What a 90-Day Job Data Audit Reveals About AI Readiness

A home service owner who suspects revenue is leaking rarely knows where to look, and that vagueness is expensive. A 90-day pull of the job data already sitting in the company's own platforms makes the loss specific and dollar-denominated before a single dollar goes to AI. The audit answers two questions at once: where revenue is leaking right now, and where the operation is too broken for automation to hold.

Why a 90-day data pull is the right starting point before any AI decision

The data needed for this exercise already exists inside the platforms owners run. ServiceTitan, Housecall Pro, and Jobber all export job records natively, with date-range filters and CSV downloads built into each product. The discipline is measurement, not technology. No new tooling is required to pull three months of call logs, estimates, dispatch records, and invoices, and that is precisely the point. An owner who cannot produce this data from their own stack has learned something important before any vendor conversation begins.

The audit puts a dollar figure on what the owner already senses. Most home service owners feel they are losing revenue somewhere, but somewhere is not actionable, and it certainly does not justify a five-figure automation investment. A 90-day pull converts that feeling into a specific list: unanswered calls at a determinable value per missed booking, quotes sent without follow-up at a determinable value per lost job, invoices sitting unsent at a determinable value per completed job. The losses become numbers, and numbers can be prioritized.

This matters because AI is not the first decision an owner faces. The first decision is whether the foundation can support automation at all. The 90-day pull answers both questions simultaneously because every data layer carries two kinds of information: a revenue figure that shows where money is leaking, and a structural signal that shows whether the processes generating that data are consistent enough to automate.

What the call data layer exposes about the front door of the business

Call data is almost always the first place a 90-day audit finds large, recoverable losses, because the front door of a home service business is a phone line that often rings into silence. Published figures for unanswered calls in the trades vary wildly, from roughly a quarter to well over half depending on which vendor is counting, and most of them come from companies selling answering tools. That is exactly why your own call log matters more than any benchmark: it shows how many calls went unanswered, at what times, and whether anyone called back. Many owners never learn these calls happened at all.

The cost of that silence is easy to underestimate. A homeowner with a leaking pipe or a dead AC unit rarely waits for a callback. Plenty of them simply call the next company on the list, and an unanswered call that never gets returned is a job that may have gone to someone else. The audit puts a number on it: unanswered calls, multiplied by your own booking rate and average ticket.

The reason is structural, not behavioral. Technicians are on job sites during peak call hours and cannot answer. The phone rings and the caller moves on. This is not a discipline failure. It is a capacity problem, and capacity problems are exactly what automation can close. The owner does not need to scold the team. The owner needs a system that answers when the team cannot.

After-hours calls are part of the picture, but a smaller part than many vendors suggest. ServiceTitan's 2025 platform data puts the share of inbound calls to residential HVAC shops that arrive outside weekday business hours at 9.8% to 14.1%, peaking in summer. For many companies the bigger leak is during the day, when technicians are on jobs and the office is stretched. The 90-day log tells you which one is yours, and that decides what is worth fixing first.

The call layer also functions as an AI-readiness diagnostic. If call routing is ad hoc, if there is no agreed script for intake, if different staff handle calls differently, the business is not ready to hand calls to an AI. The system would simply replicate the chaos at scale, answering every call with whatever inconsistent process it absorbed from the data. Conversely, if the call log shows a consistent, repeatable intake pattern and the only problem is volume or hours, that is the condition where AI call handling has the best chance of paying off: it is copying a process that already works, just at hours and volumes the team cannot cover.

What the estimate and follow-up data reveals about the quote pipeline

The estimate layer of the audit consistently reveals that a large portion of quoting effort produces zero revenue, not because the price was wrong, but because nobody followed up. This is a different kind of leak than the front door. The lead came in, the conversation happened, someone drove to the house and prepared a quote, and then the job died from neglect. The gap between quotes sent and quotes followed up is almost always large.

Time decay compounds the neglect. The longer a quote sits without a call, the colder the customer gets, and the job data timestamps make this visible in a way the owner's memory cannot. A quote sent on a Tuesday with no follow-up until the following Monday is not a slow process. It is a lost job.

What this layer reveals about AI readiness is a question of data hygiene. Automated follow-up sequences work when the estimate data is clean: job type recorded, quote amount attached, customer contact complete. A sequence that sends the wrong quote amount or addresses the customer by the wrong name does more damage than no sequence. Aseado Solutions builds inside the platforms owners already run, ServiceTitan, Housecall Pro, and Jobber, and addresses these foundation issues before layering automation on top.

The audit also checks whether follow-up is currently happening manually and inconsistently. If it is, that signals a missing SOP, a foundation problem that must be solved before an automated sequence can replace it. The audit finds both conditions in the same log: the revenue leak the owner suspected, and the structural gap that determines whether any tool can close it.

What dispatch and scheduling data reveals about operational readiness in the field

Dispatch data is the layer where the audit most often finds a business that is not broken enough to notice but not clean enough for AI to use. The operation runs, jobs get done, customers are mostly satisfied, and underneath that smooth surface the data is too inconsistent for an automated system to act on correctly. A 90-day dispatch log shows how jobs are assigned, whether by geography, skill, availability, or whoever is closest to the phone, how often jobs are rescheduled or reassigned, and how often job types are recorded consistently enough to match a technician skill profile.

Inconsistent job-type tagging is the clearest AI-readiness failure in this layer. When AC repair, HVAC service, and cooling issue all mean the same thing, a dispatching model trained on that data will misclassify jobs and send the wrong technician. The system produces confident-looking assignments that waste a truck roll, and the owner blames the AI when the real problem was the ten different labels the team used for the same job type.

Rescheduling frequency functions as a proxy for scheduling quality. A dispatch log with high reschedule rates signals that the underlying scheduling logic lives in someone's head, not in any documented form. That logic cannot be handed to an automated system because it does not exist outside the person who applies it. Where it shows ad hoc decisions made by whoever is available, the SOPs must come first.

This layer most often surprises owners. They expect the audit to find a phone problem or a follow-up problem, and it does, but the dispatch data shows something further downstream. Even if the front door is fixed and every quote is followed up, jobs will still fall through without a cleaner operational foundation in the field. The leak is not in one place; it runs through the whole job lifecycle.

What the invoice and payment data reveals about back-office leaks and automation prerequisites

Invoice data is where the audit finds the quietest leaks. Jobs are completed but invoiced late. Invoices are sent but never followed up. Manual data entry introduces errors that distort every other number in the business. None of these failures makes noise. The customer has already received the service, and the work shows up as done in the job log, so the owner assumes the revenue arrived. Each of these is a recoverable revenue number, not a cost.

Manual data entry shows up as inconsistency that is easy to spot in a 90-day pull. Duplicate records for the same customer, phone numbers in the wrong field, job totals that do not match the invoice. Any of these cause automated reminders and follow-up sequences to fire incorrectly or not fire at all. The error produces complaints, not efficiency.

The invoice layer also reveals whether QuickBooks or the accounting integration is receiving clean data from the FSM platform. If job records in the FSM platform do not match what is in QuickBooks, automation built on either system will compound the mismatch. The numbers will be wrong everywhere, and wrong numbers traveling faster through automation only become wrong numbers with more confidence.

Where the invoice data is clean and the only problem is timing and volume of follow-up, automated payment reminders and invoice sequences are among the fastest-payback automations available to a home service business. The audit does not just find four separate leaks. It maps the whole job lifecycle, from the first call through the final payment, and shows which parts of it are ready for automation and which need fixing first. That synthesis is the real deliverable.

How the audit separates fix-the-operation-first findings from automate-this-now findings

The most useful output of the audit is not a list of revenue leaks. It is the same list sorted into two piles: what to fix in the operation first, and what is ready to automate now. A sorted list is a plan.

Fix-first findings share a common structure. The underlying process is undocumented, inconsistently executed, or missing data that automation would need to act on correctly. These are not AI problems. They are operational problems that AI would make worse.

Automate-now findings also share a common structure. Follow-up sequences, invoice reminders, booking confirmation texts, and after-hours call handling are the clearest examples. The work is routine, the rules are stable, and the only friction is volume. That is the profile of automation most likely to pay for itself early.

The distinction matters because AI built on a broken foundation does not fail with an error message. It produces confident-looking outputs from bad inputs, and the team discovers the failure weeks later when customers complain or revenue numbers do not move. The priority order is the discipline that precedes deployment.

Why most home service AI projects fail after launch, and what the audit prevents

Most AI projects in small businesses do not fail on day one. They fail quietly over weeks as the team reverts to the old way, and the new system sits in the software stack producing outputs nobody reads. An audit changes that, because the owner who ran the numbers knows what a working system is worth and notices quickly when it is not delivering.

The root cause is rarely technical. AI built on what sounds good rather than what the data shows is the most common source of the problem. The impressive system launches, encounters the dirty data underneath it, and quietly stops being used.

Founder dependency compounds this in home services specifically. When the real rules for pricing, scheduling and exceptions live in the owner's head, the AI has no basis for handling edge cases, and the team overrides it constantly until it is effectively switched off. The audit requires that the process be understood and documented before any system is built on it, which is the condition most AI projects skip.

Post-launch monitoring, fixes, and team training are not optional extras. They are the difference between a system that works at month three and one that is quietly ignored by month two. The audit does not guarantee a successful AI deployment. It removes the conditions that make failure likely.

What to look for when choosing a partner to run the audit and build what it finds

The quality of the audit depends entirely on whether the partner starts with the business's actual data or with a predetermined tool recommendation. The right partner asks to see the data architecture before quoting anything. A partner who quotes a fixed price before seeing the data, or who leads with a specific vendor's product without understanding the current stack, is selling a tool, not solving a problem.

Partner type should match business stage. DIY plus an AI subscription is viable for the smallest operations where the ROI of a paid engagement is harder to justify. The owner should not pay for more engagement than the business can absorb. The partner should build inside the software the business already runs: ServiceTitan, Housecall Pro, Jobber, QuickBooks. Look for written scope before each build, ongoing support after launch, month-to-month terms, and a clean handover if you ever want to run it yourself. Those terms describe a partner that wants the client to succeed, not to stay dependent.

Aseado Solutions works this way for owner-operated home service companies at the mid-market level. The 10-Day AI Activation Audit maps every non-job task, builds one production-ready system on the biggest opportunity within the same engagement window, and its founders, not an account team, stay in a shared Slack channel with the client on a month-to-month engagement with no lock-in.

The decision in front of the owner is not which AI tool to buy. It is whether anyone in the business actually knows what the data shows. Everything else follows from there.