FieldproxyAI-native field service softwarePricing
How do AI credits work in field service software?
AI credits are a usage meter: the vendor gives each AI task a credit cost (a minute of an AI phone call, one text conversation, one page read from a PDF, one agent action) and your plan includes a monthly allowance that resets and usually does not roll over. When the allowance runs out, the software either pauses the AI, keeps going and bills overage at a set rate, or lets you buy a top-up pack, depending on the vendor and the setting you choose. To compare vendors, convert credits to the work you actually do: call minutes, text threads and documents per month, then multiply by each vendor's published rate.
- A credit is a unit, not a price: always ask how many credits each task uses and what one credit costs.
- The big users are AI phone calls (billed per minute or per conversation), text and chat threads, and documents or photos read by AI.
- Allowances reset monthly and usually expire: HubSpot, Microsoft and Salesforce all say unused credits do not roll over.
- Pick the run-out behavior in advance: pause the AI, bill overage, or top up. Jobber, for example, lets you choose pause or pay.
- Credits tied to the company, not to seats, mean adding users or subcontractors does not add AI cost by itself.
This page explains how AI usage is metered and capped. It does not rank vendors, and list rates change often, so check each vendor's live page before you budget.
Facts checked as of November 2026. Sources at the end of the page.
Step by step
How AI credits work, step by step
Most field service and CRM vendors follow the same pattern, even when they call the unit something else (conversations, resolutions, Flex Credits, Copilot Credits).
1. The vendor sets a rate card per task
Each AI task has a fixed credit cost. HubSpot publishes 50 credits per resolved customer agent conversation, 50 credits per minute of voice, 500 per invoice handled by its revenue agent and 10 per data agent response. Salesforce charges 20 Flex Credits per Agentforce action and 30 per voice action.
2. One credit has a dollar value
HubSpot sells extra credits at $10 per 1,000 ($0.01 each). Salesforce sells 100,000 Flex Credits for $500 ($0.005 each), so one action costs $0.10. Microsoft Copilot Studio sells 25,000 Copilot Credits for $200 a month, or $0.01 each pay as you go.
3. Your plan includes a monthly allowance
Plans include a pool of credits per account, not per user. The pool refreshes on the billing date. HubSpot and Microsoft both state that unused monthly credits expire, and Salesforce says unused Flex Credits do not carry into the next subscription term.
4. Each AI task draws down the pool
A single job can use several tasks: an AI receptionist answers the call (minutes), reads the work order PDF (pages), texts the customer a confirmation (a conversation), and summarizes the visit (a response). Microsoft's own example adds 10 credits for grounding and 2 for a generative answer to one prompt, so one question can cost 12 credits.
5. Something happens at zero
Three patterns exist: the AI pauses until the reset, the AI keeps running and the extra is billed in arrears, or you buy a top-up pack. Microsoft blocks agents when a prepaid tenant reaches 125% of capacity unless pay as you go is linked. Salesforce bills usage beyond the entitlement at the contracted rate with no penalty.
6. You watch usage and set limits
Good systems show usage live and let an admin cap it. HubSpot lets you set spend caps at the account, feature and activity level and sends notifications as you approach the limit. Microsoft lets an admin set a monthly limit per agent.
Failure modes
Why AI credits confuse buyers
These are the questions and worries buyers raised on our sales calls, by trade and size.
A number of credits with no unit behind it
A residential electrical contractor and a painting company both asked the same thing: what does a credit actually buy? An allowance of credits means nothing until you know the cost of a call minute, a text thread and a page read, so ask for the rate card before the quote.
No idea what happens at zero
A commercial electrical and low-voltage contractor asked what happens when the credits run out. If the answer is that the AI receptionist stops answering at 6 pm on a Friday, that is a business decision you need to make up front, not discover.
Fear of an open-ended bill
A warranty work-order marketplace with a large contractor network asked whether spending could be capped, and a facility maintenance coordinator asked whether AI is a flat fee or per credit. Usage pricing without a cap is the main reason buyers stall on AI features.
Not knowing what one workflow costs
A nationwide furniture installer working through subcontractors asked how many credits it takes for an agent to read a job, message the technician, text the client and schedule the visit. A grid inspection project asked what a bulk import and PDF extraction would use. Vendors rarely publish this per workflow, so you have to build it from the rate card.
Who uses credits: staff, contractors or customers
A risk inspection firm with contractor field agents asked whether contractors' activity uses credits. Usually it depends on the task, not the person: a contractor filling a form uses none, but a contractor's photo checked by AI or a voice note turned into fields does.
Calls and texts priced differently by country
A cleaning robot manufacturer asked how calls and texts are priced in India. Phone carrier costs per minute and per message differ by country, so a voice-heavy plan priced for the US may not hold elsewhere.
Options
How vendors meter AI today
As each vendor's own pages show it in November 2026, in USD. Rates change often; check the live page.
| Model | Example | Unit and rate | At the limit |
|---|---|---|---|
| Add-on with included conversations | Jobber Receptionist | $29 a month includes 30 conversations, then $0.79 each; unlimited on the Plus plan. A call counts when the AI engages; spam does not count. A text thread is one conversation for up to 24 hours. | You choose: keep going and pay overage, or pause until the next billing cycle |
| Credits with a per-task rate card | HubSpot credits | $0.01 per credit; 50 per resolved chat conversation, 50 per voice minute, 10 per data agent answer | Notifications near the limit; spend caps per account, feature and activity; pay as you go optional |
| Credits per action | Salesforce Agentforce (used with Salesforce Field Service) | $500 per 100,000 Flex Credits; 20 per action ($0.10), 30 per voice action ($0.15); or $2 per conversation | Overage billed monthly at the contracted rate; unused credits do not roll over |
| Credits per feature and per token | Microsoft Copilot Studio (used with Dynamics 365) | $200 for 25,000 credits a month or $0.01 pay as you go; 2 per generative answer, 5 per agent action, 8 per page processed, 35 per minute of generative voice | Agents blocked at 125% of prepaid capacity unless pay as you go is linked |
| AI included, usage not published | ServiceM8, Housecall Pro, ServiceTitan | ServiceM8 lists unlimited AI Helper assists on paid plans (10 a day on Free). Housecall Pro sells CSR AI as an add-on with no published price. ServiceTitan says Atlas is available to all customers and does not publish usage pricing. | Ask in writing what limits apply |
| Credits bundled into one company plan | Fieldproxy | Each plan bundles jobs a month, integrations and AI credits; AI roles run on the credits with no separate agent fee | Top up on any plan; monthly limit set with you |
Vendor facts are linked in Sources. Fieldproxy is one price per company with unlimited users; see /pricing for the plans.
Worked example
A worked example: one month of AI at a 600-job maintenance company
Say you run commercial maintenance and complete 600 jobs a month. You want AI to answer after-hours calls, handle text confirmations and turn emailed PDF work orders into jobs. Last month you had 300 after-hours calls averaging 3 minutes (900 minutes), 400 text threads with customers, of which 300 ended with the question answered, and 200 PDF work orders of 2 pages each (400 pages).
Count the units first, not the credits: 900 call minutes, 300 to 400 text conversations, 400 pages. Every vendor's bill is some multiple of these numbers.
The same month priced on three published meters, at list rates.
| Task | HubSpot credits at $0.01 | Microsoft Copilot Credits at $0.01 | Jobber Receptionist |
|---|---|---|---|
| 900 voice minutes | 50 a minute: 45,000 credits, $450 | 35 a minute (generative voice): 31,500 credits, $315 | Counted per conversation, not per minute |
| Text threads | 300 resolved at 50: 15,000 credits, $150 | Depends on answers per thread: about 2 credits per generative answer | Calls and texts share one count: 700 conversations, $29 for 30 plus 670 at $0.79, about $558 |
| 400 PDF pages | Not a published HubSpot task | 8 a page: 3,200 credits, $32 | Not part of Receptionist |
Illustrative only: the three products do different jobs and are not interchangeable. Rates from the vendors' pages in Sources.
Three things stand out. First, voice is the biggest line on every meter: minutes add up faster than texts or documents. Second, the phone line itself is cheap: Twilio lists $0.0085 a minute to receive a call on a US local number, so 900 minutes is about $7.65; nearly all the cost is the AI layer, not the carrier.
Third, the unit decides who wins. A per-conversation meter favors long calls; a per-minute meter favors short ones. If your after-hours calls average 8 minutes (an emergency HVAC call taking full site details), the per-minute bill nearly triples while the per-conversation bill does not move. Run your own call log through each rate card before you pick.
Finally, set the cap from the same numbers. If 900 minutes is a normal month, a cap at 1,200 minutes with an alert at 1,000 covers a storm week without an open-ended bill.
Buying checklist
What to check before you buy
Good signs
- A written rate card: credits per call minute or conversation, per text thread, per page or photo read, per agent action.
- Credits are pooled per company, so adding users or subcontractors does not change AI cost.
- A monthly cap you set, with alerts before you hit it, and a clear choice of pause or overage.
- Top-ups at a known price that you can buy the same day.
- A usage screen that shows which agent or feature used the credits, by day.
- Spam and abandoned calls are not counted.
Red flags
- An allowance stated only as a number of credits, with no rate card.
- No cap: overage is billed automatically with no limit you control.
- The AI stops answering at zero with no warning and no top-up option.
- Rules-based automations (a status change sends an email) also burn AI credits.
- Reasoning or premium models billed per token with no estimate per task.
- AI agents sold per agent on top of the plan, so each new use case adds a fee.
Token-based billing deserves extra care. Microsoft's documentation shows a single follow-up question on a reasoning model using 141 credits, and the same question starting a new conversation using 461, because long instructions and retrieved knowledge are billed as tokens. Ask any vendor using reasoning models for a per-task estimate.
Fieldproxy
How AI credits work in Fieldproxy
Fieldproxy is one price per company with unlimited users. Each plan bundles jobs a month, integrations and AI credits; see /pricing for the plans.
Credits come with the plan, not with seats
The AI credit allowance belongs to the company's plan. Office staff, technicians and subcontractors are all users at no extra cost, so onboarding more people does not change the AI budget; only the AI work does.
AI roles run on the credits, with no agent fee
AI roles such as an after-hours receptionist on the Calling agent (/agents/calling-agent), the Text agent (/agents/text-agent) answering SMS and email, and a dispatcher on the Dispatch agent (/agents/dispatch-agent) run on the plan's AI credits. There is no separate fee per agent.
AI features are in every plan
Command Center (/agents/command-center), summaries, drafting and document reading, such as the Document intelligence agent (/agents/document-intelligence-agent) turning PDF work orders into jobs, are in every plan. Automations are unlimited.
A monthly limit set with you
AI usage runs under a monthly limit agreed when the plan is set up (e.g. a cap sized to last month's call minutes plus a margin), so there is no open-ended AI bill.
Top up on any plan
If a busy month needs more, you top up credits on the plan you have instead of changing plans or waiting for the reset.
You approve what the AI does
Command Center shows a confirm card and an Undo on every change it makes, and Autopilot (/agents/autopilot) drafts follow-ups as cards a person approves before anything is sent.
Regional notes
Currency and country differences
Some vendors publish credit prices in local currency. Salesforce lists 100,000 Flex Credits at $500, €500, AU$700 or £400. Others publish USD only, so check whether your invoice is converted at a set rate or at the card rate.
Voice and SMS carrier costs differ by country. Twilio publishes a separate price page per country, so an AI receptionist on a UK, Australian or Indian number can cost a different amount per minute than a US one. Ask whether the vendor's credit rate for calls is the same in every country or varies with the carrier rate.
How switching works
How switching to Fieldproxy works
We move your data over and set the system up to the way you already work. Fieldproxy replaces the old tool; it does not run alongside it.
Send a month of calls, texts and work orders
A call log with durations, a count of customer text threads, and the number of emailed or PDF work orders. That is enough to size AI usage.We size the credits on your real volume
We map your volume to the AI roles you want (receptionist, text replies, document reading) and show the monthly credit use before you commit.We set the limit and the run-out behavior
We agree a monthly limit with you and set the AI roles up on your stages, forms and statuses.Go live and check the first month
After the first month we compare real usage with the estimate and adjust the limit or top up if needed.
FAQ
Questions people ask
- What uses AI credits in field service software?
- Usually AI phone calls (per minute or per conversation), AI text and chat replies, documents and photos read by AI, and actions an AI agent takes on its own. Normal app use, such as a technician filling a form or a dispatcher dragging a job, uses none.
- What happens when I run out of AI credits?
- It depends on the vendor and your setting: the AI pauses until the reset, overage is billed at a set rate, or you buy a top-up. Jobber lets you choose pause or pay; Microsoft blocks agents at 125% of prepaid capacity unless pay as you go is linked. In Fieldproxy you top up on any plan.
- Can I cap how much I spend on AI?
- With most credit systems, yes. HubSpot lets you set spend caps by account, feature and activity, and Microsoft lets an admin set a monthly limit per agent. Ask any vendor that cannot show you a cap setting how overage is controlled.
- Is AI a flat fee or charged per credit?
- Both exist. Some vendors sell an AI add-on for a flat monthly fee with included conversations (Jobber Receptionist), some meter every task in credits (HubSpot, Salesforce, Microsoft), and some bundle credits into the main plan (Fieldproxy).
- Do subcontractors or technicians use AI credits?
- Only when their work calls the AI, for example a photo checked by AI or a voice note turned into job fields. Logging in, filling a form or closing a job does not. Where credits are pooled per company, the number of users does not change the cost.
- How many credits does one job use?
- Add up the tasks: a 3-minute AI call, one text confirmation, a 2-page PDF read and a summary. On HubSpot's published rates the call alone is 150 credits; on Microsoft's, 105 for generative voice plus 16 for the pages. Ask the vendor for the same breakdown.
- Do unused AI credits roll over?
- Usually not. HubSpot and Microsoft both state that unused monthly credits expire, and Salesforce says unused Flex Credits do not carry into the next term. Size the allowance to a normal month and top up for peaks.
Sources
Competitor details on this page come from the vendors' own pages and documentation, checked as of November 2026. Plans and prices change, so check the vendor's current page before you decide.
- Jobber Help Center: Receptionist powered by Jobber AI: https://help.getjobber.com/en/articles/receptionistpowered-by-jobber-ai/
- HubSpot Credits: https://www.hubspot.com/products/artificial-intelligence/credits
- HubSpot: customer agent priced per resolved conversation: https://www.hubspot.com/company-news/hubspots-customer-agent-and-prospecting-agent-now-you-pay-when-the-task-is-complete
- Salesforce Agentforce pricing: https://www.salesforce.com/agentforce/pricing/
- Microsoft Copilot Studio licensing guidance: https://www.microsoft.com/licensing/guidance/Microsoft-Copilot-Studio
- Microsoft Learn: Copilot Studio billing rates and management: https://learn.microsoft.com/en-us/microsoft-copilot-studio/requirements-messages-management
- ServiceM8 pricing: https://www.servicem8.com/pricing
- Housecall Pro pricing: https://www.housecallpro.com/pricing/
- ServiceTitan Help: Atlas: https://help.servicetitan.com/docs/atlas.md
- Twilio voice pricing, United States: https://www.twilio.com/en-us/voice/pricing/us
Related
Keep reading
- Per-user vs per-job field service pricing
- Turn PDF work orders into jobs with AI
- FSM software pricing compared
- AI receptionist for contractors
- What is the best field service software for a 5-10 person HVAC team?
- Cheaper ServiceTitan alternatives for small plumbing and HVAC companies, without payment processor lock-in