// the tuesday guide
The info@ inbox that sorts itself: labels, priorities and drafts ready to go
You open your email in the morning and info@ has a bit of everything: a quote request, a supplier invoice, a customer asking where their order is, three newsletters, a certified email, an ad for a course. To figure out what actually matters you have to read them all, and meanwhile the one email that counted gets buried under the rest.
This week we take that job apart: how to get your inbox to split itself by topic, push urgent emails to the top and have a reply already written for the questions you always get. With one rule that is not up for debate: no email goes out without your OK.
What the info@ inbox looks like today, done by hand
In most small businesses the shared inbox works like this: whoever has time opens it, reads from top to bottom, answers what they can and leaves the rest “to read later.” When two or three people share it, nobody is quite sure who has already replied to what.
Time gets lost in three places. The first is figuring out what is there: to know whether an email is a quote request or a promotion, you have to open it. The second is rewriting the same replies over and over: opening hours, delivery times, documents to send, how to pay. The third is finding things again: the supplier’s March invoice, that customer’s request from two weeks ago.
Anyone who has managed a shared inbox knows the typical mistakes: the quote request spotted three days late, the complaint discovered when the customer calls back angry, the same question answered by two people in two different ways.
What gets automated and what stays with you
The mechanical part gets automated: splitting emails by topic, flagging the urgent ones, drafting a reply for the questions that always come in the same way, filing invoices and documents in the right place.
The decision always stays with you: what to tell a customer, what price to quote, how to handle a complaint, what to promise. Automation can write the draft, but hitting send is yours. One wrong reply sent automatically to a customer costs more than all the time you saved in a month.
With the tools you already have, and when you need something custom
Half the work gets done without any AI. Gmail has filters that apply a label based on the sender or a word in the subject line; Outlook has rules that move emails into a folder. With these tools, which you already have, you can sort every email you can recognize by sender: regular suppliers, certified email, the bank, couriers, newsletters.
AI is for what a rule can’t read: emails written by customers. “Hi, I’d like to know how much it would cost to…” has no recognizable sender and no fixed word in the subject. A language model, on the other hand, understands that it is a quote request, that the customer is in a hurry, and can write a first reply. Today the small models, which are built for exactly this kind of repetitive work, cost very little: we covered them in last week’s AI news.
You connect the inbox to the AI with a no-code automation tool, or with a small custom program. Custom makes sense when the volume of email is high, when the reply has to pull data from your management software (an order status, availability, a list price) or when several people share the inbox and you need to know who owns what. It is the kind of work AI agents for businesses do: they read, classify, prepare, and leave the final word to a person.
An example from start to finish: a quote request comes in
Take any workshop, lab or studio, with an info@ inbox and two people who read it. Here is what happens when an email arrives.
1. The email arrives. A new customer writes in: they ask for a quote on a job, say they would need it by the end of the month, and leave a phone number.
2. The rules go first. The sender is not a supplier, not a certified email, not a newsletter: no filter touches it, so it moves on to the AI.
3. The AI classifies it. It reads the email and assigns it a label, “Quotes,” and a priority: there is a deadline, so it rises to the top of the list. If it had been a complaint, an alert would also have gone to your phone.
4. The AI prepares the draft. Using your template replies, it writes a first response: it says thank you, confirms the quote is on its way and asks for the two pieces of information that are always missing (measurements, quantities, address). It leaves it in your drafts, already addressed to the customer.
5. The person decides. Whoever opens the inbox finds the email at the top, with the label and the draft ready. They read it, change a sentence, hit send. Two minutes instead of ten.
6. The loop closes. After sending, the email moves from “Quotes” to “Awaiting reply,” so in three days you can see right away who hasn’t answered yet.
How to start in one afternoon, and the mistake to avoid
Hour one: pick your categories. Scroll through the last hundred emails in info@ and sort them by hand into five or six groups: usually quotes, support, suppliers, admin, advertising, plus a “To review” folder for everything else. Any more categories than that will confuse the machine, and you too.
Hour two: the filters. In Gmail or Outlook, create a rule for every sender that keeps coming back: suppliers, the bank, certified email, couriers, newsletters. That alone already cleans up a good part of the inbox, with no AI at all.
Hour three: the template replies. Write the five replies you send most often, in your own voice. They are the base the AI will start from for its drafts: if you write them well, the drafts will come out close to how you write.
Then, once the filters are working, you add AI for customer emails. For the first few weeks there is one rule: when it isn’t sure, the email goes to “To review,” and once a week you check what it put in the wrong place and correct it.
The mistake to avoid is turning on automatic replies. It is the strongest temptation, because the drafts look good from day one. But all it takes is one wrong reply, a made-up price or a promise you can’t keep, sent to a customer in your name, to lose the trust that a tidy inbox was supposed to earn you. Drafts, yes. Auto-send, no.
In short
Your info@ inbox can sort itself: a few clear categories, email filters for every sender you can recognize, and AI only for emails written by customers. Urgent emails rise to the top, the most serious ones also reach your phone, and for the usual questions you find a draft ready to go. The most important part stays with you: reading, correcting and hitting send. You can get started in one afternoon with categories, filters and template replies, and auto-send never gets turned on.
Frequently asked questions
Do you need AI to sort email automatically?
Not for all of it. Emails you can recognize by sender, like suppliers, the bank, certified email, couriers and newsletters, can be sorted with the Gmail filters or Outlook rules you already have. AI is for emails written by customers, which have no fixed sender or keyword to recognize.
Can AI reply to customer emails on its own?
Technically, yes, but that is the mistake to avoid. AI should prepare the draft and leave it in the inbox, already addressed: a person reads it, fixes it if needed and sends it. One wrong reply sent automatically to a customer costs more than the time you saved.
How many categories should you use for an info@ inbox?
Five or six at most, for example quotes, support, suppliers, admin and advertising, plus a “To review” folder for unclear cases. With too many categories the machine makes more mistakes, and it gets harder for you to find your emails too.
A half-hour chat, no strings attached: we look at your email together and see what can be automated
Let's talk →