Your 9-person shop is losing $40,000+ per year to manual tasks your office manager shouldn't be doing — and missing 1-2 calls daily that represent another $75,000-$150,000 in unbilled revenue. Three automations fix 80% of this in 90 days.

ServiceTitan has AI-assisted dispatch optimization that can pre-sort jobs by geography, skill match, and urgency. However, a 9-person plumbing shop has enough nuance — which tech has the right parts on the truck, who's already near a job site, which customer is a VIP — that the owner's morning review still adds real value. AI handles the initial sort and flags conflicts; owner makes final calls. This is a 40-50% time reduction, not a replacement.

AI can draft the initial response, pull standard pricing from a rate card, and send a templated estimate for common jobs (drain cleaning, water heater swap, basic leak). Complex scopes still need a plumber's eyes. The 2-3 day turnaround the owner flagged as a pain point is almost entirely fixable with an AI-drafted first response within 30 minutes of inquiry, even if the final number needs human sign-off. This is a speed problem more than a quality problem.

AI can flag POs that exceed a threshold, compare against historical pricing, and surface anomalies. But the actual approval decision on a $3,000 water heater order or a $12,000 pipe replacement job involves vendor relationships, job-specific context, and cash flow judgment that the owner holds in their head. AI as a pre-screener and anomaly detector is the right role here — not a decision-maker.

QuickBooks Online's built-in bank reconciliation AI is genuinely good at this task for a business this size. It auto-categorizes transactions, matches deposits to invoices, and flags exceptions. The owner still needs to review flagged items and approve the weekly close — that's 10 minutes, not 45. This is one of the clearest wins in the audit.

This is a near-perfect automation candidate. The trigger is clear (job marked complete in ServiceTitan), the message is templated, the channel is SMS/email, and the stakes of a slightly imperfect message are low. ServiceTitan's marketing automation or a tool like NiceJob can fire a personalized review request within 2 hours of job completion — no human needed. The only edge case is a job that went badly, which you handle by filtering on job status or customer satisfaction score before sending.

Do not automate. Hiring a plumber in Austin's tight labor market is a trust and judgment call. You are evaluating whether this person will be in a customer's home unsupervised, whether they'll show up reliably, and whether they'll fit a 9-person team. AI screening tools have documented bias issues, create legal exposure under EEOC guidelines, and will not catch the red flags an experienced owner catches in a 10-minute phone call. The time cost of a bad hire — lost productivity, customer complaints, re-hiring — far exceeds any screening efficiency gain.

Do not automate: Hiring decisions for tradespeople involve legal risk (EEOC compliance), trust assessment for in-home service workers, and cultural fit judgment that AI tools cannot reliably replicate. A bad hire in a 9-person shop is catastrophic. The time savings do not justify the risk.

Do not automate. These are two specific, named commercial relationships that likely represent a disproportionate share of revenue. Negotiating payment terms with a property management company is a relationship conversation, not a transaction. The owner's credibility, history, and judgment are the asset here. AI can help prepare talking points or draft a follow-up email after the call — that's the assist role — but the negotiation itself must stay human.

Do not automate: Commercial client negotiations involve relationship capital, revenue risk, and judgment calls about acceptable terms that require the owner's direct involvement. These two clients likely represent 15-25% of revenue. Automating this interaction risks damaging relationships that took years to build.

This is the single highest-impact automation in the audit. You are losing 1-2 calls per day — at an average ticket of $350-600, that's $90,000-$150,000 in annual revenue leakage. AI voice agents (specifically trained for home services) can answer, qualify, and book 70-80% of inbound calls without human involvement. Complex calls (commercial accounts, multi-day jobs, upset customers) still route to the office manager. The technology is production-grade in 2025-2026 for this exact use case.

This is a fully automatable task today. ServiceTitan has native automated appointment reminders. The trigger is clear (appointment scheduled for tomorrow), the message is templated, the channel is SMS/email, and there is zero judgment required. If you are not already running this automation, you are paying a human to do something a $0 feature in your existing software handles perfectly. Turn this on today.

ServiceTitan already supports automated invoice generation on job close and mobile payment collection by the tech in the field. If your plumbers are closing jobs in ServiceTitan on their phones, the invoice can auto-generate and send without the office manager touching it. The gap is usually that techs aren't closing jobs properly in the app. Fix the field behavior first, then the automation works. Human review needed for disputed amounts or commercial invoices with custom line items.

Gusto's AutoPilot feature literally runs payroll automatically on a schedule without human intervention, provided time entries are accurate. The office manager's time here is mostly reviewing time cards and clicking approve. If ServiceTitan time tracking feeds into Gusto (integration exists), this becomes a 10-minute review task, not a 45-minute manual process. The human review step before approval is appropriate and should stay.

Automated AR follow-up sequences are well-proven technology. ServiceTitan has this built in, and tools like Invoiced or even QuickBooks Online can run 30/60/90 day reminder sequences automatically. The office manager should only touch accounts that don't respond to automated reminders after 90 days — at which point it becomes a judgment call about the relationship. This is a clear time sink that automation handles better than humans (consistent, never forgets, never feels awkward about asking).

Amazon Business has auto-reorder features for recurring supplies. Parts ordering is more complex — it depends on job-specific needs, supplier relationships, and price negotiation that varies. Office supplies can be fully automated with Amazon Business Subscribe & Save. Parts ordering should stay human-driven but can be assisted by a reorder alert when truck inventory drops below threshold (ServiceTitan has inventory tracking for this).

Texas TDLR license renewals are annual, structured, and have real consequences if missed (plumbers can't legally work). AI can help by setting calendar reminders, drafting the renewal checklist, and pre-filling forms from stored data. But the actual submission requires human verification of CE hours, license numbers, and payment. This is a compliance task — AI assists, human executes and verifies. Do not fully automate a task where an error means your plumbers can't legally work.

Physical diagnosis of plumbing problems requires sensory input (sound, smell, visual inspection, pressure testing) that AI cannot replicate in the field in 2025-2026. There are AI-assisted camera inspection tools for sewer lines, but they assist the plumber's interpretation — they don't replace it. Revisit in 12-18 months as computer vision tools for pipe inspection mature.

This is the task the owner flagged as causing 2-3 day delays and lost jobs. AI can dramatically speed this up. If the lead plumber inputs job type, scope notes, and materials needed into ServiceTitan's mobile app, AI can draft the estimate from the pricebook in under 2 minutes. The plumber reviews, adjusts, and sends. This cuts estimate turnaround from 2-3 days to same-day. The diagnosis and scope judgment stays human — the writing and pricing lookup becomes AI-assisted.

Physical skilled trade work. Not automatable. Not relevant to this audit.

Do not automate: Physical skilled trade work requiring licensed professionals. No AI application exists or is appropriate. This is the core value delivery of the business.

Apprentice mentoring is a trust and safety relationship. A lead plumber teaching an apprentice how to sweat copper or diagnose a slab leak is transferring judgment, safety habits, and professional standards — not information. AI study tools can supplement the apprentice's exam prep (see prompt library), but the mentoring relationship itself should not be touched.

Do not automate: Skilled trade mentorship involves safety-critical knowledge transfer, professional judgment, and trust-based relationships. Automating this creates liability if an apprentice learns incorrect technique from an AI tool and causes property damage or injury.

Austin's Development Services Department permit process is structured but requires local knowledge, correct form selection, and sometimes in-person interaction. AI can help draft permit applications, check required documentation checklists, and draft scope descriptions. The actual submission and follow-up with the city requires a licensed plumber's signature and judgment. AI as a prep tool, not a replacement.

Route optimization AI (Google Maps, ServiceTitan dispatch) can reduce drive time between calls. This is already partially available in ServiceTitan. The plumber still drives — AI just sequences the route better. Realistic time savings: 15-20 minutes per day per tech through better sequencing.

The diagnosis is physical and stays human. The quoting can be significantly assisted by a well-built ServiceTitan pricebook — the plumber selects the job type and materials, the app generates the price, and the customer approves on a tablet. This is less about AI and more about building the pricebook infrastructure. Once built, quote time drops from 10-15 minutes to 2-3 minutes per job.

The repair is physical skilled work — not automatable. Photo capture is already a ServiceTitan feature. AI could theoretically auto-tag or organize photos, but the time savings are minimal. The real value is ensuring plumbers actually take and upload photos consistently, which is a training/process issue, not an AI issue.

Job summary writing is a strong AI candidate. The plumber speaks or types a few notes, and AI drafts a professional job summary for the customer record. ServiceTitan has voice-to-text and AI summary features in development; ChatGPT can do this today with a simple prompt. Payment collection via ServiceTitan mobile is already automated — the plumber just needs to close the job properly.

ServiceTitan inventory tracking can flag when truck stock drops below reorder thresholds and generate a restock list automatically. The physical restocking stays human. This is a process improvement more than an AI play — the value is in not having a plumber show up to a job without the right parts.

Physical labor assistance on job sites. Not automatable. Not relevant to this audit.

Do not automate: Physical job site assistance requiring human presence, safety awareness, and real-time judgment. No AI application is appropriate or available.

Physical labor. Not automatable. Not relevant to this audit.

Do not automate: Physical skilled trade labor. No AI application exists or is appropriate for this task.

AI tutoring for trade licensing exams is genuinely useful and well-supported by current LLM capabilities. An apprentice can use ChatGPT or Claude to quiz themselves on Texas plumbing code, explain concepts they don't understand, and generate practice problems. This doesn't replace a formal prep course but meaningfully supplements it. The investment is $20/month and the payoff is faster licensure — which means higher billing rates sooner.