The Status Filter: Shopmonkey estimate follow up that only reaches people still deciding

This is the guide from the Cortex video on the Status Filter. It follows the video's three steps in the same order:

1. Estimate status only. The agent watches orders still in Estimate status and skips anything that has moved on.
2. An approved text. The agent drafts the follow-up, and your service advisor approves it before it goes.
3. The right key owner. The API key is made under an admin account the shop keeps, so it does not break when someone leaves.

Below are the exact rules, three texts your advisors can use, and the setup details that keep the filter honest.

Why one record changes everything

In a Shopmonkey API guide by Carly, a company that sells a competing product, estimates, repair orders and invoices are all the same Order object, told apart by their status. That is efficient for a shop system. It is a trap for a follow-up tool, because a tool that looks for "orders with no reply" will find cars that were approved, repaired, paid for and driven home.

Shopmonkey describes its API as a door through which information can be shared, and its developer site links the API overview and webhooks. The vendor pages do not list the plan or key details, so the specifics below come from third-party guides, and you should confirm them with Shopmonkey.

Who this is for and what you need first

This is for auto repair shops on Shopmonkey that want estimate follow-up done every day without an advisor building the list by hand.

• An admin account that belongs to the shop, not to one person. Something like a shared "shop admin" login with its own email address.
• A list of your advisors and who approves texts for which customers.
• Your shop's rules on texting: hours, opt-outs and what you never put in a text.
• A place where drafts land for approval each morning.

Step 1, estimate status only

The filter, written out so anyone can check it:

Rule	Keep the order when	Why
Status	It is Estimate	Anything else has moved on
Age	The estimate is at least 2 days old and no more than 30	Too early is pushy; too late is a different conversation
Same vehicle	No newer order exists for that vehicle	A newer order means the customer is already back
Last contact	No message to the customer in the last 2 days	One follow-up at a time
Declined items	The customer has not declined the whole estimate	Declined work is its own follow-up

The age and contact numbers are our starting suggestions, not Shopmonkey rules. Adjust them to how your customers decide. For the different case of work a customer said no to, see declined work follow-up.

One more detail from Carly's guide: webhooks fire on table-level changes, so one change in the shop can send several webhooks. It lists webhooks on 11 objects, including Order, Customer, Vehicle and Payment, with a 55 second delivery timeout. It also says a failed webhook is disabled with no automatic redelivery. So the agent should treat webhooks as a hint, re-read the order's current status before drafting anything, and run one full check each night to catch whatever a webhook missed.

Step 2, an approved text

The advisor sees each draft next to the estimate and approves, edits or skips it. Three texts to start from. Names and cars are examples.

Day 2, a question:

Hi Mr. Patel, it's Corner Garage. Any questions on the brake estimate for your Civic? Happy to walk you through it.

Day 5, the safety item first:

Hi Mr. Patel, just a quick note on your Civic. The front brakes were the item we'd fix first. Want us to hold a time this week?

Last note:

Hi Mr. Patel, we'll close out the Civic estimate at the end of the week. If you'd like it done, reply here and we'll set a time. Thanks for considering us.

After the last note, the agent stops. More messages after that tend to cost goodwill rather than win work. We explain why every text goes through a person at first in human-in-the-loop AI.

Step 3, the right key owner

According to Carly, API keys are created in Settings under Integration, API Keys, by an admin. A key carries the same permissions as the user who created it, returns a 403 error once that user is offboarded, and cannot be viewed again after it is created.

So:

• Create the key while signed in as the shop admin account, not a person's own login.
• Name it after its job, for example "estimate follow-up agent".
• Save it in the shop's password manager the moment it is shown.
• Write down who created it and when.

A Supergood report card on Shopmonkey adds that there is no sandbox or developer tier and no fine-grained key scopes. That means the key can do whatever that admin can do. Keep the agent itself to reading orders and writing drafts, and check with your builder that nothing in its code edits orders.

Mistakes to avoid

• Filtering once, sending later. Status can change between the morning list and the afternoon send. Re-check status right before a text goes out.
• Trusting a single webhook. Several webhooks can arrive for one change, and a failed one is not resent. Use them to wake the agent, not as the record.
• Using a person's key. When that person leaves, follow-up stops without anyone noticing.
• Texting after pickup. If it happens once, pause the agent and find the gap in the filter before turning it back on.

How to tell it worked

• Every draft your advisor sees is for a car whose estimate is still open.
• No customer who picked up their car gets an estimate text. Spot-check five drafts a week against the order screen.
• The follow-up keeps running the week after someone leaves the shop.
• Your advisors edit fewer drafts each week.

For the bigger picture of agents on Shopmonkey, read how an AI agent for Shopmonkey works. If you would like someone to look at your shop's follow-up with you, your free AI Operations Audit covers it.