What the Jarvis AI Brain is, and why a rhythm matters
The Jarvis AI Brain is the natural-language layer of Run with Jarvis: you ask a plain question about your own calls, calendar, payments or ad attribution, and it answers from the platform's data. This guide gives you 30 questions worth asking and arranges them into a rhythm, a short daily close-out and four weekly checks, so that asking becomes a habit tied to decisions.
As of October 2026, the Brain is part of the Elite plan, where it appears on the plan list as the Jarvis AI Assistant. The capability itself is described in detail in the Jarvis AI Brain deep-dive. What that article does not give you is a routine, and a routine is where the value is. An assistant you can ask anything tends to get asked nothing unless you decide in advance which questions matter and when.
So this is a practical list. It covers what the Brain can see, how to phrase a question so the answer is unambiguous, the 30 questions themselves, the honest limits, and how owners on Core or Pro get to the same answers by a longer route. The platform is Run with Jarvis, and everything below stays inside what its published plan list and existing documentation say the Brain does.
What the Brain can see
An assistant can only answer from data it can reach. The Brain reads across four layers of the same platform.
The calls. KeyBot answers your phone around the clock in English and Spanish. Each call leaves a record of who called, what they wanted, and whether it ended in a booking.
The calendar. GetTimePad holds the appointments, whoever booked them. That covers what is scheduled, what is open, and how a lead moved to a booked job.
The customers and the money. IntelliDrive holds the CRM, the POS, invoicing and payments, along with dispatch. This is what makes a question about collections or unpaid invoices answerable at all.
The attribution. CallFlux, which is included from the Pro plan up, tracks which number, campaign and ad source produced each call, with recording, transcription and lead scoring.
The reason this matters is that the best questions cross those layers. "Which ad source booked jobs this week?" needs the attribution, the booking and, if you want revenue, the payment. No single layer holds all three. If you want the wider picture of how the pieces fit, the overview of what Jarvis software is covers the architecture.
It also sets the limit. If something never went through the platform, the Brain does not know about it. More on that below.
How to phrase a good question
The single biggest improvement to the answers you get is in how you ask. A vague question gets a plausible answer to something slightly different from what you meant. Three ingredients fix most of that.
A time range. "Last week" can mean the last seven days or the calendar week that ended Sunday. Say which. "Between Monday and Sunday last week." "Since the first of the month." "Yesterday."
A specific measure. "How did we do?" is not a question. "How much did we collect?" is. So is "How many jobs did we book?" Notice that booked, completed, invoiced and collected are four different numbers. Name the one you want.
A comparison, when you want a judgment. A number on its own rarely tells you anything. "How much did we collect this week compared with last week?" tells you direction. "How many calls did we book this week against the week before?" tells you whether a change is real.
Put together, a strong question reads like this: "How many jobs did we book between Monday and Sunday last week, and how does that compare with the week before?" It can only mean one thing.
Two more habits help.
Ask the follow-up. The first answer is usually the start. Bookings are down, so is that fewer calls or fewer calls converted? Fewer calls, so from which source? Each follow-up is one more sentence, with the context carried forward.
Say it the way you would say it to an office manager. You do not need field names or report titles. If a person would have understood the sentence, it is a reasonable sentence to ask.
The weekly rhythm at a glance
| When | Question group | Decision it drives |
|---|---|---|
| End of every working day | Daily close-out (questions 1 to 6) | Who needs a callback tonight or first thing tomorrow, and whether tomorrow's schedule needs filling |
| Monday morning | Weekly review (questions 7 to 12) | Where to put attention this week: calls, conversion, quotes or capacity |
| Wednesday | Marketing check (questions 13 to 18) | Which ad sources to keep, fix or stop spending on |
| Thursday | Cash check (questions 19 to 24) | Which invoices to chase before the weekend and which customers to call personally |
| Friday | Team and schedule check (questions 25 to 30) | How next week is staffed and where the workload is uneven |
The days are a suggestion. What matters is that each group is tied to a decision and asked on a regular beat, so that this Thursday's answer can be held against last Thursday's.
Daily close-out: questions 1 to 6
Ask these at the end of the working day. They take a few minutes and they protect the revenue that leaks overnight.
1. How much did we collect today? This is collected, not invoiced. It is the number that tells you what actually arrived.
2. Which calls today did not end in a booking, and why? Some will be wrong numbers and price shoppers. Some will be real jobs that slipped. You are looking for the second kind.
3. Are there any missed calls or callback requests from today that have not been returned? This is the question with the most money attached. A caller who was promised a callback and did not get one is a customer who is about to become someone else's.
4. What does tomorrow's schedule look like, and where are the gaps? A gap you see tonight can be filled. A gap you discover tomorrow morning is a technician waiting.
5. Did any jobs scheduled for today not get completed? Anything left open needs rebooking or a call to the customer before they call you.
6. Were there any after-hours calls last night, and did they book? This one looks back a day on purpose. It tells you whether the overnight front door is working.
The decision. By the end of these six you should know who gets a call tonight or first thing tomorrow, and whether tomorrow needs another job on it.
Monday review: questions 7 to 12
Monday is for the shape of the week that just ended. Ask each with a comparison to the week before.
7. How many calls did we receive last week, Monday to Sunday, compared with the week before? This is demand. It moves with season, weather and marketing.
8. How many of those calls turned into booked jobs? Demand and conversion are different problems with different fixes. A drop in calls is a marketing question. A drop in calls booked is a front-desk or pricing question.
9. How much revenue did we collect last week compared with the week before? Keep it to collected revenue so the comparison is like for like.
10. Which leads from last week are still not booked? These are people who contacted you and have not said no. They are the cheapest jobs you will ever win back.
11. Which quotes older than seven days are still open? A stalled quote is rarely a refusal. It is usually a customer who got busy.
12. How booked is this week, day by day? This tells you where there is room to push work and where there is none.
The decision. One sentence: where does attention go this week? If calls were down, look at marketing on Wednesday with that in mind. If calls were steady and bookings were down, listen to some calls. If quotes are piling up, set aside time to follow up.
Marketing check: questions 13 to 18
These questions depend on call attribution, which is part of Pro and Elite. Without tracking in place, there is nothing recording which ad produced which call, and no assistant can reconstruct it afterward. The background is in the call tracking and attribution guide.
13. Which ad source produced the most calls in the last seven days? Volume first.
14. Which ad source produced the most booked jobs in the last seven days? This is frequently a different answer from question 13, and the difference is the whole point of asking.
15. Which sources produced calls that did not book? A source that generates calls and no jobs is either attracting the wrong people or sending them at the wrong time.
16. How much revenue has been collected so far from jobs that came from each source this month? Booked is a promise. Collected is the result.
17. How many of last week's calls were from existing customers and how many were new? Marketing gets credit for new customers. Repeat customers were going to call anyway.
18. How did last week's calls from our main ad source compare with the week before? One source, two periods. This is how you notice a campaign fading before the month-end report.
The decision. For each source: keep, fix or stop. A source with calls and bookings stays. A source with calls and no bookings gets investigated. A source with neither gets questioned.
One caution. The Brain reports what the platform recorded: the call, its source, the booking and the payment. Your judgment about why a source is underperforming, whether the ad, the landing page or the hour of day, is still yours to form.
Cash check: questions 19 to 24
Thursday works well because there is still a working day left to act before the weekend.
19. Which invoices are unpaid, and how old is each one? Sorted oldest first, this is a to-do list.
20. Which unpaid invoices are more than 30 days old? These need a person, not another automatic reminder.
21. How much did we collect this week so far compared with the same point last week? A mid-week reading gives you time to respond.
22. Which customers have more than one unpaid invoice? One late invoice is an oversight. Two is a pattern, and it should change how you take the next booking from that customer.
23. How much have we invoiced this month and how much of that has been collected? The gap between those two numbers is your receivables in one line.
24. Which jobs completed this week have not been invoiced yet? Work done and not billed is the most avoidable cash problem there is.
The decision. Pick the invoices to chase on Friday and the two or three customers who should get a personal call from you.
Team and schedule check: questions 25 to 30
These questions rely on jobs being assigned to technicians in the platform. If assignments are entered consistently, the answers are useful. If they are not, fix that first.
25. Which technicians have open capacity next week? This is where next week's bookings should be steered.
26. How many jobs did each technician complete this week? Look for large differences and ask why before drawing conclusions. Job types differ.
27. How much revenue was collected on each technician's jobs this month? Useful alongside question 26, since a technician with fewer, larger jobs may be carrying more than the count suggests.
28. Which appointments next week look at risk of a no-show? Unconfirmed appointments are the usual candidates.
29. Are any days next week overbooked or underbooked? An uneven week can often be evened out with a few calls.
30. Which customers have not booked with us in twelve months? This is a reactivation list for the slow days you found in question 29.
The decision. How next week is staffed, which days need more work, and which days need relief.
A word on using these fairly. Numbers about individual technicians are a starting point for a conversation, not a verdict. A technician who handles the difficult jobs will look slower on a count.
Follow-up questions are where the value is
The 30 above are openers. The real use comes from the second and third question.
Take question 8. Bookings from calls dropped. The natural follow-ups are: Was the drop on particular days? Was it after hours or during the day? Were the unbooked callers asking for a specific service? Each one narrows the cause, and each is a sentence.
This is the practical difference from a report. A report gives you the first answer in a fixed shape. A conversation lets you keep going until you reach something you can act on. The comparison between the two approaches, including the places where a fixed screen is still better, is set out in asking your business a question vs reading a dashboard.
It also explains why the rhythm matters. A fixed set of opening questions, asked on a schedule, gives you the consistency of a report. The follow-ups give you the flexibility a report lacks.
What it cannot answer
Being clear about limits is what lets you trust the rest.
Anything that never entered the platform. A cash payment that was not recorded. A job agreed in a parking lot and never booked. A quote given verbally. If there is no record, the honest answer is that there is no record.
Spending and activity in tools it does not hold. The Brain works over the platform's own data. Costs that live elsewhere, such as fuel, payroll or a supplier account, are not something it can report on.
Why something happened. It can tell you that calls from one source stopped booking. The reason is a judgment you form from listening to calls and looking at the offer.
What to do about it. Whether to discount, to hire, to cut a campaign or to call an unhappy customer is a decision. The Brain shortens the time it takes to see the situation. It does not make the call.
Attribution without tracking. On a setup without call tracking, the questions in the marketing group have no data behind them.
When the Brain tells you it has no record of something you know happened, treat that as useful. It has just shown you a gap in what your business is capturing.
Getting the same answers on Core or Pro
The Brain is on Elite. That does not mean owners on the other plans are in the dark. Nearly every question above can be answered from the platform's own records. The difference is who does the assembling.
On Core. The calendar, the CRM, the POS, invoicing and payments are all in the plan. Questions 1, 4, 5, 9, 12 and the whole cash group are answerable by opening the schedule and the invoice and payment records. You do the looking.
On Pro. Call tracking, attribution, transcription, lead scoring and call recording are added. That makes the marketing group answerable from the call-tracking data, and it lets you listen to the calls behind question 2.
What the Brain adds. It removes the assembling, and it makes cross-layer questions and follow-ups cheap. Question 16, revenue collected by ad source, means matching attribution to payments. That is possible by hand on Pro and takes a sentence on Elite.
A reasonable approach on Core or Pro is to pick the handful of numbers you check on a schedule and keep them in one place. The short list worth pinning is covered in the service business KPI dashboard guide. When you notice that you keep wanting to ask one-off questions that the fixed list does not cover, that is the signal the Brain would earn its place. Because the plans are month-to-month, you can test that directly.
Calibrating trust in the first two weeks
Do not act on answers you have not checked. Spend the first two weeks asking questions you already know the answer to.
Start with yesterday. Ask how much was collected yesterday, then check it against your payment records. Ask how many jobs were booked, then count them on the calendar.
Test the edges. Ask about a date range that crosses a month end. Ask about a customer with an unusual history. Ask about something you know was never recorded and see that the answer says so.
Tighten your phrasing. Where an answer surprised you, look at the question first. Most surprises come from an ambiguous time range or measure.
Fix the data, not the assistant. If the answer is wrong because a job was never closed out or a payment never entered, the fix is in how the team records work.
After two weeks you will know which questions you can act on immediately and which need a second look. That is the same process you would use with a new office manager, and the idea of software as a team member with a job description is explored in AI employee for service businesses.
Where to start
Do not begin with all 30. Begin with the six daily close-out questions and ask them every working day for two weeks, checking the answers against your own records. Add the Monday review in week three. Add the marketing, cash and team checks one at a time after that.
Write the questions down in the exact wording that gives you a clean answer, and reuse that wording. Consistency is what turns a clever feature into an operating rhythm.
The Jarvis AI Brain is part of Elite at $1,200 per month. Core is $500 and Pro is $750, all with zero setup fees, unlimited users and month-to-month terms. Details are on the pricing page. To see these questions asked against your own operation, get in touch.
