How We Automated Trial Follow-Ups with AI Agents without touching CRM

When a trial comes in, the right move is a fast, personal email that references what they actually did in the product. But that takes time: look up their activity, figure out where they are, write something specific, schedule a follow-up, log it. With 5 to 10 trials a week, that process quietly breaks down.

We connected our live trial data to Claude using Knowi MCP and the agent now handles every step automatically. When a new trial comes in, the agent detects it, checks what they built, filters out noise, writes a personalized email timed to their local business hours, schedules the follow-up sequence, and logs everything. No one has to open a spreadsheet, CRM, or product dashboard to make it happen.

Here is exactly how it works.

Table of Contents

Quick Summary (TL;DR)

• Knowi MCP gives Claude direct access to live trial data. No exports, no manual lookups, no integrations to build.
• When a new trial comes in, the agent checks what they connected, what they built, and where they are located before doing anything else.
• Current clients and spam signups are filtered out automatically before any email is drafted.
• Emails are timed to arrive at local business hours based on signup timestamp and geo location.
• A follow-up sequence runs automatically on day 3 and day 7, with different angles depending on whether the trial engaged or went quiet.
• Every action is logged to a live tracker dataset. No CRM entry needed.

Step 1: Detect New Trials

The agent runs on a schedule, typically at noon, 2 PM, 4 PM, and 6 PM local time. Each run opens with a single query to the live Trials dataset in Knowi:

“Show me all trial signups from the last 24 hours that are not already in the Lead Tracker.”

Knowi MCP returns the results in real time: company name, email, employee count, location, signup timestamp, datasources connected, widgets created, and login count. No export, no dashboard, no manual check. The agent has everything it needs to start the workflow.

If there are no new trials, the run takes two seconds and outputs a single line: “No new trials since last check.” If there are new trials, the workflow continues.

Step 2: Filter Out Noise Before Writing Anything

Before drafting a single email, the agent removes the leads that should never receive outreach. This happens automatically on every run.

Current clients: The agent cross-references the signup email domain against the existing customer list in Knowi. If the domain matches a paying account, the lead is logged as “existing client” and skipped. No awkward email asking a $40,000 annual customer if they want to try the product.

Spam and test signups: Addresses with generic domains (gmail, yahoo, hotmail), obvious test patterns (test@, demo@, admin@), or no company name attached are flagged and removed. The agent does not draft emails for these. They go into a “filtered” log for review if needed.

Already contacted: If a lead is in the tracker with an outreach date in the last 7 days, the agent skips them and checks whether a follow-up is due instead.

What is left after the filter is a clean list of net-new trial signups worth a personal touch.

Step 3: Read What They Actually Did in the Product

For each trial that passes the filter, the agent reads three signals from the live dataset before deciding what to write:

• Datasources connected: Did they connect anything? If yes, which ones? MongoDB, Snowflake, a REST API, and Elasticsearch each point to a different use case and a different email angle.
• Widgets created: Did they build anything? Someone who connected a datasource and built five widgets is deep in evaluation mode. Someone who signed up and never connected anything needs a different message entirely.
• Login frequency: One login versus returning three times in a week signals very different levels of intent.

The agent uses these signals to determine the email angle. Not a generic template, but a specific message based on what that person did. If they connected MongoDB Atlas and built four widgets, the email references MongoDB specifically. If they signed up but built nothing, the email leads with a use case relevant to their industry or role.

Step 4: Time the Email to Their Local Business Hours

The agent checks the signup timestamp and infers the trial’s time zone before drafting the send time. A signup at 5:45 AM UTC points to India or Southeast Asia. A signup at 3 PM UTC points to the US West Coast.

Emails are scheduled to arrive at 9 to 10 AM local time on a business day. This is not a cosmetic detail. Emails that arrive at the start of the local workday get meaningfully higher open rates than those sent at the sender’s convenience.

The routing also determines who sends the email:

• Large enterprise logos (Fortune 500 or recognizable brands): GTM lead handles personally, regardless of location
• US company, 50 or more employees: flagged to sales to call within 24 hours, GTM lead sends email
• US company, under 50 employees: sales email follow-up
• Outside US: GTM lead sends personal email, timed to local business hours

Step 5: Draft the Personalized Email

Here is what two emails from the same Monday morning run looked like, for two trials with completely different product activity.

Active Trial: Connected MongoDB, Built Widgets (India)

Arjun at Databridge Labs connected MongoDB Atlas and Cloud9Charts, then built five widgets. The agent drafts:

“Saw you connected MongoDB Atlas and have been building. We work with a lot of MongoDB teams and have deep native support: real queries, no flattening, no ETL. Happy to show you what is possible beyond what you have set up so far.”

Send time: 10 AM IST. The MongoDB specificity makes it clear someone looked at the account, not just the signup form.

Inactive Trial: No Datasource, No Activity (London)

James at a mid-size branding agency signed up but never connected anything. The agent reads the company type and role, then leads with a relevant use case:

“A lot of teams like yours use Knowi to pull together client data from multiple sources into one live dashboard. No manual exports, no waiting on engineering. Happy to show you a quick example with data that looks like yours.”

Send time: 9 AM GMT. The angle speaks to the workflow problem, not the product features. No SQL reference, no data engineering pitch.

Step 6: Schedule the Follow-Up Sequence Automatically

The agent does not send one email and wait. After the first touch, it writes two follow-up dates into the Lead Tracker and handles them on each scheduled run.

Day 3 follow-up: If no reply, the agent checks whether the trial’s product activity changed since the first email. If they connected a datasource after the first touch, the follow-up references it directly. If nothing changed, the follow-up shifts angle. For example, from a product-activity hook to a “what were you hoping to see” open question.

Day 7 follow-up: If still no reply, one more message, shorter. Something like: “Happy to answer questions over email if a call does not work. What were you hoping to do with the data?” After this, the lead moves to “nurture” stage and exits active follow-up.

Every follow-up is drafted fresh at send time, not pre-written. The agent re-reads the current trial data before drafting, so if the person became more active in the product between day 1 and day 3, the follow-up reflects that.

Step 7: Log Everything Without Touching a CRM

After each action, the agent pushes a structured row to the Lead Tracker dataset in Knowi using knowi_push_data. The row captures the full state of the lead: stage, owner, email angle used, send time, next follow-up date, and whether the trial’s product activity changed since last touch.

The tracker becomes a live dashboard queryable at any time. Want to see all trials in active follow-up? One query. Want to see which trials from the past 30 days connected a datasource but never replied? One query. Because it lives in Knowi, it can be filtered, visualized, or used to trigger an alert, the same as any other dataset.

What This Replaces

Frequently Asked Questions

Want to connect your trial data to an AI agent and automate your follow-up workflow? Start a free trial or talk to our team.