INSIGHTS
AI & Automation Insights for Operators
Practical guides on AI assistants, workflow automation, lead intelligence, and field operations. Written for business owners and operators who want results, not hype.
Your Endpoint Was Down. Which Webhooks Are Gone?
How much data a webhook outage costs you is set by the source with the shortest retry window, not the longest. Slack stops retrying after about five minutes, Shopify after roughly four hours, HubSpot after 24, Stripe after three days. Shopify removes the subscription and Slack disables it after sustained failure, so the pipe is gone rather than merely backed up. Recovery means a reconciliation sweep, and only events whose facts survive in the current record state, or in a vendor event log like Stripe's 30-day events list, can be rebuilt at all.
Quiet Hours: Why Your Automation Texts at 6 a.m.
Automated messages go out during legal quiet hours because the send-time guard checks the wrong clock and checks it at the wrong moment: a timezone derived from an area code that no longer matches where the person lives, evaluated when the batch is queued rather than when each message is actually delivered. Oklahoma and Florida both presume by statute that a message sent to their area code reached someone in their state, so the correct rule is to send only when the window implied by the area code and the window implied by the recipient's known location are both open.
Automation Can't Find the Record It Just Created
A search or find step that returns nothing for a record you just created is almost never a missing record. Most vendor APIs expose two read paths, and search endpoints read an asynchronous index while get-by-id, list, and unique-property endpoints read the record itself. Stripe, HubSpot, and Atlassian all document the lag. The fix is to call the direct read path or carry the ID forward from the create response, not to add a wait step.
Totals Off by a Penny? You Are Rounding Twice
A one-cent mismatch between two systems is almost never a floating-point bug. It happens because both systems rounded the same number under different rules, most often one rounding tax per line item and the other rounding tax on the subtotal, and both answers are defensible. The fix is to name exactly one system as the rounding authority for each figure, pass money between systems as integer minor units, and use largest-remainder allocation whenever a rounded total has to be split.
When an Employee Leaves, Which Automations Break?
Deactivating a departing employee does not break automations in one predictable way. Some connections die the moment IT resets the password, weeks before the departure. Some keep running indefinitely under a person who no longer works there. And the ownership transfer you perform to clean up the first problem silently changes the inbound webhook URL of every Zap that catches hooks, taking down a second set of integrations that were working fine.
Why Your CSV Import Silently Mangles IDs and Dates
CSV files carry no type information and no encoding declaration, so every program that reads one guesses independently, and the guesses are silent and irreversible. A ZIP code of 02138 becomes 2138, a 16-digit order number loses its last digit to a zero, and a part number of 3/4 becomes a date. The damage happens at five separate hops in a typical pipeline, not just when someone opens the file in Excel, and it cannot be undone by formatting the column afterward because the original characters are already gone.
Automated Emails Not Threading? Store the Thread ID
Automated emails start a new thread because the sending step never receives the identity of the message it should be replying to. Zapier, Make, and n8n each ship a reply action that threads correctly, but all three require a thread or message ID that only exists if an email started the run, so workflows triggered by a form, a CRM update, or a schedule have to store the thread handle on the record at first send.
Duplicating a workflow is not a staging environment
Duplicating an automation does not create a staging environment, because the copy inherits the same app connections and writes to the same live systems. On a polling or scheduled trigger the copy becomes a second live consumer that doubles every side effect. Real isolation is decided per connected system: a vendor test mode where one exists, a second free account where it does not, and a hard recipient allowlist in front of the send step where neither is possible.
When Google Sheets breaks as an automation backend
Google Sheets stops working as an automation backend because of how records are addressed and written, not because of how many rows it holds. In Zapier and Make a row's position is its identity, so a single delete or sort re-points the automation at a different record, and Sheets enforces no unique constraint and no row-level lock, so two runs updating the same row lose one write with no error. Both failures appear at a few hundred rows, long before the 10 million cell limit matters.
Why your automation is stuck waiting for approval
An automation stalls at an approval step because three separate clocks run on every approval: the workflow's wait timeout, the lifetime of the callback behind the button you sent, and the reviewer's own response time. Setting only the wait timeout leaves you with an approve button that still looks live after the run behind it is gone, so the durable fix is to store the approval as a decision record the workflow reads, not as a pause it holds open.
Why automation projects fail: three wrong assumptions
Automation projects usually fail because the model of human behavior encoded in the workflow is wrong, not because an integration broke. The three most expensive wrong assumptions are that a field edit is a decision, that a customer reply is a command, and that your inventory of what already exists is accurate. All three pass testing and only surface under real use.
Why your CRM shows three of the same customer
Repeated customer records usually come from a schema that stores one person per customer, not from weak matching. When a household contains several people with different phone numbers and emails, every channel creates its own record, and tightening the matcher makes it worse because aggressive matching is what merges the wrong people permanently. The fix is a contacts layer beneath the customer record, plus an ordered resolution rule that says which record wins.
Want a system like the ones we write about?
We build practical AI and automation systems for businesses that need better workflows, connected tools, and cleaner operations.
Request a Workflow Review