Get early access
AI for customer support and success

AI agents for support and success, nothing sent without you.

AI for customer support and success teams works best behind the scenes. In Arclyx, five agents sort the support queue and draft routine replies, group the week's requests by root cause, rate renewals at risk, brief you before customer calls and write the handoff when you name a closed deal. None of them talks to a customer.

Get early accessMeet the agents ↓
What lands in Slack
Support triage#support · Weekdays, 9:00

Support triage · 4 urgent, 0 need a person, 6 routine

  • Customer A · Urgent · no draft

    Everything failing since last night.

    An outage. Page the on-call engineer.

  • Customer B · Urgent · no draft

    Possible personal data in an automated message.

    Possible data exposure. Needs a person today.

10 requests · 6 drafts saved in Gmail · nothing sent

The agents

How can support and success teams use AI agents?

On the reading and sorting between conversations, never the conversations themselves. Two agents work the support queue, three work the accounts. Each starts from one sentence, reads only the tools it is given and sends one short message to you or your team.

  • Weekdays, 9:00

    Support triage

    Every weekday, today's requests sorted into urgent, needs a person or routine, with a Gmail draft waiting for each routine request.

    Reads
    TablesDocument search
    Drafts in
    Gmail
    Posts to
    #support, or emails you
    See Support triage →
  • Fridays, 16:00

    Support digest

    Every Friday, the week's requests grouped by root cause, with the customers named and the open GitHub issue for each problem.

    Reads
    TablesGitHub
    Posts to
    #support-digest, or emails you
    See Support digest →
  • Mondays, 8:00

    Renewal risk radar

    Every Monday, each renewal in the next 90 days rated red, amber or green on four signals, with one next step for its CSM.

    Reads
    Tables
    Posts to
    #renewals, or emails you
    See Renewal risk radar →
  • When you ask

    Customer call brief

    Before a customer call: the account, open tickets, the usage trend and every promise from the last call, kept or not.

    Reads
    TablesDocument searchGoogle Calendar
    Posts to
    your Slack DMs, or emails you
    See Customer call brief →
  • When you ask

    Success handoff

    When you name a closed deal: what they bought and why, every promise checked against the contract, and who's who.

    Reads
    HubSpotDocument search
    Posts to
    your Slack DMs, or emails you
    See Success handoff →

All agents, by team →

One customer, five agents

Where does each agent fit in a customer's year?

Each one takes a moment that usually slips: the day the deal closes, the requests that pile up, the problems that repeat, the call you prepare for in a minute, and the renewal you notice too late. They don't hand work to each other; each reads your data at its own moment.

  1. 01

    The deal closes

    Ask Success handoff for the deal: it reads the HubSpot deal and the discovery notes, then writes what success needs on day one.

  2. 02

    Requests come in

    Support triage sorts every new request each morning and drafts the routine replies for a person to send.

  3. 03

    The same problem repeats

    Support digest groups the week's requests by root cause and checks whether engineering tracks each one.

  4. 04

    A call is booked

    Customer call brief pulls the account, tickets, usage and last call's promises into one brief.

  5. 05

    Renewal comes into view

    Renewal risk radar rates every renewal in the next 90 days and names who acts next.

Safe by design

Can an AI agent reply to customers for us?

Not these five. In Gmail their only writing tool saves drafts, so a person reviews and sends every reply, and nothing urgent or sensitive is drafted at all. Your tables, HubSpot and GitHub are only read. Each agent's one message goes to you or your team, and every run is recorded.

They can

  • Read the tables and documents you upload
  • Read HubSpot deals, GitHub issues and your calendar
  • Save Gmail drafts addressed to a request's sender
  • Post one message to Slack, or email you

They can't

  • Send an email or message a customer
  • Draft for urgent or sensitive requests
  • Change HubSpot, GitHub issues or your tables
  • Fill in missing data
  • Email anyone but you

How Arclyx keeps agents inside their access →

A week of the team

How do you use AI in customer success?

Give each job that slips its own agent, and let it run on a schedule or when you ask. Here is a week of a support and success team on Arclyx: each step is the sentence that built an agent, then what it posted in Slack. Try it runs a new agent once as a dry run first.

  1. Monday, 8:00 · Renewal risk radar
    “Every Monday at 8, find every customer renewing in the next 90 days in our customer, usage and ticket tables, rate each red, amber or green on usage drop, open urgent tickets, days since contact and seats in use, and post the list to #renewals with one next step for each account's CSM.”
    #renewals
    Renewal risk radarRenewal risk radar · 9 renewals in the next 90 days
    • Customer A · Red · renews 20 Nov

      $3,200 MRR. 37 to 14 active seats, 2 open high tickets, 73 days since contact.

      All four signals. Next: Priya to run an exec call this week on both open tickets.

    • Customer B · Red · renews 5 Dec

      $1,250 MRR. 22 to 9 active seats, an urgent ticket open, 86 days since contact.

      All four signals. Next: Marcus to close the urgent ticket this week and right-size 25 seats to about 9.

    See Renewal risk radar →
  2. Tuesday, 9:00, as every weekday · Support triage
    “Every weekday at 9, sort today's new support requests into urgent, needs a person or routine using our help notes, save a Gmail draft reply for each routine one a help note answers, never send anything, and post the list to #support.”
    #support
    Support triageSupport triage · 4 urgent, 0 need a person, 6 routine
    • Customer A · Urgent · no draft

      Everything failing since last night.

      An outage. Page the on-call engineer.

    • Customer B · Urgent · no draft

      Possible personal data in an automated message.

      Possible data exposure. Needs a person today.

    See Support triage →
  3. Wednesday, 11:30, asked before a call · Customer call brief
    “When I name a customer, brief me for the call from our customer, usage and ticket tables, the call notes and my calendar: plan, seats, MRR and renewal, open tickets, usage trend, what they asked for and what we promised, the next meeting, and two or three things to raise.”
    your Slack DMs
    Customer call briefCustomer call brief · Customer B
    • Customer B · Account

      Team plan, 25 seats, $1,250 MRR. Renews 5 December, in 42 days.

      Owned by Marcus. The renewal is close enough to come up on the call.

    • Customer B · Open tickets

      Two open: an urgent integration failure, 16 days old, and a question about paying for 25 seats.

      The urgent one blocks the reason they bought. Know its fix status first.

    See Customer call brief →
  4. Thursday, 15:40, asked as a deal closed · Success handoff
    “When I name a closed deal, read it in HubSpot and our discovery notes, and send me the handoff for customer success: what they bought, why they bought, what we promised, who's who on both sides, risks and the first onboarding steps.”
    your Slack DMs
    Success handoffHandoff · Customer N · Business plan
    • Customer N · Deal

      $33,600 a year for 35 seats, closed won 22 October, 12 months from 1 November.

      A 10% first-year discount, tied to a case study by month six.

    • Customer N · Why they bought

      Drafts-only email, so a person approves every message to their customers, and agents their ops staff can build without an engineer.

      A weekly report in Slack was the third reason.

    See Success handoff →
  5. Friday, 16:00 · Support digest
    “Every Friday at 4, read this week's support requests, group them by root cause, count each group and name the customers, find the open GitHub issue that tracks each one or say there isn't one, and post the digest to #support-digest, biggest group first.”
    #support-digest
    Support digestSupport digest · Monday to Friday · 22 requests in 9 groups
    • Customer A · Reports stop after a schedule is edited · 5

      Five requests, from Customers A, G, H, I and J.

      Tracked by #214: editing a schedule's time clears where its report is sent.

    • Customer B · Integration shows connected, but syncs fail · 4

      Four requests, from Customers B, A, D and E.

      Tracked by #187: mark an integration unhealthy when its sign-in fails.

    See Support digest →
  6. Done

    A week of support and success. Nothing sent to a customer.

    • Renewals rated, in #renewals
    • Routine answers drafted in Gmail, waiting for a person
    • A brief before Wednesday's call, in your Slack DMs
    • Thursday's deal handed to success
    • The week's recurring issues, in #support-digest

Create an agent from one sentence →How a dry run works →

Questions

AI for customer support and success, answered.

What are AI agents for customer support and success?

Software that takes a recurring support or success job, such as sorting the queue or checking which renewals are at risk, works from your own data and reports to your team. In Arclyx each one reads only the tools it's given and sends one short message.

Are these a support chatbot?

No. A chatbot answers customers. These agents work for your team behind the scenes, and none of them talks to a customer. The closest is Support triage, which drafts answers to routine requests in Gmail for a person to review and send.

Do they work with Zendesk or Intercom?

Through exports, not a connection. Export your requests, tickets and usage as CSV from your help desk, product analytics or billing tool, upload them to Arclyx, and the agents read them as tables.

Can an AI agent predict churn?

Renewal risk radar rates every renewal in the next 90 days on four signals: usage drop, open urgent tickets, days since contact and seats in use. It isn't a trained model, so every rating shows the numbers behind it.

Do they work with HubSpot and GitHub?

Yes, read-only. Success handoff searches and reads HubSpot deals, companies and contacts; Support digest lists, searches and reads open GitHub issues. Neither can change a record or an issue.

More of Arclyx

Last updated 25 October 2026

Build your first support agent.

One sentence in, a working agent out. Join the beta and we'll set you up.

or write to hello@arclyx.ai →