OpenAI Pricing for Home Service Businesses: What 1,000 Leads a Month Actually Costs

GPT-4o costs $2.50 per million input tokens and $10 per million output tokens, per OpenAI's published 2026 API pricing. GPT-5-nano, the smallest production model, runs $0.05 input and $0.40 output per million tokens.

Those numbers mean nothing to a contractor. So here is what they actually translate into for a real home services shop running 1,000 leads a month through AI. Plus what the real bill looks like once you add up the parts OpenAI does not sell you.

The OpenAI price list that matters for contractors

Per OpenAI's 2026 API pricing and industry references compiled by Finout, PricePerToken, and CloudZero, here are the models a home services build actually uses:

GPT-4o (the workhorse): $2.50 / 1M input, $10.00 / 1M output, $1.25 / 1M cached input.

GPT-4o-mini: roughly $0.15 / 1M input, $0.60 / 1M output. Cheap enough to use liberally on classification and simple extraction.

GPT-5-nano: ~$0.05 / 1M input, ~$0.40 / 1M output. Smallest and cheapest production model, per industry references.

text-embedding-3-small: $0.02 / 1M tokens. This is how you index your price book and docs for retrieval.

Realtime API (voice): ~$0.06 per minute of audio input, ~$0.24 per minute of audio output on the gpt-realtime model.

Batch API: 50% off everything, as long as you can wait up to 24 hours.

Token math. 1,000 tokens is roughly 750 words of English. A typical customer chat turn is 200 to 400 tokens each way.

The 1,000-leads-a-month example

Take a mid-sized HVAC or plumbing shop. 1,000 inbound leads a month across calls, web forms, Google LSA, Facebook, and returning customers.

Every lead triggers the following AI work:

Inbound classification (is this a new lead, a returning customer, spam, or a vendor?). ~500 input tokens, 100 output tokens on GPT-4o-mini.
Missed-call text-back and qualification chat. ~3,000 tokens total per conversation on GPT-4o-mini.
Quote follow-up drafting on quiet leads. ~1,500 tokens per follow-up on GPT-4o.
Morning brief summary across the owner's day. ~4,000 tokens on GPT-4o-mini, once per day per user.

Adding up the OpenAI API bill

Classification: 1,000 leads × (500 input + 100 output) = 500K input, 100K output. Cost on GPT-4o-mini: ~$0.08 + $0.06 = $0.14.

Qualification chat: Assume 40% of leads engage. 400 conversations × 3,000 tokens ≈ 1.2M tokens, split 70/30 input/output. Cost on GPT-4o-mini: ~$0.13 + $0.22 = $0.35.

Quote follow-up: Assume 200 quote follow-ups a month. 200 × 1,500 tokens ≈ 300K tokens. Cost on GPT-4o: ~$0.38 + $1.50 = $1.88.

Morning brief: One owner, 30 briefs a month × 4,000 tokens. Cost on GPT-4o-mini: ~$0.02.

Embeddings on your docs/price book, one-time: a full price book and service FAQ is maybe 50K tokens. At $0.02 / 1M that is $0.001. Not a typo.

Total OpenAI API spend on a realistic 1,000-lead shop: under $3 a month.

Even if you multiply every estimate by 10 to account for retries, tool-calling overhead, context re-sends, and more complex agents, you are still under $30 a month in raw API fees.

Why this looks fake

Every contractor reads that and assumes the math is wrong. It is not wrong. It just is not the full bill.

OpenAI's API is genuinely cheap. That is not what costs money in an AI build. The real costs are everything around it.

What the real bill looks like

Developer time: Building a production agent stack (missed-call text-back, qualification, CRM sync, compliance, monitoring) runs $25K to $75K with a competent freelancer or contractor. That is a real number multiple operators have shared on the Owned and Operated podcast hosted by John Wilson.

Maintenance: OpenAI deprecates models on a 12 to 18 month cycle. Your CRM API changes. Google changes SMS rules. Plan on 5 to 10 hours of dev time per month ongoing, or $500 to $1,500 a month.

Phone and SMS infra: Twilio messaging runs $0.0079 to $0.02 per SMS depending on volume and carrier fees. A number costs a dollar a month. At 10 messages per lead × 1,000 leads, that is 10,000 SMS ≈ $80 to $200 a month.

Hosting and logging: Vercel or AWS for the app, plus a logging stack like LogTail or Datadog. $50 to $300 a month at small-shop scale.

Human fallback: You still need a person to handle what the bot cannot. That is an office manager seat regardless. No AI lets you zero that out at $1M to $10M revenue.

All-in for a DIY build: month one is $30K-$75K. Month two through twelve is $700 to $2,000 a month plus the occasional dev fire.

The LocaliQ number that puts this in perspective

LocaliQ's 2025 home services ad benchmarks put the average cost per lead at $90.92. Roofing leads run $228. Doors and windows run $200. HVAC, plumbing, and electrical sit in the $80 to $150 range.

Now flip the math. You are paying $90 to acquire a lead. Under 5% of you respond within the critical 5-minute window, per a Valve+Meter industry study. Instant Sales Funnels cites an average HVAC callback delay of 4.2 hours, by which time 67% of those leads have already booked a competitor.

You are paying $90 a lead to miss two-thirds of them. Spending even $1 or $2 per lead in AI to actually respond is a rounding error. The question is never "can I afford AI." The question is "what builds or buys the infrastructure that uses the AI."

The ChatGPT Plus and Business seat math

Separate from API usage, OpenAI sells ChatGPT subscriptions. Plus is $20 per month per user. Business is $25 per seat per month.

Put your owner and your office manager on Plus. That is $40 a month for writing, scripts, training, ad copy, and drafting. Every shop should do this.

If you have an ops manager, salesperson, and marketing lead, go Business. Five seats at $25 is $125 a month for the whole team.

That is not an AI strategy. That is your writing tool. It does not connect to your CRM, answer your phones, or text back missed calls.

Pre-built AI for home services: the pricing alternative

Vertical AI products wrap OpenAI and add everything OpenAI does not sell: the CRM integrations, the compliance, the vertical-specific prompts, the guardrails, and the working UI.

Hatch, CallRail Voice Assist, Podium, and similar contractor-focused AI tools typically run $300 to $1,500 a month per location depending on tier. They replace the $30K-$75K DIY build.

Sully sits in this tier. It is built specifically for $1M to $10M home service contractors and includes missed-call follow-up, lead qualification, quote follow-up, AI chat trained on your company data, and a morning brief. It is pre-wired into Jobber, Housecall Pro, ServiceTitan, Workiz, GoHighLevel, Gmail, Google Calendar, Slack, QuickBooks, and HubSpot.

The underlying OpenAI costs are pennies. The rest of the product is the work. That is what you are paying for.

The honest cost comparison

Option	Month 1	Monthly ongoing	Breaks if you skip
ChatGPT Plus (2 seats)	$40	$40	Nothing, it is a writing tool
DIY on OpenAI API	$30K-$75K	$700-$2,000 + dev fires	Compliance, maintenance, monitoring
Vertical AI platform (Sully, Hatch, etc.)	$300-$1,500	Same	Nothing, vendor owns the stack

For a shop doing $1M to $10M and running 1,000 leads a month, the math is almost always "buy the vertical tool." The OpenAI API alone is cheap. Building a production system on top of it is not.

The one-line rule

AI is cheap. The infrastructure around AI is not. Pay for the infrastructure, not for the tokens.

If you are reading this and sizing up a DIY build against the $30 a month API bill, you are measuring the wrong thing. Go look at your missed-call rate, your speed-to-lead, and your quote follow-through. Those are the numbers that matter. The AI is the easy part.

Sources