Your office manager is spending roughly 22 hours a week on tasks AI can fully or mostly handle today — that's $28,600/year in recoverable labor, and it doesn't count the 1-2 missed calls daily that are likely costing you $40,000-$80,000 in lost revenue annually at your average ticket size.

ServiceTitan has AI-assisted dispatch optimization built in, and tools like Workiz and Jobber offer smart scheduling. However, dispatching plumbers in a 9-person shop involves real judgment: who's closest, who has the right skills for the job type, who's already in a difficult situation on a prior job. AI can surface the optimal schedule and flag conflicts, but the owner's contextual knowledge of each tech's current status makes full automation risky. This is a human-AI hybrid — AI does the math, owner makes the call in 2 minutes instead of 20.

This is your stated pain point — 2-3 day estimate turnaround is killing close rates. AI can handle the first-response layer: acknowledge the request, ask qualifying questions (what's the issue, property type, urgency, photos), and generate a ballpark range from your historical job data. The final number still needs a plumber's eye, but AI can cut the response time from 2-3 days to 2-3 hours by doing the intake work automatically. Tools like Hatch or even a ServiceTitan-connected AI intake form can do this today.

AI can flag anomalies — a PO that's 40% above your typical spend for that part category, a vendor you've never used, a quantity that doesn't match the job scope. But the approval decision itself involves supplier relationships, cash flow timing, and job context that the owner holds in their head. Use AI to surface the data and flag outliers; keep the human in the approval seat. QuickBooks Online's anomaly detection and Bill.com's approval workflows handle the assist layer well.

Bank reconciliation in QBO is one of the most mature AI automation use cases in small business finance. QBO's bank feed with AI categorization handles 85-90% of transaction matching automatically. The remaining 10-15% are genuinely ambiguous transactions that need human eyes — split transactions, unusual vendors, mismatched amounts. Owner should spend 15 minutes reviewing exceptions weekly, not 2 hours doing it all manually. This is a quick win.

This is a textbook automation win. The trigger is clear (job marked complete in ServiceTitan), the message is templated, the channel is SMS or email, and the stakes of a slightly imperfect message are low. Tools like NiceJob, Podium, and ServiceTitan's built-in review automation handle this natively. The only human judgment needed: don't send a review request after a job that went badly. That exception logic is easy to configure (e.g., only trigger if invoice is paid and no complaint flag).

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 small crew poisons the whole team. AI screening tools have documented bias issues, and using them for skilled trades hiring in Texas creates legal exposure under EEOC guidelines. More practically: the owner's gut read on whether someone will show up reliably, treat customers well, and work safely is irreplaceable here. AI can help post job listings and screen resumes for license status — that's it.

Do not automate: Hiring decisions for a 9-person skilled trades shop involve trust, culture fit, and safety judgment that AI cannot reliably assess. AI screening tools carry documented bias risks and EEOC legal exposure. A bad plumber hire creates liability, customer complaints, and team dysfunction that costs far more than the time saved. The owner should own this process entirely.

Do not automate. These are your two commercial anchor clients. The relationship, the trust, and the ability to read the room in a negotiation are entirely human. An AI-drafted negotiation email to a property manager who's been your client for three years will read as impersonal at best and insulting at worst. The owner's personal relationship is the competitive moat here — protect it.

Do not automate: Commercial client negotiations depend on relationship capital built over time. These two clients likely represent a disproportionate share of revenue. Automating or delegating this negotiation to AI risks the relationship and the contract. This is the owner's job, full stop.

This is your single highest-impact automation. You're losing 1-2 calls per day — at an average residential ticket of $350-600, that's $90,000-$220,000 in annual revenue leakage. AI voice agents for home services (specifically ServiceTitan-integrated ones) can answer calls 24/7, qualify the job type, check availability, and book the appointment directly into ServiceTitan. They handle 70-80% of inbound calls without human intervention. Complex calls (commercial clients, warranty disputes, angry customers) route to a human. This is not experimental — Hatch, Numa, and ServiceTitan's own AI receptionist are in production at shops your size.

This is pure automation. The trigger is deterministic (job scheduled for tomorrow), the message is templated (time window, plumber name, what to expect), and the channel is SMS/email. ServiceTitan does this natively. If you're not already using it, you're leaving the easiest hour of your office manager's day on the table. Set it up once, it runs forever. The only edge case: customer requests a reschedule via reply — that routes to a human.

ServiceTitan generates invoices automatically from completed job records. The plumber in the field can close the job, collect payment via the mobile app, and the invoice syncs to QBO without the office manager touching it. The remaining human work: reviewing invoices for jobs with scope changes, handling payment disputes, and managing commercial clients who pay on net-30 terms. That's maybe 20% of the volume. The other 80% is fully automatable today.

Gusto's AutoPilot feature runs payroll automatically on a set schedule with zero manual intervention for weeks where hours are consistent. The human work is reviewing time entries before the cutoff — especially for overtime, PTO, or hours that look wrong. Gusto integrates with ServiceTitan for time tracking. The office manager should be spending 20 minutes reviewing exceptions, not 2 hours running payroll manually.

Collections follow-up is highly structured: the trigger is days-past-due, the message escalates by bucket (friendly reminder at 30, firmer at 60, final notice at 90), and the channel is SMS/email. ServiceTitan has built-in AR automation, and tools like Collectly or even QBO's payment reminders handle this well. The human judgment call: when to pick up the phone and call a commercial client personally, or when to escalate to collections. That's maybe 10% of cases. The other 90% is automated sequences.

Amazon Business has auto-reorder and approval workflow features that handle recurring supply orders well. Parts ordering is trickier — job-specific parts require judgment about spec, availability, and substitution. For consumables (drain cleaner, fittings, office supplies), set up Amazon Business auto-reorder with quantity thresholds. For job-specific parts, AI can generate the purchase list from the job scope, but a human should confirm before ordering. This saves time without creating parts-mismatch problems on job sites.

License renewals with the Texas State Board of Plumbing Examiners (TSBPE) are structured but infrequent and consequential. AI can help by: (1) maintaining a license expiration calendar with automated reminders, (2) pre-filling renewal forms from stored employee data, and (3) flagging CE hour requirements. But the actual submission and verification should have a human in the loop — a missed renewal means a plumber can't legally work. Use AI for tracking and prep; human for submission and confirmation.

Do not automate. On-site plumbing diagnosis requires physical inspection, sensory judgment (sounds, smells, water pressure feel), and years of pattern recognition that AI cannot replicate remotely. Misdiagnosis leads to failed repairs, property damage, and liability. This is the core skilled labor value of a lead plumber.

Do not automate: On-site diagnosis of complex plumbing problems requires physical presence, sensory input, and expert judgment. No current AI tool can replace this. Attempting to automate it creates safety and liability risk.

This directly addresses your pain point about 2-3 day estimate turnaround. AI can dramatically accelerate estimate writing by: pulling up similar past jobs from ServiceTitan, suggesting line items and labor hours based on job type, and generating a formatted proposal. The lead plumber still sets the final numbers — they know the site conditions, the difficulty, and the customer. But going from a blank page to a 90% complete estimate in 5 minutes instead of 45 minutes is achievable today. ServiceTitan's estimate templates and AI-assisted line item suggestions handle this.

Physical skilled trades work. Not automatable by any current or near-future AI. This is the product you sell.

Do not automate: Physical plumbing installation and repair requires licensed skilled labor. No AI tool performs physical work. This is not an automation question.

Apprentice mentoring is a trust and safety relationship. A lead plumber teaching an apprentice to sweat copper or handle a gas line is not a task that benefits from AI intermediation. AI can supplement with reference materials, but the mentoring relationship itself must remain human.

Do not automate: Skilled trades mentorship involves safety-critical instruction, relationship-based learning, and real-time judgment correction. AI cannot supervise an apprentice on a live job site. This is a human responsibility with safety and liability implications.

Austin's Development Services Department permit process is structured but requires local knowledge and occasional human follow-up when applications are flagged. AI can help by: pre-filling permit applications from job data, tracking permit status, and flagging expiration dates. The actual submission and any back-and-forth with the city requires a human. This is a time-assist, not a replacement.

Route optimization is a real AI win here. Tools like ServiceTitan's dispatch board with route optimization, or standalone tools like OptimoRoute, can sequence daily service calls to minimize drive time. For a 3-plumber service team in Austin traffic, this can save 30-45 minutes per plumber per day. The human still drives; AI just sequences the route intelligently.

The diagnosis is human. The quote can be AI-assisted. ServiceTitan's flat-rate pricebook means the service plumber selects the job type and the price populates automatically — no mental math, no inconsistent pricing. The customer approval conversation is human. AI assist here means the plumber spends 2 minutes generating a professional quote on their tablet instead of 15 minutes writing it up, and pricing is consistent across all three service plumbers.

The repair is physical and human. Photo capture is already done via ServiceTitan mobile — the AI assist here is minimal. Some shops use AI to auto-tag and organize photos by job type for warranty and liability documentation, but this is a minor efficiency gain. Not worth prioritizing.

Job summary writing is a strong AI candidate. ServiceTitan's AI can auto-generate job notes from the selected job type, parts used, and any voice notes the plumber dictates. Payment collection via the mobile app is already largely automated. The plumber's job should be: select what was done, dictate any unusual notes, collect payment. AI handles the write-up. This saves 10-15 minutes per job across 3 plumbers running 4-6 jobs each per day.

AI can generate a restock list from the day's job records — what parts were used, what's below par level, what's needed for tomorrow's scheduled jobs. ServiceTitan tracks parts usage per job. The physical restocking is human. The intelligence layer (what to pull and in what quantity) can be AI-generated. This is a 10-minute task that AI can make a 2-minute task.

Physical labor assistance on job sites. Not automatable.

Do not automate: Physical job site assistance requires human presence and cannot be automated. This is core labor.

Physical labor. Not automatable by any current AI.

Do not automate: Physical site preparation and equipment handling requires human labor. Not an automation question.

AI tutoring tools are genuinely useful for exam prep. A tool like Khanmigo, Claude, or a trades-specific study app can generate practice questions, explain concepts, and quiz apprentices on Texas plumbing code. This doesn't save the business money directly, but it accelerates apprentice advancement to journeyman status — which increases their billable rate and reduces your dependency on expensive lead plumbers for every complex job. Worth mentioning as a retention and development tool.