field. Otherwise the reply goes out from the personal mailbox.
The prompt action requires a Power Automate Premium license and capacity in AI Builder credits or Copilot Credits. Starting November 1, 2026, new or renewed licenses no longer include AI Builder credits, and usage is billed through Copilot Credits instead. Test the workflow first with a test mailbox and a few real, anonymized inquiries.
How to build it with n8n
n8n is a workflow tool that runs as a cloud service or on your own servers. The workflow is the same: a trigger reacts to new emails in Outlook or Gmail, and an AI step sorts the inquiry and writes the draft. For approval, the Outlook node offers the "Send and Wait for Response" operation. It sends the draft to the responsible person and waits for their response. With the "Approval" response type, the person can only approve or decline. If they should be able to edit the text, choose "Free Text" or "Custom Form". The same approval also works through Teams, Slack or Gmail. Then the node sends the reply to the original email with "Reply". Often the simpler route is to have the workflow save the reply as a draft in the mailbox, so the person reviews, edits and sends it themselves.
If you don't want to build anything yet
You can test the core idea by hand. Put your frequently asked questions and reply templates in one document. Then use a business AI account, such as ChatGPT Business or Copilot Chat, to draft a reply for every new inquiry for a week. You will see how good the drafts are and what information is missing. That is the best preparation for a fixed workflow.
What happens to the data
Inquiries contain names, contact details and often details about the project. They belong in an AI service with a data processing agreement, not in a personal account. According to Microsoft, prompts in Power Automate run on Azure OpenAI and are not used to train the models. With n8n, it depends on which AI provider you connect. For more, see our guide Can your team put client data into ChatGPT?
What the AI shouldn't decide
Prices, dates, discounts, commitments about feasibility, and the decision whether to take on a job at all stay with a person. The AI may ask and suggest, but not commit. Also decide which types of inquiries never get a draft and are forwarded directly.
Where this reaches its limits
The workflow depends on the material behind it. If answers to common questions are missing, the AI writes a lot of follow-up questions and the drafts don't help much. With very few inquiries per week, a fixed workflow rarely pays off. And if several people approve, you need a clear rule about who gets which inquiry. Otherwise the draft just waits again.
How aenuma helps
aenuma builds this workflow in your environment, for example with Microsoft 365, n8n, or Make. We test it with real inquiries from your day-to-day work. Together, we agree in advance which inquiries get a draft and who approves. Then we introduce the workflow to your team and, if you want, keep supporting it.
Frequently asked questions
Will customers notice that AI prepared the reply?
If your material is good and a person reviews the draft, the reply reads like any other reply from your team. Whether and how you mention the use of AI is up to you. What matters is that someone on your team owns the reply.
Can the AI reply without approval?
Technically, yes. For a simple acknowledgment with a booking link, that can make sense. For anything with substance, we recommend approval, at least until you have seen over a few weeks how good the drafts are.
Do we need a CRM for this?
No. The workflow works with a mailbox and your material. If you use a CRM, it can also create the inquiry there.
Sources
The Short Life of Online Sales Leads (Harvard Business Review, March 2011)
Office 365 Outlook connector (Microsoft Learn)
Use your prompt in Power Automate (Microsoft Learn)
FAQ for prompts (Microsoft Learn)
Microsoft Outlook node (n8n Docs)
By Joshua Schütt, founder of aenuma. Last updated: October 2026.
How long do your inquiries wait for a reply today?
Tell us how inquiries reach your company and who answers them. We will tell you which part can be prepared and what is better left with your team. Not sure where to start? The Time Sink Check shows which task makes a good first workflow.
Book a call