Support work is a queue that never empties. The same ten questions arrive in a hundred different moods. A refund request needs a warm, fast yes. A billing dispute needs a careful, exact answer. A confused first-time user needs patience and steps, not policy. You know how to handle all of it. What slows you down is the typing: rewording the same answer for the fortieth time, hunting for the right macro, reading a forty-message thread before you can escalate it.
One agent, running on Keimodel credits, is built for exactly this. It drafts replies in your brand voice, reads an uploaded product doc and turns it into a clean help-center answer, and summarizes a long, angry thread into the three facts an engineer actually needs. The difference from a plain chatbot is that it remembers your tone rules, your refund policy, and your product once you tell it, so you are not pasting the same context into every chat. It drafts, you send. A human still owns the customer relationship.
Set your voice and policies in Memory first, save your best replies as Skills, and put the morning triage on a schedule. The prompts below turn a backed-up queue into a few minutes of review per ticket.
Capabilities this leans on
Do this first. Every reply the agent drafts will already match your tone and your policies, so you edit less and send faster.
Remember these facts about our support voice and policies: we're a SaaS company, our tone is warm, calm, and plain, we never blame the customer and never say 'unfortunately'. We refund within 30 days, no questions asked, and offer a prorated credit after that. We always give the customer a next step and a name to follow up with. Sign off as 'the [Company] team'. Never promise a feature or a ship date we haven't confirmed.
Paste the customer's message; get a reply that already sounds like you.
A customer wrote: 'I was charged twice this month and I'm pretty annoyed. Can you fix this today?' Draft a reply in our voice: acknowledge the double charge, confirm we'll refund the duplicate today, and give them a reference to follow up. Under 90 words.
Now give me a shorter version for live chat, two sentences, same tone.
Rewrite it once more assuming the duplicate charge turned out to be two different valid invoices, so we're explaining, not refunding. Stay warm and clear.
What you get: Three send-ready drafts for one ticket, in your voice, covering the likely outcomes.
Stop rewriting the same ten replies; save them as Skills you can fire by name.
Here are the three canned replies I use most: password reset, cancel and refund, and 'where is my export'. Rewrite each in our brand voice, then save all three as a Skill called 'Support macros' so I can call them up instantly.
Add a fourth macro to that Skill: a polite reply for a feature request we can't commit to yet, that thanks them, logs the request, and sets no false expectation.
When I say 'macro: refund', return the refund-and-cancel reply with a blank for the customer's name and order number.
What you get: A named macro library you invoke in one line instead of digging through a doc.
Same facts, different mood: the agent adjusts warmth and length, not the policy.
This customer is clearly angry: 'Third time I've written about this and nobody has fixed it. Honestly considering canceling.' Draft a reply that takes ownership, does not get defensive, states exactly what we'll do next and by when, and offers to get them on a call. Under 100 words.
This customer is confused, not angry: 'I think I did something wrong, my dashboard is empty and I don't know why.' Draft a gentle, step-by-step reply, no jargon, that walks them through the likely fix.
Both replies should keep our standard tone rules. Flag any line where I'm promising something I should double-check before sending.
What you get: Replies pitched to the person in front of you, with a flag on anything you should verify first.
Hand off to engineering or a manager without making them read forty messages.
Here is a 38-message support thread (pasted below). Summarize it for an engineer: the exact problem, what the customer has already tried, what we've already told them, the account and plan, and the one open question that's blocking resolution. Keep it under 150 words.
Now write the two-line internal note I'd post when I escalate this, with a clear ask and the priority I think it deserves.
Draft the holding reply I'll send the customer so they know it's been escalated and when to expect an update.
What you get: A clean handoff packet and a customer-facing update, both pulled from the thread you already have.
Upload the internal spec or PDF; get a public-facing article in your voice.
I'm uploading our internal doc on the new export feature. Turn it into a help-center article: a one-line summary, step-by-step instructions, a short troubleshooting section, and a 'still stuck?' line that points to support. Plain language, our tone, no internal jargon.
From the same doc, write the three most likely follow-up questions a customer will ask and a short answer to each, so I can add them as an FAQ.
Give me a 280-character version of the summary I can paste into a chat reply when someone asks how exports work.
What you get: A publishable article plus an FAQ and a quick-reply snippet, all grounded in the doc you uploaded.
Scheduled tasks and a connected channel keep the queue from surprising you.
Every weekday at 8am, take the list of open tickets I paste in and group them: refunds, bugs, billing, how-to, and at-risk customers. For each group, give a count, the three oldest, and which one I should answer first. Send it as a short digest.
(From Slack) Draft a calm reply to this one: customer says the app has been down for them all morning. Acknowledge, confirm we're looking, and give them a time we'll update them by.
Once a week, scan our public status page and recent reviews and tell me the top three things customers are complaining about right now.
What you get: A morning triage digest that sorts itself, plus replies you can fire from Slack before you open the dashboard.
SaaS founders
Landing copy, changelogs, lifecycle emails, support macros, and churn analysis.
E-commerce / DTC owners
Pricing math, product copy in your voice, competitor research, and customer replies.
Content marketing
Draft long-form, atomize it everywhere, and compare models to pick the best line.
Open the Agent, paste any prompt above, and change the details to fit your business.