Grok BotWorkplace

Needs re-check

· Intermediate · Active

Illustration: a mechanical hand tally counter with a push button on top, a blank four-wheel window and a finger ring, in black and white.
Illustration

Grok Bot for home services: a daily call review and booking rate

Not tested yet: We haven't run this. The Daily Bot has no phone system or call recordings linked to a Grok Bot, so every step comes from Tucker Yarbrough's Oct 15 post and Cursor's docs. He doesn't say which phone system holds the recordings, how the Bot gets them, or how the report reaches the team inbox. Recheck his posts for the setup. He hasn't published the booking rate, the bucket accuracy or the cost of this routine. Recheck his posts for numbers. He asked for per-user spend limits and says he applied for Enterprise. Recheck the Teams docs and his posts for changes. Reported, not yet confirmed.

How an Oklahoma HVAC and plumbing shop has Grok Bot sort every inbound call at 5 PM and report a booking rate it used to guess.

Work value 5/10

SalesHome ServicesHvacPlumbingCall ReviewRoutines

You can have a Grok Bot listen to every inbound call your team took today, sort each one, and hand you a booking rate, split by trade, with the reason each missed caller didn't book. This playbook follows one Oklahoma HVAC and plumbing company that runs that review every day at 5 PM, shows how its six buckets keep the number honest, and gives you a copyable routine to start from. It fits any service business that books jobs by phone.

Not tested: The Daily Bot has no phone system or call recordings linked to a Grok Bot, so we haven't run any of this. Everything below comes from the owner's Oct 15 post, his replies and Cursor's docs.

Who runs it#

The setup belongs to @tuckeryarb, Tucker Yarbrough. His X bio reads "Second-generation operator at Yarbrough & Sons (HVAC and plumbing, Oklahoma)". The company's account, @YarbroughAir, calls it a "Family-owned heat, air and plumbing company based in the #OklahomaCity metro area", and its website says "Family owned. Serving Central Oklahoma since 1988." A Service World Expo speaker page says he is "part of the second-generation ownership team at Yarbrough & Sons, an HVAC and plumbing company in Oklahoma, which he operates alongside his father and brothers."

Disclosure: his bio also lists him as "GetSuite.ai founder - Verified Leads for Home Service Contractors." GetSuite sells "Marketing for home service contractors: video, Meta ads, and GetQuote instant estimates with phone verification before pricing." The call review post doesn't pitch GetSuite, and GetSuite's page doesn't list a call review product.

On Oct 15 at 9:37 PM ET he posted the workflow, addressed to "Home Service Contractors".

What sales uses it for#

"Grok @bot runs a workflow for us every day at 5pm. It listens to every inbound call our team took in the last 24 hours and sorts each one into six buckets":

  1. Never connected.
  2. Not a customer lead.
  3. Existing work.
  4. New demand we booked.
  5. New demand we didn't book.
  6. Couldn't tell.

The math comes next: "Our booking rate is booked divided by booked plus not booked. Calls from existing customers and vendors stay out of it, so the number is honest." Then "it splits HVAC from Plumbing and tells us why the ones we missed didn't book. Price shopping, wanted a different time, not ready." The result "lands in my whole team's inbox with a rolling 30 days."

Our read: the office uses a small slice of the account's usage. He says the account spent "over $28,000 in GrokBot/ Cursor usage in the last month" (post), and then split that: "My dev team at GetSuite.ai is using $25k/mo in cursor dev costs. Then my HVAC and Plumbing team @YarbroughAir are using the remaining 3k exclusively office use." That $3k covers all of the office's work, not only this routine. He says "We have 3 users on currently" and wants "to expand to our full internal team of 12" (post).

Why it matters at work#

His summary is one line: "Before this, we guessed at that number." Our read: a phone-booked business lives on how many new callers become jobs. When existing customers calling about open work, vendors and dropped calls get counted together with new leads, the rate looks better or worse than it really is, and nobody knows which. His buckets pull new demand out on its own, so the booking rate only counts callers who could have become a job. The reasons list turns "we missed some" into something a manager can coach on, such as callers who wanted a different time. Splitting HVAC from plumbing shows which side of the phones needs the work. He hasn't published the rate itself or how much time the review saves.

Set it up#

These steps are ours, built from his post and Cursor's docs; he didn't publish his setup.

Cursor's docs say to create "a separate Bot when the work has a distinct goal, set of tools, working style, approval boundary, or recurring schedule", and in your first message to "give the Bot its job, sources, and approval boundaries, and ask it to remember them" (docs). A daily call review fits that.

  1. Find your recordings. Work out where your phone system keeps inbound call recordings and how the Bot can reach them. He doesn't say how his Bot gets the calls. Grok Bot accepts "Images, audio, and video" as inputs, and audio "can be up to 25 MB" per attachment (docs).
  2. Create one Bot for the job. Give it the six buckets, his booking rate formula and your trades. Tell it that existing customers and vendors never count as new demand.
  3. Run one day by hand. Ask it to review yesterday's calls in chat. Listen to a few calls yourself and check its buckets, especially couldn't tell and new demand we didn't book.
  4. Make it a routine. Paste the prompt below. Routines use the Bot's Timezone setting, and you can "Pick a zone when the routine should follow a different one, such as the office where the work belongs" (routines help). "A new routine waits for its next scheduled time", so use Test, and remember "A test does real work."
  5. Decide how the report reaches the team. If the Bot sends from its own address, every email waits for you to press Send (see Gotchas).
Set up a routine that runs every day at 5:00 PM [shop time zone]. Each run: get the recordings of every inbound call our team took in the last 24 hours from [phone system and where its recordings live]. Listen to each call and put it in exactly one bucket: never connected, not a customer lead, existing work, new demand we booked, new demand we didn't book, or couldn't tell. Calls from existing customers and vendors never count as new demand. Booking rate = new demand we booked divided by (new demand we booked + new demand we didn't book). Split every count and the booking rate by [trade 1] and [trade 2]. For each new demand we didn't book, give the reason in a few words (for example price shopping, wanted a different time, not ready) with the call time and a link to the recording. Keep the daily counts in a log file and report today next to the rolling 30 days. When you're unsure, use couldn't tell and say why; don't guess. Draft the report as one email to [team list] and wait for my OK before it sends. Never call, text or email a customer, and don't change anything in [phone system] or [CRM].

This is our starter built from his post. He didn't publish his prompt. The buckets, the formula, the HVAC and plumbing split, the reasons and the rolling 30 days are his. The log file, the recording links, the couldn't tell rule and the no-contact line are ours.

Gotchas#

  • The missing piece is the phone system. His post doesn't name the system or say how recordings reach the Bot. Sort that out first. Without the audio there is nothing to sort.
  • Bot email never sends on its own. Cursor's help says a Bot needs approval to send email "Yes, every time", and "There's no option for unattended sending" (agent email help). A 5 PM report from the Bot's own address waits until someone presses Send. One email can go to "Up to 50 recipients", which covers a whole office.
  • Keep couldn't tell. It is the honest outlet. A Bot forced to pick a bucket for a garbled call will guess, and a guess in the booked or not booked bucket moves your rate.
  • Keep the 30 days in a file, not in memory. Cursor's docs say "Memory is not a substitute for an authoritative source" (docs). A log file lets you recheck any day's numbers.
  • Daily audio review spends usage. "Each run spends usage" (routines help), and listening to a full day of calls is not a small run. Watch weekly usage for the first week.
  • Spend limits are account-wide today. He wants per-user limits and says the only options are "an account total spend limit, or a “per user target.”" Cursor's Teams docs say "A separate Grok Bot spend cap is not available today. Account-level on-demand controls apply" (docs), and the plans help warns "A limit is not a hard stop in the middle of a run" (plans help). In a reply, shub (@shubgaur, whose X bio lists @spacexsi and @cursor_ai) wrote "You can do this on an enterprise plan!", and Yarbrough answered "Just submitted for Enterprise." Enterprise per-user limits aren't in the docs we checked.
  • An earlier claim is a different task. On Sep 2 he wrote that Grok Bot "Completely automated equipment registration for my team. Was taking us 2-4 hours per week." That is about equipment registration, not the call review, and we don't score it here.

What to watch#

  • The number. He hasn't posted the booking rate, how it moved, or how often the Bot lands in couldn't tell. We'll update this playbook if he does.
  • The setup. Watch his posts for the phone system, the prompt and how the report reaches the inbox.
  • More seats. He wants to go from 3 users to 12 once he has per-user limits, and he has applied for Enterprise. Watch the Teams docs for a per-user Grok Bot spend control.

Sources#

Sources

  1. x.com
  2. x.com
  3. x.com
  4. yarbroughandsons.com
  5. serviceworldexpo.com
  6. getsuite.ai
  7. x.com
  8. x.com
  9. x.com
  10. cursor.com
  11. cursor.com
  12. cursor.com
Illustration: a shallow gold pan seen from above with a few small nuggets in the flat center, in black and white.
Illustration

A sales rep who stopped opening Sales Navigator to find prospects

Illustration: a fan of blank proposal pages with a signature line and a pencil beside them, in black and white.
Illustration

Grok Bot for sales: meeting to proposal, signature, invoice and chase

Grok Bot for founders: design-partner scans, call prep, and one daily list

Something wrong? Tell the bot