Let AI Handle Your Inbox: Gmail and Outlook Setup
Email is the highest volume, lowest value task in most sales jobs. Matching, forwarding, filing, and drafting replies eats the morning and produces nothing you can point at. That is exactly the profile of work worth handing over.
The realistic outcome of inbox automation is not an empty inbox. It is an inbox where the messages that need you have already been separated from the ones that do not, and where the reply is drafted before you open the thread.
What inbox automation actually does
Three jobs, in order of how much pain they remove.
- Triage. It reads incoming mail and sorts it into the categories you defined: leads, clients, support questions, invoices, noise.
- Drafting. It writes a reply to the ones that need one, in your voice, for you to send or adjust.
- Filing. It moves the rest to the right folder, or flags it for review, so nothing important sits under three newsletters.
Decide the scope before you build anything
The first decision is not technical, it is political. Which mail does the agent see, and what is it allowed to do with it?
A sensible starting scope for most teams is incoming leads and client replies, held in draft. That covers the work that most often slips and keeps the agent away from invoices, legal threads, and anything personal.
Write the exclusions explicitly. Legal terms, contracts, refund threats, and anything from someone whose name you do not recognise in a sensitive thread should go to a human, untouched. The instruction should say so in those words, because a vague boundary gets tested in week one.
Connect the account and write the routine
The setup is short. Connect the Gmail or Outlook account you actually send from, then create a routine with a schedule and a precise instruction.
A working instruction has five parts. What to read, over what period, how to classify, what to draft, and what to leave alone. Adding a summary line at the end makes it reviewable in seconds.
One pass worth including early: a forward or copy of every incoming message into the agent's view. Without that step the agent is working blind on new mail and only sees what you remember to show it.
The rules that make it useful
The difference between a helpful agent and an annoying one is the rule list.
- Never send without approval while the routine is new.
- Never reply to a legal, billing, or complaint thread. Route it.
- Never promise a date, price, or discount. Draft around it and flag it.
- Always include the thread history in the reply so the client is not reading a response with no context.
- Always produce a one line summary of what was handled.
Keep the list short enough to remember. Five rules that are followed beat twenty that are skimmed.
Drafting replies without sending nonsense
The failure mode people fear is an automated reply that embarrasses them. The way to avoid it is to make drafting and sending separate decisions for the first month. The agent drafts, you approve.
Quality improves quickly if you correct in the tool rather than outside it. When a draft is wrong, tell the agent why in a sentence and ask it to remember. Voice instructions are easier and faster than typing, and the agent applies the correction to every later draft.
A useful test is to ask for drafts against your last thirty days of mail, read them in one sitting, and count how many you would have sent. That number tells you whether it is ready to draft unsupervised.
Test it on your own inbox first
Before it touches a client, run it on your own address for a week. Send it the awkward emails: the complaint, the introduction request, the one that needs a price. Watch what it does with each.
Then check the boring things. Does it handle a thread with nine messages in it without losing the point. Does it notice a name change. Does it stop when the answer requires a decision it is not allowed to make.
When it passes, expand the scope by one category at a time. Expanding everything at once is how people end up with an agent that files an important client email into a folder they never open.
If your replies are landing in the wrong folder, why emails go to spam is the next thing to fix. And if the list feeding the inbox is stale, cleaning it with an agent handles it.
Frequently asked questions
Is it safe to connect a work inbox?
It is a real question and it deserves a real answer. Check where the data is stored, whether it is used to train models, and who on your team can see the account. Then start with one mailbox rather than the whole company.
Does this actually speed up replies?
It changes what your time is spent on. Instead of reading and writing, you are reading and approving. Most people find that cuts the morning email block substantially, and the replies that go out are more consistent.
Will it miss anything important?
It will, occasionally, which is why the summary line exists. Skim the summary daily. If something important was misfiled you will see it in the summary rather than discovering it a week later.
The point of inbox automation is not to remove you from email. It is to make sure the messages that need you arrive already understood and already drafted.
If you would rather spend your time selling than configuring, start a free 7-day Dubb trial.
Let the agent draft in the thread you already use
Dubb AI reads Gmail and Outlook, drafts the reply and can attach a video, so the answer goes out while the lead is still warm.