# Know which renewals are at risk, 90 days ahead.

Renewal risk radar is an AI churn prediction agent. It finds every customer renewing in the next 90 days, rates each one red, amber or green on four signals: usage drop, open urgent tickets, days since contact and seat use. Then it posts one Slack list with a next step naming each account's CSM.

## Why do renewals slip without warning?

Because the warning signs sit in three places nobody checks together. Usage lives in product analytics, urgent tickets in the help desk and the last call in someone's notes. Each looks fine on its own, so an account can lose most of its users and go two months unheard before anyone sees the renewal is at risk.

Without it:

- Usage drops nobody notices until the renewal call.
- An urgent ticket open for weeks on a renewing account.
- Accounts no one has spoken to since onboarding.
- A health score nobody can explain.

With Renewal risk radar:

- Every renewal in the next 90 days, rated every Monday.
- Each rating shows the signals behind it.
- One next step per account, with the CSM named.
- Data gaps listed, never filled in.

## What signals predict churn?

Four signs a customer may not renew, each checked against your own data on every run. Two or more makes an account red.

### Usage drop

Active seats in the latest week compared with the first of the last eight. More than 25% down is a signal. For example: Weekly active seats, 8-week trend.

### Open urgent tickets

Any open or pending support ticket marked high or urgent on the account. For example: Severity, Status.

### Contact gap

More than 60 days since anyone last spoke to the customer. A missing date counts, and is listed as a gap. For example: Last contact, Data gaps.

### Seats in use

Fewer than half the seats the customer pays for active in the latest week. For example: Seats bought, Seats active.

## How does AI churn prediction work?

Every run follows the same five steps, then posts one list. Each step uses one tool, and only the last one writes anything.

1. **Finds your tables** (Tables list, reads): Customers and renewals, weekly product usage and support tickets.
2. **Reads the columns** (Tables schema, reads): Every column name checked before a query, never guessed.
3. **Finds the renewals** (Tables query, reads): Every customer renewing in the next 90 days from today.
4. **Rates each account** (Tables query, reads): Two or more signals is red, one is amber, none is green.
5. **Posts the list** (Slack, posts): Red first, each account with one next step naming its CSM.

When it runs:

- Schedule: Mondays at 8:00, before the customer success stand-up
- @mention: @renewal-risk-radar which renewals are red this week?
- Webhook: Run when your billing system calls its webhook after a fresh export

## From one sentence to a renewal list.

Describe the job, let Arclyx draft the agent, test it once on your real tables, create it, then read what lands in #renewals. Every rating shows the numbers behind it.

Example output in #renewals, "Renewal risk radar · 9 renewals in the next 90 days" (placeholder companies):

- 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. Source: example-a.com/customers
- 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. Source: example-a.com/customers
- Customer C · Amber · renews 12 Dec: $6,000 MRR. 21 of 50 seats active, 1 open ticket, 15 days since contact. One signal: under half its seats in use. Next: Sofia to present a seat plan before the renewal. Source: example-a.com/product-usage
- Customer D · Green · renews 10 Jan: $4,800 MRR. 52 to 41 active seats, no open tickets, 52 days since contact. A 21% dip, under the 25% line. Next: Sofia to start renewal prep and watch the usage. Source: example-a.com/product-usage

2 red · 2 amber · 5 green · rules in one line · no data gaps

## One list, four readers.

The same Monday list answers a different question for each person who owns part of a renewal.

- Success leads: Which renewals need a plan this week, red first.
- CSMs: Their own accounts, each with the next step named.
- Revenue and finance: MRR up for renewal, and how much of it is at risk.
- Founders: The renewals that decide the quarter, without a meeting.

## How do you identify churn risk with AI?

Pick a few signals you trust, such as falling usage, unresolved urgent tickets and long silences, and give an AI agent your data and a rule for each. Every week it rates the renewals coming up, explains each rating and names who acts next. In Arclyx that's five steps, starting with one sentence.

1. **Describe the job**: One sentence: how far ahead to look, which signals count, and where the list goes.
2. **Upload your tables**: Customers and renewals, weekly usage and support tickets, as CSV. Arclyx reads each one as a table.
3. **Pick its Slack channel**: Slack is connected once for your workspace. Pick the agent's channel on its page and invite the Arclyx app there.
4. **Watch a dry run**: Optional. See the full list for your real renewals before anything is posted.
5. **Schedule it**: Mondays at 8:00, or @mention it before a renewal meeting.

The sentence that built it (Schedule, Mondays, 8:00):

> 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.

### Or start from one of these sentences

- Before a renewal meeting: "When I ask, rate every customer renewing this quarter red, amber or green on our four signals and send me the red ones first."
- One CSM's book: "Every Monday at 8, rate the renewals in the next 90 days for Priya's accounts only and post them to #cs-priya."
- Stricter rules: "Every Monday, rate renewals in the next 120 days, counting a 20% usage drop or 45 days without contact as a signal, in #renewals."
- Red only: "Every Monday at 8, post only the red renewals in the next 90 days to #cs-leads, each with the signals behind it and one next step."

## AI churn prediction vs a health score.

A spreadsheet and a CRM health score each cover part of the job. Here is what each one does, and doesn't, do.

| What you need | By hand | HubSpot health score | Renewal risk radar |
| --- | --- | --- | --- |
| What it reads | Whatever someone exports | Rules on contact and company records | Your customer, usage and ticket tables |
| How it rates | Judgement, account by account | Points added and taken away, then a status you name | Four fixed signals: red, amber or green |
| Next step | If someone writes one | No: CSMs use the status to prioritise | One per account, naming the CSM |
| What you get | A spreadsheet | A score and status on each record | One Slack list, red first |
| Setup | Hours before each review | Service Hub Professional or Enterprise | One sentence and three tables |

Source: [HubSpot knowledge base: create a health score in the customer success workspace](https://knowledge.hubspot.com/help-desk/customize-a-health-score-in-the-customer-success-workspace), checked 24 October 2026.

## Is it safe to give an AI agent your customer data?

It only reads: read-only queries on the tables you upload, and one message to Slack or your inbox. You can see every run, and runs have a spend ceiling by default.

### Every run is recorded

- Every step, in order: Every run records the tools it called, with a short preview of each result. Only the person who started a run can open its content.
- Every rating, explained: Each account shows the numbers behind its colour, and the rules are restated on the last line.
- What each run cost: Tokens and cost are recorded per run and per agent, next to your plan's limits.
- A spend ceiling, on by default: Each run has a dollar ceiling unless you remove it. A run stops before it goes over it, and tells you.

### It can

- Read the customer, usage and ticket tables you can access in Arclyx
- Run read-only queries
- Post to one Slack channel you name, message you, or email you

### It can't

- Change your tables
- Contact a customer
- Fill in missing data
- Email anyone but you
- Post anywhere else

## AI churn prediction, answered.

### What is AI churn prediction?

Using software to spot which customers are likely to leave before they do, from signals such as falling usage, unresolved tickets and silence. Renewal risk radar applies it where it counts most: the customers renewing in the next 90 days, each with a next step for the person who owns the account.

### Is it a machine learning model?

No. It rates every account on four fixed signals you can read in the message, so you always know why a customer is red. Change a threshold in the sentence, such as a 30% usage drop, and the next run uses it.

### Which signals does it use?

Active seats more than 25% below eight weeks ago; an open or pending ticket marked high or urgent; more than 60 days since the last contact; and fewer than half the seats bought in use. Two or more is red, one is amber, none is green.

### What data does it need?

Three tables, uploaded to Arclyx as CSV: customers with renewal date, MRR, seats and CSM; weekly active seats; and support tickets with severity and status. Exports from your billing tool, product analytics and help desk work.

### What if some data is missing?

It doesn't guess. A customer with no contact date counts as a contact gap and is listed as a data gap; one with no usage rows leaves the usage signals unmet and is flagged. The last line of every list names the gaps.

### Does it change our data or contact customers?

No. It runs read-only queries on your tables and posts one Slack message, to a channel you choose or to you directly. It has no tool that writes to a table or reaches a customer.

### Can it look further ahead than 90 days?

Yes. The window is part of the sentence, so ask for 120 days, or this quarter's renewals only, and it uses that. It sorts every group by renewal date, so the nearest stay on top.

### Do I need to write code?

No. Describe the job in one sentence. Arclyx drafts the agent and its instructions, you upload your tables, and you can watch a dry run before you create it.

## Agents that work well with it

- [Customer call brief](https://arclyx.ai/agents/customer-meeting-prep): A brief before every customer call
- [Support digest](https://arclyx.ai/agents/support-ticket-analysis): The week's support requests grouped by root cause
- [KPI reporter](https://arclyx.ai/agents/kpi-reporting): Monthly KPIs with the big moves flagged

[AI for customer support and success](https://arclyx.ai/solutions/customer-support): AI customer service agents that work behind the scenes: the queue sorted, replies drafted for you, renewals at risk rated and every handoff written.

## Early access

Build your renewal risk radar. One sentence in, a working agent out. Join the beta and we'll set you up. Join at https://arclyx.ai/#join or write to hello@arclyx.ai.

Last updated 24 October 2026.