Five lessons, about an hour and ten minutes. This is where the job actually starts: getting into a brand's support tools without becoming the reason they get hacked, learning to move through a stacked queue fast without becoming careless, sounding like the brand instead of like yourself, knowing when a saved reply is right and when it's an insult, and closing out a shift so the next person isn't guessing.
Tolu's first message from Bumi Naturals' founder, Ada, after the trial shift was hired, was: "I'll just give you my Gmail password, easier that way." It is easier, for about a week — until Ada wants to log the founder back into her own account, or Tolu leaves the role and nobody remembers to change anything, or the shared password sits in a WhatsApp message forever, readable by anyone who ever sees that phone.
The right access is the smallest amount that lets you do the job, granted in a way that can be taken back without a scramble.
Most helpdesk tools — a shared inbox, a live-chat dashboard, a WhatsApp Business number — support proper multi-user access: your own login, your own name on replies, permissions that can be switched off in seconds. Where a tool doesn't support that, the fallback is a role email alias (support@brand.com forwarding to your inbox) rather than the founder's personal one. Where even that isn't set up yet, it is your job — politely, in the first week — to ask for it, because you are the reason it now matters.
"Before I start on the queue properly, could we set me up with my own login on [helpdesk tool] rather than the shared Gmail? It means my replies show as coming from me, and you can switch my access off instantly if you ever need to — versus a shared password, which is harder to fully undo. Should take ten minutes on your end."
Notice what this does: it frames the request as protecting Ada, not as Tolu being difficult. Most founders have never been asked this and will be quietly relieved someone thought of it.
A founder who trusts you may offer more than you need — the password to the business bank account "just in case," or access to a personal email that happens to also receive support mail. Take the access the job needs and nothing more. If something wider is genuinely required later, that is a conversation to have deliberately, not a default to accept because declining feels awkward.
Write the access request you'd send in your own first week on a new support role. Keep it to three sentences, following the worked example's shape.
A stacked queue — say ninety unread messages after a weekend — feels like it should be worked top to bottom, oldest first. That instinct is wrong, and following it is how a legitimately urgent message sits for two hours behind eleven "thank you!" replies that need nothing at all.
Before writing a single reply, scroll the whole queue once and sort every message into one of four piles, by severity rather than arrival time:
A calmly worded message can be a "now": "Hi, just checking — my card was charged twice for order #2291, could someone look?" No exclamation marks, no anger, and it is genuinely urgent — a double charge left alone becomes a chargeback, which costs the brand more than the refund would have. Meanwhile an angry-sounding "THIS IS RIDICULOUS, WHERE IS MY ORDER" about a package that is, per tracking, arriving tomorrow, is a "today" — real, but not actively getting worse.
A live-chat message sitting unread for twenty minutes reads as ignored, because chat sets an expectation of near-instant reply. The same twenty-minute gap on email is unremarkable. Sort by channel-adjusted urgency, not by a single universal timer.
Take any twenty messages from a real inbox, forum, or the drill file for this module, and sort them into the four piles in under ten minutes, timing yourself.
Customers do not experience "Tolu, filling in for support today." They experience "Bumi Naturals." A reply that sounds like a different person every shift, or like a generic call-centre script, quietly tells the customer the brand is bigger and more impersonal than it actually is — which is the opposite of what a small, founder-led brand is usually selling.
Before your first reply, read ten pieces of the brand's own writing — its Instagram captions, its product descriptions, a founder's past replies if any exist. Note three things: how formal or casual it is, whether it uses emoji and how many, and what it never does — Bumi Naturals, for instance, never says "no worries" (Ada finds it dismissive) and always signs off with the customer's first name.
"Dear Customer, Thank you for contacting us. Your order will be delivered within 5-7 business days. Please let us know if you have any further questions. Best regards, Support Team."
"Hi Funmi! Your order's on its way — usually 5–7 working days to Lagos, and I'll flag you the tracking link the moment it ships. Let me know if it's not with you by next Wednesday and I'll chase it personally. — Tolu, Bumi Naturals"
The second version costs almost nothing extra to write once you know the voice, and it reads like a specific person at a specific small brand — which is the entire reason a customer chose Bumi Naturals over a bigger competitor in the first place.
A voice held only in your head disappears the day you're on leave, or the day a second rep joins. Write it down once — tone, emoji rules, sign-off, three phrases the brand uses and three it never does — and it becomes an asset the brand owns, not something that lives only in your memory. This is worksheet page 3.
Pick a real small brand's Instagram or website. Write three sentences describing its voice, then draft one reply to a made-up customer question in that voice.
A saved reply — a macro — is not a shortcut you should feel guilty about. Used well, it is how a brand answers its fortieth "does this ship to Kenya" question exactly as carefully as its first. Used badly, it is how a customer with a genuine, specific problem gets an answer that clearly wasn't written for them, and notices.
A macro answering a factual, repeated question — shipping zones, ingredients, return windows — passes this test easily, because the facts don't change person to person. A macro attempting to handle a complaint, an apology, or anything with emotion in it almost always fails it, because it reads as exactly what it is: the same paragraph sent to everyone, including the one person for whom it's obviously wrong.
"Yes! We ship to Kenya via [carrier] — typically 8–12 working days, and duties are included in the checkout price so there's nothing extra on delivery."
Customer: "This is the second order that's arrived with a broken pump, I'm honestly considering not ordering again." Reply sent: "We're sorry for any inconvenience. Please contact us with your order number and we'll be happy to assist!" — generic, doesn't acknowledge it's the second time, doesn't actually resolve anything.
The fix in the second case isn't to ban macros — it's a hybrid: open with a macro-free, specific acknowledgement ("I can see this is the second broken pump you've had, that's genuinely not good enough"), then use a macro for the mechanical part (the replacement process), if one exists for that.
Editing one word of a macro — the customer's name, a specific detail — takes five seconds and is often the difference between a reply that reads as considered and one that reads as automated, even when 95% of the wording is identical.
Take three real or made-up support messages. Decide macro, human, or hybrid for each, and say why in one line.
The queue doesn't stop existing when your shift does. A handover is the difference between the next person — or the founder, checking in before bed — knowing exactly what's in flight, and them having to reopen every ticket to find out.
Not a transcript of everything you did. Three things: anything still open that needs a follow-up ("waiting on warehouse re: order #2291's broken pump — told customer 24–48 hours"), anything you were unsure about and made a judgement call on, and anything worth flagging as a pattern rather than a one-off ("three broken-pump complaints this week, might be worth a word with the supplier").
"Queue clear except: #2291 waiting on warehouse for a replacement pump, told customer 24–48h. #2305 asked to cancel before shipping — held the order, needs Ada's confirmation it hasn't shipped yet. Also: three broken-pump complaints this week (2288, 2291, 2299), all same product line — might be worth flagging to the supplier."
This takes under two minutes to write and saves the next person from reconstructing the queue's state from scratch — or worse, from missing #2305's hold and shipping an order the customer explicitly asked to cancel.
Write a handover note for a made-up shift with two open items and one pattern, in the shape of the worked example.
A weekend's worth of Bumi Naturals' queue is supplied: sixty messages across email, live chat and WhatsApp, unsorted. Clear it in one hour.
Whether the judgement from all five lessons holds up under volume and a clock, not just one message at a time in a calm read-through. Every module's drill tests judgement under a different kind of mess — this one tests it under sheer stacked volume.
You should now have: an access request template, a four-pile triage method, a written voice file, a macro-or-human test, and a handover template. Module 3 turns to the moment triage isn't enough on its own — the customer who is genuinely, sometimes justifiably, angry.