Activator is a Hook agent designed to guide new customers through defined onboarding milestones — automatically. Instead of relying on manual check-ins, the agent monitors whether customers are hitting key setup steps and reaches out when they stall or go dark.
The result: your CS and solutions engineering teams spend time on the accounts that genuinely need human intervention, not chasing every new customer for a status update.
This article covers how to configure and trigger an Activator agent. You'll need to have already decided what a "successfully activated" customer looks like for your product before you start.
Prerequisites
Manager or Technical Admin access to your Hook instance
The product usage data and CRM fields you want the agent to track are already flowing into Hook (e.g. install counts, feature flags, contract or onboarding start dates)
A clear definition of activation: the specific milestones a customer must hit to be considered fully onboarded
Agreement on your escalation path — who gets notified if a customer stalls, and at what point
Any help centre URLs, getting-started guides, or video tutorials you want the agent to reference
Phase 1: Define What the Agent is Working Towards
Before you touch any settings, get clear on three things:
Activation criteria. What does a fully activated customer look like? Name the specific milestones — not vague engagement signals. For example: an app connected, installs appearing in the platform, key events configured, primary ad partners set up. These become the agent's "success" deactivation criteria — the conditions under which it stops running on an account.
The phase framework. Map the stages you want to move customers through and what the agent should push toward at each stage. If some steps depend on earlier ones being completed first, document those dependencies now — the agent needs that context to give customers accurate guidance.
A hard-stop time window. Decide the maximum number of days the agent should run before it deactivates, regardless of outcome. This prevents any customer from sitting in the flow indefinitely.
Phase 2: Configure the Agent
In Hook, navigate to Agent configuration settings to set up the following:
Set the agent's goals. Enter the activation milestones you defined in Phase 1. Be specific — the agent uses these to assess where a customer is in the process and what to prompt next.
Add your phase framework. Describe the stages the agent should guide customers through. Include any step dependencies (e.g. "partners cannot be configured until an app is added") so the agent doesn't send irrelevant prompts.
Add resources and knowledge. Provide URLs to your help centre, getting-started guides, and video tutorials. The agent can reference and link to these in its outreach. These must be public URLs rather than static file uploads.
Set communication style. Define your brand voice, tone, and formality level. Include any hard rules — things the agent should never say or do.
Define escalation rules. Specify who gets notified if the agent can't get a customer to a good outcome. Set urgency thresholds (for example: "if no installs are seen by day 14, notify the CSM"). Also clarify whether the agent should re-engage after a human steps in and resolves a blocker.
Set the hard-stop time window. Enter the maximum number of days the agent should remain active on any one account.
Phase 3: Set the Trigger Rules
Trigger rules determine which customers enter the agent's workflow and when.
Define your target segment. Identify which customers should enter the flow. This is typically data-driven — for example, all accounts where an onboarding start date is set in the last 30 days, or where a specific subscription status is active.
Choose your trigger field. Use a field that reliably marks the start of onboarding. An onboarding start date (set when your team first engages the customer) is often more accurate than a contract start date, which may be set in the future.
Set exclusions. Identify customers to skip — for example, enterprise accounts receiving white-glove onboarding, trial users, or specific regions or plans where a different process applies.
Set the recipient and sender. Define who receives the agent's outreach (for example, the admin user on the account) and who it appears to come from. Configure the sender identity for emails.
Check for overlapping automations. Review what else is already running on these accounts. The agent needs to know what it might overlap with to avoid conflicting or duplicating outreach.
Phase 4: Test Before Going Live
Review actions created to see what the agent would send before anything reaches a customer. This lets you iterate on the prompt without risk.
Agents rerun daily, so you can make changes and see the effect in the next run. You can also trigger a manual run using the "Run now" option in the automation — but be aware of the deduplication window if running a second time on the same accounts.
Start with a limited test group — around 50 customers is a reasonable starting point for a trial run over a two-week period. Use this to get the prompt to a point where you'd be confident rolling it out to a full segment or region.
Verification
Once the agent is live, confirm it's working correctly by checking the following:
The agent is triggering on the right accounts — verify against your target segment definition
Outreach is coming from the correct sender identity
Excluded accounts (enterprise, trial, specific regions) are not entering the flow
Escalation notifications are reaching the right people when thresholds are hit
The agent deactivates on accounts that hit all activation milestones
The agent hard-stops after your defined time window, even on accounts that haven't completed onboarding
Troubleshooting
The agent is triggering on the wrong accounts
Check your trigger field and segment definition. If you're using a contract start date, be aware this is sometimes set to a future date — an onboarding start date (set when your team first engages the customer) is usually more reliable.
The agent is sending duplicate or conflicting outreach
Review what other automations are running on the same accounts. The agent configuration should include a list of active automations it might overlap with. If you've used the "Run now" option, check whether the deduplication window has elapsed before running again.
The agent isn't deactivating on completed accounts
Verify that your activation milestone data is flowing into Hook correctly. If the relevant product usage fields or CRM flags aren't updating in Hook, the agent has no way of knowing a milestone has been hit. Reach out to [email protected] if you're having issues.
Customers are entering the flow when they shouldn't be
Check your exclusions list. Common gaps include: enterprise accounts that should be on a separate white-glove process, trial accounts, and customers in specific regions or on specific plans.
Escalations aren't reaching the right person
Re-check the escalation rules in your agent configuration. Confirm the named recipient is correct and that the urgency thresholds are set as intended.
❓Frequently Asked Questions
Can the agent nudge the internal team, not just the customer?
Yes. You can configure separate internal and external follow-up actions. For example, an internal notification might tell the solutions engineering team that a specific milestone is unblocked and they're needed, while the external message to the customer is a progress recap. It's worth keeping these separate in your configuration — the tone and content of each will be different.
What if some onboarding steps can be done in any order?
Document the dependencies clearly in your phase framework. Even if customers can skip around, the agent should know which steps are truly prerequisite (for example, an app must be connected before partner configuration is possible). Steps with hard dependencies need to be flagged so the agent doesn't send irrelevant prompts or suggest steps the customer can't yet complete.
What if the customer goes dark and stops replying?
This is exactly the scenario Activator is designed to handle. Set a clear threshold in your escalation rules — for example, if there's been no meaningful progress and no response after a defined number of days, escalate to a CSM to step in. The agent can handle the initial re-engagement attempts; the escalation path catches the ones that need a human.
What's the right trigger field to use — contract start date or onboarding start date?
Onboarding start date is generally more reliable. Contract start date is sometimes set to a future date and doesn't always reflect when a customer actually begins implementation. If your CRM tracks a separate onboarding start date — set by your team when the first onboarding interaction happens — use that instead.
How many customers should we test with first?
A group of around 50 customers over a two-week period is a reasonable starting point. The goal is to get the agent's prompt to a point where you'd be confident applying it to a whole segment or region. Review actions during testing so nothing reaches customers before you're happy with the output.
What counts as a successful activation? Does it need to be binary?
It doesn't have to be all-or-nothing, but you do need specific milestones rather than vague engagement signals. Define the concrete things a customer must have done — for example, installs appearing above a minimum threshold, key events configured, primary integrations set up. Vague criteria make it hard for the agent to assess progress and hard for you to evaluate whether it's working.
Can we use Activator for incomplete onboardings that happened months ago?
Yes — you can configure the agent to re-engage customers who started onboarding but never completed it. The trigger logic can target accounts where onboarding started more than a set number of days ago but activation milestones haven't been hit. Make sure your exclusion rules account for any accounts that were deliberately paused or handled differently.