Your 9-person shop is losing $47,000/year in recoverable labor and missed revenue — your office manager spends 22+ hours weekly on tasks AI handles today, and you're dropping 1-2 calls daily that represent $75,000-$150,000 in annual revenue at your ticket averages.

ServiceTitan has AI-assisted dispatch optimization built in, and it can surface scheduling conflicts, drive-time optimization, and tech skill matching. But this owner is making judgment calls about which plumber handles which commercial client, managing personality dynamics, and accounting for truck inventory — none of which AI handles cleanly. AI reduces the cognitive load and prep time; it does not replace the morning dispatch decision.

AI can draft the initial response, acknowledge the request, ask clarifying questions, and even generate a ballpark range for common jobs (water heater swap, drain clean) using your historical pricing. But the owner is currently doing this because lead plumbers are slow — the real fix is a structured estimate workflow, not just AI drafting. AI gets you a faster first touch; a human still needs to confirm scope and price. This also directly addresses the pain point about losing jobs to faster competitors.

AI can flag POs that exceed a threshold, compare against historical pricing, and surface whether a vendor's quote is above market. But the approval decision itself — especially for $2,000+ equipment on a job with a thin margin — requires the owner's judgment about cash flow, job profitability, and vendor relationship. AI as a pre-screener and price-checker is valuable; AI as the approver is not appropriate here.

QuickBooks Online's AI reconciliation is genuinely good for a business this size. It auto-matches transactions, learns your categorization patterns, and flags exceptions. The owner still needs to review flagged items and approve the reconciliation — but the manual matching work is largely gone. This is one of the clearest wins in the audit: structured data, mature tooling, low blast radius on errors caught in review.

This is a near-perfect automation candidate. The trigger is clear (job marked complete in ServiceTitan), the message is templated with light personalization, the channel is SMS/email, and the stakes of a slightly imperfect message are low. ServiceTitan Marketing Pro or a Zapier workflow handles this end-to-end. The only human element worth preserving: if a job had a complaint or callback, you do not want an automated review request going out. That exception logic is easy to build.

Do not automate. Hiring a plumber in a 9-person shop is a trust and culture decision as much as a skills decision. A bad hire in a field service business damages customer relationships, creates liability, and is expensive to unwind. AI screening tools have documented bias issues in skilled trades hiring, and the legal exposure for a small employer using automated screening without proper compliance review is real. The owner's judgment here is irreplaceable.

Do not automate: Hiring decisions in a 9-person field service business are trust, culture, and liability decisions. AI screening tools carry documented bias risks in skilled trades contexts and create legal exposure for small employers without HR compliance infrastructure. A bad plumber hire damages customer relationships and creates workers' comp and licensing liability. The owner's direct judgment is not replaceable here — and the frequency (a few hires per year) does not justify the risk of automating.

Do not automate. These are two specific, named commercial relationships that represent a meaningful share of revenue for a $1.5-2.5M shop. Negotiating payment terms is a relationship and trust exercise. An AI-drafted negotiation email to a property manager who already knows the owner personally would be immediately obvious and potentially damaging. The owner's credibility and relationship equity are the asset here.

Do not automate: These are named, high-value commercial relationships. Negotiating net-30 terms is a trust and relationship exercise where the owner's personal credibility is the leverage. Automating this — even with AI drafting — risks signaling to property managers that they are not worth the owner's personal attention, which could jeopardize contracts that likely represent $200K-$500K in annual revenue. AI can assist with prep (drafting talking points, modeling cash flow impact of different terms), but the negotiation itself must be human.

This is the single highest-impact automation in the audit and directly addresses the owner's stated pain point of dropping 1-2 calls daily. AI voice agents for home services booking are production-grade in 2026 — they handle job type triage, address capture, scheduling against ServiceTitan availability, and confirmation. The 'mostly' rather than 'full' rating reflects that complex calls (upset customers, unusual job types, commercial accounts) still need human handling. Realistically, 70-80% of inbound booking calls are routine and fully handleable by AI.

This is the cleanest full-automation in the audit. The trigger is deterministic (appointment scheduled for tomorrow), the message is templated, the channel is SMS/email, and the content is low-stakes. ServiceTitan does this natively. If you are not already running automated appointment confirmations, you are manually doing something a $0 feature handles. There is no meaningful human element here.

ServiceTitan generates invoices from completed job records automatically, and payment collection in the field (via the tech's mobile app) is already a mature workflow. The office manager's role here is largely exception handling: jobs where the tech didn't collect, commercial accounts on net terms, and disputed line items. AI handles the routine 80%; the office manager handles the 20% that are exceptions. This is already partially automated in most ServiceTitan shops — the question is whether yours is configured correctly.

Gusto's automated payroll is genuinely good for a shop this size. With direct deposit and auto-run enabled, payroll runs itself once hours are approved. The office manager's real work is approving timesheets and catching anomalies (overtime, missing punches). That review step should stay human — but the actual payroll submission is automatable. The integration between ServiceTitan time tracking and Gusto is the key unlock here.

AR follow-up sequences are a solved problem in 2026. The trigger is clear (invoice age), the message escalates by bucket (30/60/90 days), and the channel is SMS/email. ServiceTitan has this natively; so does QuickBooks Online. The exception: your two commercial property-management clients on net-30 terms should NOT be on an automated dunning sequence — those relationships require a personal call from the owner. Build the automation with an exclusion list for commercial accounts.

Amazon Business has auto-reorder features for recurring supplies, and AI can help maintain a standard order list. But parts ordering for a plumbing shop is not purely routine — job-specific parts, substitutions when items are out of stock, and vendor relationships (Ferguson, Waxman) require judgment. Automate the consumables (office supplies, common fittings in bulk); keep human oversight on job-specific parts.

Texas TDLR license renewals are a structured annual task, but the filing itself requires accurate license numbers, CE completion verification, and payment — errors have real consequences (lapsed licenses mean techs cannot legally work). AI can draft a renewal checklist, set calendar reminders, and help the office manager track CE hours across the team. It should not submit the actual renewal. The risk of an AI error on a regulatory filing is too high for a 9-person shop where one lapsed license grounds a tech.

On-site physical diagnosis of plumbing systems requires sensory input (sound, smell, visual inspection, pressure testing) that no current AI tool handles in a field service context. AI diagnostic assistants for plumbing are early-stage at best. Revisit in 12-18 months as camera inspection AI (for sewer lines specifically) matures. Do not invest here now.

This directly addresses the owner's pain point about 2-3 day estimate turnaround. AI cannot generate the estimate from scratch — the plumber still needs to diagnose and scope the job. But AI can dramatically accelerate the write-up: given job type, scope notes, and your standard price book, an AI tool can draft a professional estimate in 2-3 minutes instead of 20-30. ServiceTitan's estimate templates plus a GPT-assisted write-up workflow is the right approach. The plumber reviews and approves; AI does the drafting.

Physical skilled trade work. Not automatable by any current or near-future AI. This is the core value delivery of the business.

Do not automate: Physical skilled trade work requiring licensed plumbers. No AI tool performs pipe replacement, water heater installation, or re-pipes. This is the irreplaceable core of the business and should not be in scope for any automation discussion.

Apprentice mentoring in a field service context is a safety, skills, and relationship function. A lead plumber teaching an apprentice to sweat copper or diagnose a slab leak is not a task that benefits from AI involvement. AI study tools can supplement the apprentice's exam prep (see apprentice role), but the mentoring relationship itself is human.

Do not automate: On-the-job mentoring in a licensed trade is a safety and skills transfer function that requires direct human supervision. Automating or AI-augmenting this creates liability if an apprentice learns incorrect technique from an AI tool and causes property damage or injury. The lead plumber's mentoring role is also a retention and culture function — apprentices who feel invested in stay longer.

Austin's Development Services Department permit process is structured but requires navigating a specific portal, submitting correct documentation, and following up on status. AI can help draft permit application narratives, check document checklists, and track submission status. But the actual submission and any back-and-forth with the city requires a licensed plumber's signature and judgment. AI as a prep and tracking tool; human for submission and follow-up.

Route optimization AI (Google Maps, ServiceTitan dispatch board) can reduce drive time between calls. This is already available and worth using. But the plumber is driving — AI is just optimizing the sequence. Meaningful but not transformative at a 3-tech scale in Austin.

The diagnosis is physical and irreplaceable. The quote can be accelerated with a well-built ServiceTitan price book — the tech selects the job type and line items, and the price populates automatically. This is a configuration improvement, not an AI deployment. The customer approval conversation is a sales and trust interaction that should stay human.

The repair itself is physical and irreplaceable. Photo capture is already a ServiceTitan mobile workflow. AI can auto-tag and organize photos by job, and some tools can flag if photos are missing before job close. This is a minor efficiency gain, not a transformation.

Job summary writing is a strong AI candidate. The tech has the job type, parts used, and work performed — AI can draft the customer-facing summary from those inputs in seconds. Payment collection via ServiceTitan mobile is already streamlined. The main friction is techs who skip the summary or write one-liners. AI drafting from structured inputs (job type + parts + duration) removes the blank-page problem.

ServiceTitan tracks parts used per job and can generate a restock list automatically. This is a configuration improvement — the tech still physically restocks the truck, but AI tells them exactly what to pull from the shop inventory. Saves the mental overhead of remembering what was used across 4-6 calls.

Physical labor assistance on licensed trade work. Not automatable. Not in scope.

Do not automate: Physical labor on licensed plumbing jobs. No AI application. This task exists because the work requires human hands and is supervised by a licensed plumber for safety and code compliance reasons.

Physical labor. Not automatable by any current AI. Not in scope for this audit.

Do not automate: Physical site preparation and demolition work. No AI application exists or is near-term for this type of task in a field service context.

AI tutoring for trade licensing exams is a legitimate and growing use case. Texas journeyman plumber exam prep has structured content (TDLR exam domains, code references, math). AI can generate practice questions, explain code sections, quiz on weak areas, and provide on-demand explanations. This is not a business cost savings — it's an investment in apprentice advancement that reduces the time to journeyman licensure, which increases their billable rate and reduces your dependency on expensive lead plumbers.