Loading...
Built for HVAC, plumbing, electrical, and recurring service teams.
Open betaUnlimited users. Pricing from $200/month.
Built for HVAC, plumbing, electrical, and recurring service teams.
Loading...
Thorbis AI monitors every job, invoice, customer, and technician schedule every 60 seconds — and tells your dispatcher exactly what to do next, ordered by risk.
Invoice #847 — 22 days overdue, $3,200
Williams HVAC has always paid within 10 days. No contact in 3 weeks. Call immediately — risk of dispute.
Suggested: Call now and offer a payment plan
Marcus has a 2-hour gap at 11 AM today
Rodriguez maintenance job is 12 min away. Pending since Monday — customer has been asking for an earlier slot.
Suggested: Assign Rodriguez job to Marcus 11:00–1:00 PM
Johnson estimate has had no response for 3 days
AC estimate for $1,450 sent Tuesday. Season is peaking — customer may be shopping around.
Suggested: Send follow-up SMS with a 3-day hold offer
Six categories of operational risk, all tracked automatically — nothing slips through the cracks
Every Pending Action belongs to one of four categories. Each has a distinct icon, a different kind of context, and a different recommended response pattern.
Outreach actions — follow-ups, check-ins, reminders
Triggered when a lead or customer has gone silent past a defined threshold, an estimate hasn't been approved, or a job completion survey hasn't been acknowledged.
Example action
"Follow up with Johnson — AC estimate ($1,450) has been open 3 days. Season is peaking. Customer may be shopping around."
Suggested: Send follow-up SMS with a 3-day hold offer
Payment actions — overdue invoices, collection risk
Triggered when an invoice passes the due date, when a customer with a strong payment history suddenly goes quiet, or when a disputed charge needs a response before it escalates.
Example action
"Invoice #847 is 14 days overdue — $3,200. Williams HVAC has always paid within 10 days. No contact since the invoice was sent."
Suggested: Call now — customer has paid reliably for 2 years
Route actions — gaps, reassignments, capacity
Triggered when a technician's route has a gap that could accommodate a waiting job, when a job is running long and threatens the next appointment, or when a technician calls out and their jobs need redistribution.
Example action
"Tech Marcus has a 2-hour gap at 11 AM. Rodriguez maintenance job is 12 min away and has been pending since Monday."
Suggested: Assign Rodriguez job to Marcus 11:00 AM — 1:00 PM
Relationship actions — churn risk, equipment, retention
Triggered when a customer hasn't booked in 12+ months, when their equipment is approaching a manufacturer service interval, when a warranty is expiring, or when a negative review has gone unacknowledged.
Example action
"Customer Sarah Chen hasn't booked in 14 months. Equipment warranty expires next month. High lifetime value — likely to switch if not contacted."
Suggested: Call with a pre-season maintenance offer
Every action is assigned a risk level the moment it's generated. Dispatchers see the most urgent issues first — never buried in noise.
Act right now
Immediate action required. Revenue loss, customer escalation, or safety risk is imminent.
Act today
Revenue at active risk. Delay makes the problem meaningfully harder to resolve.
Act this week
An opportunity or growing problem. Won't explode today but will if ignored.
Informational
Routine suggestions and informational nudges. No urgency, but worth acting on when time allows.
Every 60 seconds, the AI runs through your entire operation. Here's exactly what it analyzes and what triggers a Pending Action.
Scans all open invoices for days past due. Cross-references each customer's payment history to determine whether late payment is unusual — which elevates the risk level.
Checks every unsent or unapproved estimate against the configured follow-up window. Factors in job value and season to determine urgency — high-value summer AC jobs escalate faster.
Analyzes every tech's schedule for open windows longer than the minimum job time. Queries pending jobs sorted by distance from the gap location to find fillable matches.
Compares current job elapsed time against the original estimate. Flags jobs that are running long before they impact the technician's next appointment or the customer's patience.
Calculates months since last booking for every active customer. Customers crossing the 12-month threshold for the first time generate a churn-risk action immediately.
Checks every piece of customer equipment against its scheduled maintenance date or manufacturer service interval. Generates outreach actions 60, 30, and 7 days before expiration.
Monitors open complaints, negative reviews, and disputed charges for acknowledgment. Escalates risk level every 24 hours without a dispatcher response or resolution note.
Scans all active SMS, email, and in-app threads for unanswered messages. Flags any thread where a customer message has not received a response within the configured SLA window.
Tracks how long each lead has been in each pipeline stage. Generates actions when a lead has been stalled longer than the median for that stage — indicating it needs human intervention.
The AI does the analysis. The dispatcher makes the call. Here's exactly how the workflow flows from detection to resolution.
The monitoring engine completes its full scan of all active records — invoices, jobs, schedules, customers, communications. When a threshold is crossed for the first time (invoice goes past due, estimate sits unanswered, route gap opens up), a new Pending Action is created and assigned a risk level.
The new action slides into the Pending Actions sidebar panel, sorted above all lower-risk actions. If it's Critical, the header badge count increments and begins pulsing red. The action displays its type icon, risk badge, customer name, job reference, the AI's description of what it found, and the specific suggested action text.
The dispatcher reads the action — customer name, what was detected, and the recommended response. They can click the customer or job reference to open the full record if they need more context. Most actions are clear enough to act on immediately from the panel itself.
Clicking Accept triggers the suggested action: sends the follow-up SMS, assigns the job to the technician, opens the invoice for collection follow-up, or flags the customer record for a callback. Clicking Dismiss records why it was dismissed (optional) and removes it from the queue — so the panel stays clean and only shows things that still need attention.
If an action is neither accepted nor dismissed, the AI continues monitoring the underlying issue. If the situation worsens — the invoice gets another day older, the schedule gap closes without a job assigned — the risk level can escalate from Medium to High or High to Critical. Escalated actions are re-surfaced at the top of the queue so they can't be missed.
The AI doesn't replace your dispatchers — it makes each one dramatically more effective by removing the cognitive load of watching everything at once.
Every overdue invoice, every unresponsive lead, every schedule gap — the AI catches it before it costs you money. Dispatchers stop firefighting and start executing.
Instead of waiting for customers to call and complain, the AI flags problems while they're still small — before they escalate into disputes, churn, or unpaid invoices.
Critical and High actions rise to the top. Low-priority suggestions stay visible but don't compete with urgent issues. Dispatchers work the queue from top to bottom with confidence.
The AI recommends. The dispatcher decides. Nothing is executed automatically — every action requires a human Accept. This prevents automation errors on sensitive customer interactions.
No training data required. The AI begins monitoring the moment you're live on Thorbis. It uses your existing job, invoice, and customer records to generate its first actions within minutes.
Operational intelligence is part of the Thorbis platform. There's no separate AI module, no usage fees per action, and no configuration required to start receiving Pending Actions.
One recovered overdue invoice. One schedule gap filled. One at-risk customer retained. Most Thorbis customers report covering their platform cost within the first month just from actions the AI surfaced that would have otherwise been missed.
These aren't edge cases. These are the situations that fall through the cracks every day in a busy dispatch operation.
Overdue invoice
Invoice #847 goes 22 days overdue. Dispatcher only notices when the owner asks about cash flow at end of month. By then the customer has already disputed the charge.
Schedule gap
Marcus finishes a job at 10:45 AM and his next appointment is at 1:00 PM. He drives back to the shop and waits. The Rodriguez job sits pending all day. Customer calls in frustrated.
Estimate gone cold
Johnson's AC estimate sat for 5 days with no follow-up. Customer booked with a competitor. $1,450 job lost. The dispatcher didn't know the estimate was still open.
Customer churns silently
Sarah Chen hasn't booked in 14 months. No one noticed. Her equipment warranty expired and she found a new HVAC company when it broke down. A $4,000/year account lost.
Late job cascade
Tech runs 45 minutes over on a repair. His 2 PM customer calls in angry. Dispatcher scrambles to manage the situation — no one had visibility until the customer called.
Overdue invoice
AI flags Invoice #847 at day 8 when it crosses the threshold unusual for this customer. Dispatcher calls the same morning. Invoice paid within 2 days. No dispute, no collections.
Schedule gap
AI detects Marcus's gap at 11 AM and matches it to the Rodriguez job 12 minutes away. Dispatcher clicks Accept. Rodriguez gets called and scheduled. Marcus fills the gap. 2 billable hours recovered.
Estimate gone cold
AI flags Johnson's estimate on day 3. Dispatcher clicks Accept — sends a follow-up SMS with a 3-day hold. Johnson replies within an hour and approves. Job booked. $1,450 captured.
Customer churns silently
AI surfaces Sarah Chen at the 12-month mark with a warranty expiration notice. Dispatcher clicks Accept — sends a pre-season maintenance offer. Sarah books a tune-up. Account retained.
Late job cascade
AI detects the job running long at the 30-minute mark. Dispatcher proactively texts the 2 PM customer with an updated arrival window. Customer appreciates the heads-up. No angry call.
Most service businesses track operations with spreadsheets or basic software alerts. Here's what you get with Thorbis Pending Actions instead.
| Capability | Spreadsheets | Basic Software Alerts | Thorbis AI |
|---|---|---|---|
| Monitors invoices automatically | |||
| Considers customer payment history when flagging | |||
| Detects and fills schedule gaps | |||
| Tracks estimate response time | |||
| Surfaces customer churn risk proactively | |||
| Risk-levels each issue (Critical / High / Medium / Low) | |||
| Escalates unaddressed issues automatically | |||
| Provides dispatcher-ready suggested action text | |||
| Updates every 60 seconds without manual refresh | |||
| Works from existing data — no setup required |
Basic software alerts = generic threshold-based notifications without context or prioritization
Answers to the questions dispatchers and owners ask most before going live with Pending Actions.
Every 60 seconds, the AI runs a complete scan of your active records — open invoices, unresolved estimates, technician schedules, customer booking history, equipment records, and communication threads. It compares the current state of each record against thresholds you define (or sensible defaults if you don't). When a threshold is crossed for the first time, a new Pending Action is created and assigned a risk level based on the severity of the issue and how it compares to historical patterns for that customer or job type.
The AI generates four action types. Send Communication actions identify customers or leads who need a follow-up — an estimate that's been open 3 days, a job completion survey that was never acknowledged. Financial actions flag payment issues like overdue invoices or customers whose payment patterns have suddenly changed. Scheduling actions detect gaps in technician routes and surface pending jobs that could fill them. Customer actions track relationship signals — customers who haven't booked in 12+ months, equipment approaching a manufacturer service interval, warranty expirations.
Critical means immediate action is required — within the hour, not by end of day. This includes emergency jobs without a technician assigned, invoices 21+ days overdue with broken customer contact, technicians who haven't checked in during an active job for 4+ hours, or negative reviews with customers threatening a chargeback. Critical actions show an animated pulsing indicator. High means revenue is actively at risk but there's still time to act today — invoices 14 to 20 days overdue, schedule gaps that reduce today's billable hours, or estimates not approved after 5+ days on a high-value job. High actions show in the same destructive color without the pulse.
By design, the AI recommends actions — it never executes them automatically. Your dispatcher reviews each Pending Action and clicks Accept to execute or Dismiss to remove it. This keeps your team in control of every customer interaction and every financial decision. The AI's job is to make sure nothing is missed, not to act on behalf of your business without human judgment. This also means your team can course-correct instantly if a recommendation doesn't apply — for example, dismissing a follow-up action because the dispatcher already spoke to the customer by phone.
If a Pending Action is left unaddressed, the AI continues monitoring the underlying issue. As the situation worsens — the invoice ages another day, the estimate goes another 24 hours without approval — the risk level can escalate. A Medium action can become High. A High action can become Critical. When an action escalates, it moves to the top of the queue and the header badge updates to reflect the new count of Critical items. This escalation loop ensures that ignoring a problem actively raises its visibility rather than letting it hide at a fixed risk level.
No. The AI begins monitoring your operation from the moment you're live on Thorbis — no configuration required, no training data to upload, no custom model to build. It uses your existing jobs, invoices, customer records, and communication history to generate its first Pending Actions within minutes of going live. Default thresholds are set based on field service industry norms, but you can adjust follow-up windows, invoice aging thresholds, and customer inactivity periods to match how your specific business operates.
Discover how these features work together to power your field service business
Join service businesses that use Thorbis AI to catch every overdue invoice, fill every schedule gap, and retain every at-risk customer — automatically.