AI for customer support

An AI teammate for support that clears the weekend backlog

Glimmer works in your team chat. It answers routine questions from your own policies, in the customer's language. It checks an angry complaint against the order before anyone replies. Before the Friday meeting, it counts what customers complained about. The unusual cases come to your team.

41

weekend messages sorted by Monday morning.

38 answered from your policies. 3 left for a person, with the details attached.

Three jobs it takes off your team

Where the hours go in customer support

The same questions arrive in three languages across four channels. The one message that needs a person gets missed.

Customer support

The weekend's 41 messages, cleared by Monday standup

Every Monday opens on a weekend backlog across WhatsApp, Instagram DMs and email, most of it the same delivery, return and warranty questions, in whichever language the customer prefers, with the two genuinely tricky ones buried somewhere in the middle.

The backlog answers itself. Routine questions get the company's actual policies, in the customer's language, only the unusual ones reach a human, with context attached.

LarkAsk GlimmerSupport ยท InboxIllustrative
CW
Charmaine WeeMon, 8:05 AM

@Glimmer morning, how bad is the weekend pile? Anything I need to touch before 9?

GlimmerBotMon, 8:09 AM

41 messages came in over the weekend. 38 are handled, 3 are waiting for you:

ChannelMsgsHandledFor you
WhatsApp19181
Instagram DM12111
Email1091

Of the 38: 33 were straight off the policy sheet, delivery windows, the 14-day return policy, warranty registration, and 5 needed a Shopify order-status check first. Six were in Mandarin and two in Malay; I replied in the same language, same tone. The 3 for you: a bulk order asking for corporate invoicing, a damaged-in-transit claim with photos, and someone quoting a 20% off price from an expired IG story, I couldn't find that story to verify the price, so I've held it with her screenshot attached rather than guess.

CW
Charmaine WeeMon, 8:12 AM

Three I can do before coffee finishes. Last Monday this was my whole morning ๐Ÿ˜…

Customer supportFounders and GMs

The 1-star threat, defused with receipts

An angry message threatening a public review lands at night, and whoever sees it first replies from memory, without checking the order history, so the apology is either too generous or, worse, wrong.

Checked before anyone replies. The complaint is validated against the order history and courier scans, and the make-good goes out only after the owner says so.

Microsoft TeamsAsk GlimmerSupport ยท ComplaintsIllustrative
DN
Dinesh NairThu, 9:38 PM

@Glimmer just got a really angry WhatsApp, order still not here after almost two weeks, she's saying she'll post a 1-star everywhere tomorrow. Before I reply, is she actually right?

GlimmerBotThu, 9:44 PM

She's right. Order XXXXXX placed 28 Jun, promised 4 to 6 working days, it's now day 9. Courier tracking shows the parcel scanned into the Tuas hub on 2 Jul and no scan since, so it's likely lost, not late. Worth knowing: this is her 3rd order with you (S$212 lifetime), and her first two arrived on time, she's a good customer having a genuinely bad experience.

Make-good, needs your approval
Hi XXXXXX, you're absolutely right and I'm sorry, your parcel has been stuck with the courier since 2 Jul. We're not making you wait for their investigation: a replacement ships tomorrow by a different courier, your S$4.90 delivery fee is refunded tonight via PayNow, and here's S$10 off your next order., Dinesh

I've also prefilled the lost-parcel claim with the courier, but it needs your sign-off before submission. Nothing goes out until you approve.

DN
Dinesh NairThu, 9:47 PM

Wah, she has every right to be angry then. Send it and file the claim, glad I didn't reply from memory ๐Ÿ™

Customer supportFounders and GMs

Friday's voice-of-customer digest, causes not tickets

Support closes tickets one by one, but nobody has time to step back and count, so the same root cause generates the same 20 tickets week after week, and the boss only hears about it when a review goes public.

One table, three root causes. A week of scattered complaints across every channel, grouped and counted before the Friday ops huddle.

LarkAsk GlimmerSupport ยท InsightsIllustrative
RL
Rachel LimFri, 4:45 PM

@Glimmer ops huddle in 15, what were customers actually unhappy about this week, across everything? Marcus keeps asking and I only ever have anecdotes.

GlimmerBotFri, 4:51 PM

126 conversations this week across WhatsApp, IG, email and Shopee chat. 68 of them fall into three themes:

ThemeMentionsvs last wkLikely root cause
"Where's my order?"31+2Courier's SMS tracking link broken since 3 Jul
Tops run small25+9Size chart still shows the old cut's measurements
Refund taking long12-4Shopee refunds now landing in 7 to 9 days

The one to fix first is the size chart, it's the fastest-growing theme and it drives returns, not just tickets: 9 of this week's 14 returns cite sizing. Fixing the chart page is a 10-minute edit. Want me to draft the corrected measurements from the new cut's spec sheet for review?

MP
Marcus PereiraFri, 4:58 PM

So it's the size chart, not the courier, costing us returns. Draft it, chart gets fixed today, not another week of tickets ๐ŸŽ‰

In your industry

More jobs for this team

One job, before and after

The Monday weekend backlog

An example week for an online shop that sells on WhatsApp, Instagram and email. We count your own hours on the first call.

Today

Open the weekend messages on Monday morningSupport exec
Check order status in ShopifySupport exec
Answer each one in the customer's languageSupport exec
Handle the 3 unusual casesSupport exec

With Glimmer

Open the weekend messages on Monday morningSupport exec
Glimmer answers 38 from your policies and ShopifyAI
Handle the 3 unusual casesSupport exec

Your systems

It works with the channels your customers already use

Nothing gets replaced. Common tools connect through connectors we have checked. For the rest, we build the connector with you, or Glimmer uses the website the way a person would.

LarkMicrosoft TeamsSlack
Instagram DMsEmailShopee chatCourier trackingPolicy sheet

Singapore grants

The government pays part. You pay the rest

A first AI build for one job is the kind of project Singapore's AI grants were made for. GenAI x Digital Leaders from IMDA covers a focused first build of up to S$20,000. We help you apply.

Check your grant in 60 seconds

How we start

One team, one job, about 3 months

We pick the job that costs you the most hours, build the first agent with you and hand it over. Your team runs it after that.

See how we start

Questions owners ask

Straight answers, before you book a call

Will it reply to customers on its own?

Only where you let it. You choose which replies need a person's OK. For example, it answers delivery and return questions from your policy sheet. Refunds and apologies go to a person first.

Our customers write in Mandarin and Malay. Can it reply?

Yes. Glimmer replies in the language the customer used, with the same policies. Anything it is unsure of comes to your team.

Can it tell the boss what customers complain about?

Yes. Ask it before the weekly meeting. It groups the week's messages from every channel, counts each problem and names the likely cause.

What about our data?

Your data stays yours. Glimmer runs in your own cloud account, or on a private server we set up for you. We write down exactly where your data goes before we start. What we build for you belongs to you.

What does it cost, and can a grant pay part?

We price it after a call about your work, because the cost depends on the size of the build. Most first builds qualify for a Singapore grant. The government pays part, you pay the rest, and we help you apply. See which grant fits.

Can I try it before we commit to a build?

Yes. Our leadership workshop puts an AI agent in your hands for an afternoon, with no coding. Seats start from S$779. If you then build with us, your company gets S$500 off its first build.