A plain-English walkthrough.

You’re hiring an AI agent — think of it as a tireless digital assistant — that has one job: read your inbox, learn how you write, and prepare reply drafts for you to review.

It is not a chatbot you open and talk to. It is nota separate app. It works in the background, inside the email you already use (Outlook, Gmail, Tencent Exmail, NetEase, Aliyun — or any other IMAP mailbox). You don’t change how you work. You just find drafts waiting for you when you open your email.

Every hour the agent checks your inbox for new mail. It looks at each message and makes one of three decisions:

• Must reply — a buyer or vendor is asking you directly. The agent writes a draft in your voice and saves it to your own Drafts folder.
• Should read — a status update, FYI, or multi-day situation you should know about. No draft. Flagged so you can skim it.
• No attention— routine approvals where you’re just CC’d, automated notifications, bulk mail. Quietly tagged so you can ignore them without opening.

You open your email the way you always have. The drafts are waiting. You read, tweak a line or two, hit send. That’s the whole workflow.

This is the part that makes MerchGent AI different from any generic AI email tool. We don’t hand you a chatbot and ask you to write instructions to it. We do the opposite — we teach the AI to sound like you.

Day 1 — the deep-learning bootstrap.When you first connect your mailbox, the agent reads your last three months of sent email (with smart sampling, so a high-volume sender doesn’t get a voice profile dominated by one busy week). From that corpus it builds a structured voice profile: your typical greetings, your sign-offs, your average sentence length, the specific words you reach for, your characteristic phrases. It also learns the things your work actually runs on — the recurring projects, the people you correspond with most, the products and accounts and terminology your day-to-day touches. This pass takes about 24 hours.

A different voice for every relationship.Most people don’t write the same way to everyone — your tone shifts depending on who’s on the other end. Formal with leadership, looser with peers. Direct with vendors, more relationship-aware with clients. Concise with people who know the context, fuller with people who don’t. The voice profile captures every register you use and picks the right one for each draft, automatically.

Week-by-week refinement.Every line you change before sending teaches the agent. If you swap “Best regards” for “Thanks,” in every reply, it learns. If you always trim the agent’s opening sentence because you don’t bother with pleasantries on a busy morning, it learns. By week 4 most customers say drafts already sound like them.

Weekly tune-up.Every Sunday the agent runs a voice-refresh pass — it takes your last week’s sent mail and edited drafts and rolls them into the profile so it stays current with new patterns in your business (new vendors, new buyers, new product lines, a change in how you write because life happens). You don’t have to do anything. It just stays current.

Triage isn’t a one-off classification — it’s how the agent organises your mailbox for you, automatically and on demand.

Every new message gets tagged. The agent applies a label (or category, for Outlook) to every inbound it sees: MerchGent: Must Reply, MerchGent: Should Read, or MerchGent: Routine. You can filter by these in your mail client to look at only the things that need you, or sort them into folders with one rule. Most customers set up an Outlook rule that moves “Routine” messages to a sub-folder so the inbox view stays clean.

On-demand inbox cleanup.When you ask, the agent can also run a backfill pass — go through your historical inbox (last 30, 60 or 90 days) and apply the same labels retroactively. Useful when you onboard MerchGent AI and want last quarter’s mail organised the same way the new mail is. Just email hello@merchgent.com with the date range, and the agent processes it on the next hourly tick.

Custom rules.If you want certain senders or subjects always routed into a specific lane (e.g. “always routine for the QC certification bot at this vendor”) we can configure that on your account — drop us a note with the rule and we’ll set it up. A self-serve rules editor is on the roadmap.

Picture Tuesday morning. You walk in at 9am. Your inbox shows 38 new messages from overnight — vendors in Asia who worked all day while you slept, buyers in Europe who got to their desk an hour before you got to yours.

The agent has already been through them all. (It ran at 4am, 5am, 6am, 7am, 8am.) Here’s what you find:

• ~20 messages tagged “routine.” Routine acknowledgments, sample approvals you’re CC’d on, automated notifications. You can scroll past them or skim-read in a minute.
• ~10 messages tagged “should read.” A production slip from one of your factories. A FYI from a buyer’s merchandising team about an upcoming RFP. A shipping forwarder’s update on a container at customs. You scan these in five minutes — you don’t need to reply, you just need to know.
• ~8 messages tagged “must reply.” For each one a draft is already waiting in your Drafts folder. Sample-ready confirmations. A buyer’s push for a quote. A pricing pushback to a factory that bumped overnight. The drafts are in your voice — direct to the factories, polished to the buyer.

You open your Drafts folder. Six of the eight drafts you read and send with one or two small tweaks. Two of them you write from scratch — they’re strategic, they need your judgement, you’d never let AI do them anyway. That’s fine. The agent saved you the 80% that doesn’t need your judgement.

Total time: maybe 30 minutes. Without MerchGent: 2–3 hours of typing the same kinds of replies you type every morning.

MerchGent AI can prepare a weekly work report that lands in any inbox you choose every Monday morning — yourself, a partner, a manager, whoever needs the bird’s-eye view.

The report aggregates what your team is dealing with across all connected mailboxes and surfaces the operational picture:

• Key issues being addressed — production slips, quality escalations, customer complaints, payment disputes, the factory that bumped a price again. Synthesized from the must-reply lane across the week.
• What the team handled — number of drafts created, by which user, by counterpart side (upstream vs downstream), with the top vendors and buyers your team communicated with most.
• Pending escalations— items the agent flagged that haven’t been responded to yet, so the recipient can follow up directly.
• Trend signals— has the volume of escalations from a particular factory been creeping up? Has a buyer’s tone shifted? The agent surfaces patterns the inbox owner might be too close to see.

The report is generated entirely from operational metadata + a short LLM synthesis — never the raw email bodies. The recipient sees the WHAT and WHY but not the actual email content, which keeps trust between team members intact.

To enable it, open your dashboard and find the “Work report” card. Enter the recipient address and choose your cadence (weekly Monday morning is the default — daily and bi-weekly are also available). Clearing the recipient turns the report off.

We’re deliberate about scope. Things MerchGent AI does not do:

• It is not a CRM.We don’t replace Salesforce, HubSpot, your custom ERP, or your spreadsheet. We don’t track deals as deals, we don’t hold pipeline. We just read mail and draft replies.
• It is not a chatbot.You don’t open it, you don’t talk to it, you don’t write prompts to it. There’s no “ask MerchGent AI a question” interface. It works in the background, in your existing email.
• It is not a calendar or scheduler.If an email says “let’s set up a call,” the agent will mention scheduling in its draft, but it won’t book the meeting. That’s your job and your tools’ job.
• It will not send email on your behalf — ever. The send permission is not part of our integration with Outlook, Gmail, or any other provider. We physically cannot send mail from your account. Every reply passes through your eyes before it leaves.
• It will not show your email to anyone. Not to MerchGent AI employees, not to outside companies, not to AI training datasets. See the Privacy Policy for the full mechanics.

Start with a 14-day free trial on Starter.A card is required to start, but you’re charged $0 today. The agent runs against your real inbox for the full 14 days — including the deep-learning bootstrap — so you see real drafts before you pay anything. Pro and Power have no trial — they’re billed when your agent goes live.

Three flat-rate tiers. Same intelligence in every tier — same voice learning, same all-providers IMAP, same drafts-only guarantee. The only difference is how often the agent ticks and how aggressively it sweeps your inbox:

• Starter — $29/month. Agent runs every 60 minutes, new mail since the last tick. For solo operators and light-volume inboxes.
• Pro — $59/month. Agent runs every 30 minutes — twice as fast as Starter. For active inboxes with time-sensitive replies.
• Power — $99/month. Fastest cadence (every 15 minutes) plus a full-inbox sweep of the last 7 days on every tick, and a twice-weekly voice refresh. For high-volume work.

Every tier includes one mailbox; add extra mailboxes at $19/month each (hourly cadence) after signup.

Cancel anytime.No contracts, no minimum term. Cancel during the trial and you’re never charged. Cancel later and you keep access through the end of the billing period; we delete your data within 7 days.

Two ways in:

• Try the public demo — paste a sample inbound email, see what the agent would do. No signup, no payment, takes 30 seconds.
• Start your 14-day free trial on Starter — connect your inbox, wait 24 hours, find drafts in your Drafts folder tomorrow morning. Card required, $0 today; Starter ($29/mo) auto-renews on day 15, while Pro ($59/mo) and Power ($99/mo) have no trial and bill at activation. Cancel anytime; we delete your data within 7 days.