By Md Abdur Rouf Mia | Updated 16 October 2026 | Prices verified 16 October 2026
The AI voice agent cost per minute runs from about $0.07 to $0.31 on the platforms’ own published pricing, and most small-business setups land between $0.08 and $0.14. At three-minute calls, that is $240 to $420 per 1,000 calls. At five-minute calls, the same 1,000 calls cost $400 to $700.
The range is wide because a voice agent is not one product. It is five layers: telephony, speech-to-text, a language model, text-to-speech, and the orchestration that ties them together. Some vendors bill each layer separately, one bundles all of them, and the headline number on each pricing page covers a different slice. Vapi’s $0.05 a minute and Bland’s $0.14 a minute are both real prices, and they cannot be compared until you add the layers Vapi leaves out.
This guide reads the pricing pages of Vapi, Retell AI, ElevenLabs Agents and Bland AI, converts each into an AI voice agent cost per minute, then runs the result at 100, 1,000, and 10,000 calls. It is a cost model built from published prices, not a test of how well any platform answers a phone. The method and its limits are near the end.
The Short Answer: AI Voice Agent Cost Per Minute at 100, 1,000 and 10,000 Calls
Three bands cover the AI voice agent cost per minute across the four platforms: lean at about $0.08 a minute, typical at $0.11 and premium at $0.14. Later sections show which setup on which platform lands in each band.
Three-minute calls
| Volume | Total minutes | Lean ($0.08/min) | Typical ($0.11/min) | Premium ($0.14/min) |
|---|---|---|---|---|
| 100 calls | 300 | $24 | $33 | $42 |
| 1,000 calls | 3,000 | $240 | $330 | $420 |
| 10,000 calls | 30,000 | $2,400 | $3,300 | $4,200 |
Five-minute calls
| Volume | Total minutes | Lean ($0.08/min) | Typical ($0.11/min) | Premium ($0.14/min) |
|---|---|---|---|---|
| 100 calls | 500 | $40 | $55 | $70 |
| 1,000 calls | 5,000 | $400 | $550 | $700 |
| 10,000 calls | 50,000 | $4,000 | $5,500 | $7,000 |
The only difference between the two tables is call length, and it scales the bill by 1.67 times. If you have to guess one number when budgeting, guess your average call length, and measure it as soon as you have real calls, because it is the assumption most likely to be wrong.
A typical $0.11 per minute means a three-minute call costs about $0.33 and a five-minute call about $0.55. Those are the per-call numbers to compare against whatever a missed call costs you today.
What Each Platform Publishes About AI Voice Agent Cost Per Minute
| Platform | What the headline rate covers | Published rate | Fixed fees and limits |
|---|---|---|---|
| Vapi | Hosting and orchestration only. Model, transcriber, and voice pass-through at provider cost | $0.05 per minute | No monthly fee on a usage-only account. Optional Core ($29/month) and Pro (10% of hosting, $999 minimum). 4 concurrent calls on usage-only, $10 per extra line a month |
| Retell AI | Every component priced separately on one page | $0.07 to $0.31 per minute all-in | No platform fee. 20 concurrent calls free, then $8 per line a month |
| ElevenLabs Agents | ElevenLabs hosting. Model tokens and telephony are billed separately | $0.08 per minute on every plan | Plans prepay minutes, from 15 (Free) to 12,375 (Business, $990). Burst above your concurrency limit costs $0.16 per minute |
| Bland AI | One bundled rate that includes the model, speech and telephony | $0.14 / $0.12 / $0.11 per minute | Start $0, Build $299/month, Scale $499/month. Daily call caps of 100 / 2,000 / 5,000 |
Read the second column first, because it decides what the AI voice agent cost per minute actually includes. Only Bland states one rate that already includes the model, the speech and telephony. The other three publish a platform rate and leave some of the model, voice or phone line as separate charges, so their headline numbers will always look smaller than Bland’s.
Why the Headline AI Voice Agent Cost Per Minute Misleads
Four things make a headline rate unreliable as a budget.
It covers a different slice of the stack. Vapi’s $0.05 a minute pays for hosting and orchestration. Its own pricing calculator shows what surrounds that fee: for 1,000 minutes it adds a transcriber, a language model, and a voice, and the estimated total is about $82 to $129 for the month, or $0.082 to $0.129 a minute. Bland’s $0.14 already includes those layers. Comparing $0.05 with $0.14 compares two different products.
The units differ. Per minute, per message, per plan-included minute, per concurrent line. ElevenLabs sells plans that prepay minutes. Retell bills components per minute, to the second. Bland adds a monthly fee on the plans with the lower per-minute rate. Vapi can add a monthly success package. Turning all of that into one AI voice agent cost per minute is arithmetic the pricing pages leave to you.
Fixed fees and caps change the answer at the edges. Bland’s $0.11 rate costs $499 a month before the first call, and its free plan stops at 100 calls a day. ElevenLabs charges $0.16 a minute for burst calls above your concurrency limit, double its standard rate.
The cost per minute is not constant across call lengths. When the model is billed per token, conversation history is resent on every turn, so a longer call costs more per minute than a short one. The worked example below puts a number on it.
The Five Layers Behind the AI Voice Agent Cost Per Minute
Every voice agent, whatever the vendor calls it, is some version of these five layers. The ranges below come from the four pricing pages.
| Layer | Range in published pricing | Where it comes from |
|---|---|---|
| Telephony | $0.0055 to $0.014 per minute; free on Vapi’s own telephony | Vapi pricing page: Telnyx $0.0055, Twilio $0.008 inbound and $0.014 outbound |
| Speech-to-text | About $0.0095 to $0.0099 per minute | Deepgram, as estimated in Vapi’s calculator |
| Language model | $0.0016 to $0.16 per minute on the standard tier, up to $0.32 on the fast tier | Retell pricing page, GPT 5 nano to GPT 5.5. Vapi’s calculator shows $0.0077 to $0.0452 for an OpenAI model |
| Text-to-speech | $0.015 to $0.10 per minute | Retell pricing page. Vapi’s calculator shows $0.0146 to $0.0238 for an ElevenLabs voice |
| Orchestration | $0.05 (Vapi) to $0.055 (Retell); $0.08 for ElevenLabs hosting (the page lists only model tokens and telephony as extra) | Each vendor’s pricing page |
Two observations matter more than any single number.
The voice and the model are the swing layers. Inside Retell’s own table the voice runs from $0.015 to $0.10 a minute and the model from $0.0016 to $0.16 on the standard tier. Those two choices move the AI voice agent cost per minute far more than the platform does.
The platform fee is close to a constant. Vapi charges $0.05 and Retell charges $0.055 for the infrastructure layer. If you are choosing between platforms to save money, you are mostly choosing between voices and models, not between platform fees.
A self-hosted orchestration layer, using an open-source framework such as Pipecat or LiveKit, would remove the platform fee. I have not priced it here — see the self-hosted cost analysis for that stack on a $5-VPS.
Where the Money Goes: A Worked LLM Example
Most guides say the language model is cheap. That is true per token and incomplete per call, because a voice agent resends its full context on every turn.
The assumptions, so you can change them: a 1,000-token system prompt (instructions, persona, tool definitions), 40 tokens for each caller turn and 100 tokens for each agent reply. A three-minute call is six exchanges and a five-minute call is ten. The rates are illustrative, not any vendor’s quote: $2 per million input tokens, $10 per million output tokens, and $0.10 per million for cached input.
| Three-minute call | Five-minute call | |
|---|---|---|
| Exchanges (caller turn plus agent reply) | 6 | 10 |
| Input tokens sent over the whole call | 8,340 | 16,700 |
| Output tokens | 600 | 1,000 |
| Input cost ($2 per million) | $0.0167 | $0.0334 |
| Output cost ($10 per million) | $0.0060 | $0.0100 |
| Total per call | $0.0227 | $0.0434 |
| Cost per minute | $0.0076 | $0.0087 |
The last row is the point. The model costs about 15% more per minute on the five-minute call. The history grows faster than the call does, so the same agent costs more per minute the longer the conversation runs. Every other layer is linear. This one is not.
Now turn on prompt caching. The system prompt does not change between turns, so it can be cached:
| 3-minute, no caching | 3-minute, cached | 5-minute, no caching | 5-minute, cached | |
|---|---|---|---|---|
| Cached input tokens (at $0.10 per million) | 0 | 6,000 | 0 | 10,000 |
| Uncached input tokens (at $2 per million) | 8,340 | 2,340 | 16,700 | 6,700 |
| Total per call | $0.0227 | $0.0113 | $0.0434 | $0.0244 |
| Per minute | $0.0076 | $0.0038 | $0.0087 | $0.0049 |
Caching cuts the model line by about half on the three-minute call and 44% on the five-minute call. It is usually a configuration setting rather than an engineering project, but providers set a minimum prompt length for caching, often around 1,024 tokens, so a short prompt may not qualify. Check your provider.
Two caveats. First, this matters on platforms that pass model usage through at token prices, such as Vapi and ElevenLabs. Retell publishes a fixed per-minute price for each model, so on Retell the context growth is priced into the rate. Second, at this prompt size, the model is a small part of the minute, about $0.0076 against a $0.05 platform fee.
A long prompt, large tool definitions or a premium model change that quickly: Vapi’s calculator shows an OpenAI model anywhere from $0.0077 to $0.0452 a minute.
Vapi
Vapi bills $0.05 a minute for hosting and orchestration, prepaid, and passes the model, transcriber, and voice through at the provider’s price with no markup. A usage-only account includes $5 of free credit (100 minutes of hosting), four concurrent calls, and one phone number.
Optional success packages add limits and support. Core is $29 a month for 10 concurrent calls and five phone numbers. Pro is 10% of the hosting fee with a $999 monthly minimum, which means it only costs more than the minimum above 199,800 minutes a month. Extra concurrency is $10 per line a month, and HIPAA-eligible data handling is $2,000 a month.
Telephony depends on what you connect. Vapi’s own telephony and SIP are listed as free. If you bring Twilio, the page lists $0.008 a minute inbound and $0.014 outbound, with Telnyx at $0.0055 and Vonage at $0.00814.
What it costs per minute. The AI voice agent cost per minute on Vapi, using the component ranges in its calculator, is between $0.082 and $0.129. The midpoint is about $0.105: $0.05 hosting, about $0.0097 for Deepgram transcription, $0.0077 to $0.0452 for an OpenAI model, $0.0146 to $0.0238 for an ElevenLabs voice and $0 for Vapi telephony. At 3,000 minutes, which is 1,000 three-minute calls, that midpoint is about $316.
Who it fits. Builders who want to pick every layer themselves. The trade-off is that you integrate and monitor several vendors, and the part of the bill that varies is the part you choose.
Retell AI
Retell publishes every component on one page, which makes its AI voice agent cost per minute the easiest of the four to price precisely. There is no platform fee. You add the voice infrastructure ($0.055 a minute), a voice ($0.015 a minute for Retell’s own and several third-party voices, up to $0.10 for ElevenLabs v3 or v4) and a model priced per minute, from $0.0016 for the smallest to $0.16 for GPT 5.5 on the standard tier. The fast tier costs more, up to double for some models.
| Retell setup | Voice infrastructure | Voice | Model | All-in per minute |
|---|---|---|---|---|
| Lean | $0.055 | $0.015 (Retell platform voice) | $0.008 (GPT 5 mini) | $0.078 |
| Default estimate | $0.055 | $0.015 (Retell platform voice) | $0.040 (GPT 5) | $0.110 |
| Premium | $0.055 | $0.100 (ElevenLabs v3) | $0.160 (GPT 5.5, standard tier) | $0.315 |
The default row matches the $0.11 a minute that Retell’s own estimator shows for a default US setup. The premium row reaches the top of the $0.07 to $0.31 headline range. That top end is the price of a premium voice plus a premium model, not a general rate: swapping the $0.10 voice for a $0.015 one removes $0.085 of the $0.315, about 27%.
Other published terms: a $10 free credit, 20 concurrent calls included and $8 per extra concurrent call a month, billing to the second, no charge for calls that fail to connect, and for voicemail only the time the agent is active on the line.
Add-ons are priced per minute: knowledge base $0.005, advanced denoising $0.005, safety guardrails $0.005, personal data removal $0.01, and AI quality assurance at $0.10 after the first 100 minutes. Branded caller ID is $0.10 per outbound call.
Retell’s estimator shows no telephony charge for the default US setup, and other countries are priced per minute through a country selector, so check yours.
Who it fits. Teams that want every price visible before they build, and that prefer a fixed per-minute model price to token-level billing.
ElevenLabs Agents
ElevenLabs charges $0.08 a minute for call time on every plan. Plans differ in how many minutes are included and how many calls can run at once: Free has 15 minutes and 4 concurrent calls, Starter 75 and 6, Creator 275 and 10, Pro 1,238 and 20, Scale 3,738 and 30, and Business 12,375 and 40. The model is billed separately by usage, and telephony is billed at cost.
Because each plan’s price works out to $0.08 per included minute (Starter is $6 for 75 minutes and Business is $990 for 12,375), a bigger plan does not lower the per-minute rate. It buys concurrency, seats, and features. One reviewer reports that yearly billing lowers the effective rate on included minutes to about $0.067. Burst minutes above your concurrency limit cost $0.16, and text messages cost $0.003 each.
What it costs per minute. The model line is the variable part of the ElevenLabs AI voice agent cost per minute. One review that read ElevenLabs’ calculator on 19 September 2026 reports about $0.0005 a minute for a very small model, $0.0012 for Gemini 2.5 Flash, and $0.0446 for GPT-5.5. Adding Twilio’s inbound rate of $0.008 gives $0.0892 per minute for a lean setup. Adding the $0.0446 model and Twilio’s outbound rate of $0.014 gives $0.1386 for a premium-model outbound setup. On 1,000 three-minute calls, that is about $268 to $416.
A note on why ElevenLabs numbers vary online. The agent product is priced per minute. The separate text-to-speech product is priced in credits per character. Articles that mix the two quote figures that cannot be compared with the $0.08 above.
Who it fits. Teams that care most about the voice and want one vendor for speech and agent hosting, and that can estimate their monthly minutes well enough to pick a plan.
Bland AI
Bland states one bundled AI voice agent cost per minute for each plan. Start is free at $0.14 a minute, Build is $299 a month at $0.12, and Scale is $499 a month at $0.11, with Enterprise on custom pricing. Bland says the rate includes the model, speech-to-text, text-to-speech and telephony, and bills to the second. Transfer time on Bland’s numbers is $0.05, $0.04 and $0.03 a minute on the three plans. The rates took effect on 5 December 2025, replacing a flat $0.09.
The caps are the part to read. Start allows 100 calls a day and 10 concurrent calls; Build 2,000 and 50, and Scale 5,000 and 100. Start’s cap alone is about 3,000 calls a month.
The fixed fees create break-even points you can calculate: Build beats Start above 14,950 minutes a month ($299 divided by the $0.02 saving), Scale beats Start above about 16,630 minutes ($499 divided by $0.03), and Scale beats Build above 20,000 minutes ($200 divided by $0.01). Two independent reviews also report a $0.015 charge for outbound attempts that fail or last under 10 seconds.
Who it fits. Teams that want one predictable number and run outbound or inbound volume high enough to justify a plan. The trade-off is less control over which model and voice you use.
The AI Voice Agent Cost Per Minute at Three Volumes
The tables use the setups described above, with every component taken from a pricing page. Three-minute calls first, then five-minute calls.
Three-minute calls
| Setup (per minute) | 100 calls | 1,000 calls | 10,000 calls |
|---|---|---|---|
| Retell, lean ($0.078) | $23 | $234 | $2,340 |
| Vapi, calculator midpoint ($0.105) | $32 | $316 | $3,161 |
| ElevenLabs, lean model ($0.089) | $27 | $268 | $2,676 |
| Retell, default estimate ($0.110) | $33 | $330 | $3,300 |
| ElevenLabs, premium model ($0.139) | $42 | $416 | $4,158 |
| Bland, cheapest eligible plan ($0.14 / $0.11 + fee) | $42 | $420 | $3,799 |
| Retell, premium voice and model ($0.315) | $95 | $945 | $9,450 |
Five-minute calls
| Setup (per minute) | 100 calls | 1,000 calls | 10,000 calls |
|---|---|---|---|
| Retell, lean ($0.078) | $39 | $390 | $3,900 |
| Vapi, calculator midpoint ($0.105) | $53 | $527 | $5,268 |
| ElevenLabs, lean model ($0.089) | $45 | $446 | $4,460 |
| Retell, default estimate ($0.110) | $55 | $550 | $5,500 |
| ElevenLabs, premium model ($0.139) | $69 | $693 | $6,930 |
| Bland, cheapest eligible plan ($0.14 / $0.11 + fee) | $70 | $700 | $5,999 |
| Retell, premium voice and model ($0.315) | $158 | $1,575 | $15,750 |
Bland’s row uses the cheapest plan that can run the volume. At 100 and 1,000 calls, that is Start. At 10,000 calls a month, about 333 a day against Start’s cap of 100, it has to be Scale ($499 plus $0.11 a minute), which comes to $3,799 at three minutes versus $3,899 on Build.
Three things stand out. The cheapest and the typical setups sit within about $0.03 of each other, so at 1,000 calls the difference between them is under $100 a month. The premium Retell row is the outlier, at nearly three times the typical bill. And call length matters more than the platform: moving from three to five minutes adds more to the bill than switching between most of the setups above.
The Hidden Costs Behind the AI Voice Agent Cost Per Minute
None of these is in the headline rate, and each can change the AI voice agent cost per minute you actually pay.
Concurrency and burst charges. Every platform limits simultaneous calls. Vapi charges $10 a line a month, Retell $8, ElevenLabs charges double the minute rate for burst calls, and Bland ties limits to the plan. Find the ceiling before an outbound campaign finds it for you.
You pay for silence. Retell states that billing covers the full duration of a call, including silence, because speech recognition keeps listening. Assume the same elsewhere unless a vendor says otherwise.
Add-ons. Retell’s quality assurance is $0.10 a minute and personal data removal is $0.01. Vapi lists HIPAA-eligible handling at $2,000 a month. These matter if you are in a regulated field.
Phone numbers and country rates. Retell charges $2 a month per number and $10 a month for a verified number. International telephony is priced per country.
Plan minimums. Bland’s $299 and $499 fees, Vapi Pro’s $999 minimum and ElevenLabs’ prepaid minutes all arrive before the first call.
Transfers. Bland bills transfer time on its own numbers at $0.03 to $0.05 a minute. On Retell the agent fee stops after a transfer, but the telephony charge continues.
Your own time. No rate includes the hours spent writing the prompt, testing calls and reviewing transcripts. For a small team that is often the largest cost, and it never appears on an invoice.
Build vs Buy: When a Per-Minute Agent Is the Wrong Purchase
A per-minute voice agent is the right purchase when you want to control the conversation, the voice and the logic. It is the wrong one when what you actually want is someone else to answer the phone.
In my AI answering service cost comparison, the four AI services cost $129 to $800 a month at 500 three-minute calls. The same 500 calls on a voice agent platform are 1,500 minutes, which is $120 to $210 a month at $0.08 to $0.14 a minute, before your time. The raw platform bill is in the same range as the cheapest managed services. The difference is who builds, tunes and supports the agent.
| Buy the outcome (answering service) | Build the stack (voice agent platform) | |
|---|---|---|
| Pricing unit | Per call, per minute or monthly plan | Per minute, plus model, voice and phone line on some platforms |
| What you own | Nothing; it is a managed service | The prompt, the voice choice, the logic and the escalation rules |
| Your time | Little to none | Prompt writing, testing and monitoring |
| Breaks when | You need custom logic | A component vendor changes its price or limits |
| Better fit | Low, steady volumes | Higher volumes with predictable call shapes |
If you are below a few hundred calls a month and do not want to own a prompt, start with the answering service comparison. There is also a third model, paying per outcome instead of per minute. It sounds cleaner, and it has traps, covered in the outcome-based AI agent pricing guide.
How This Cost Model Was Built
This estimate of AI voice agent cost per minute is a cost model built from the vendors’ published prices, not an invoice from a real deployment. Every figure was read on 16 October 2026 and will change.
What I used. Vapi’s pricing page and calculator, Retell’s pricing page and estimator, ElevenLabs’ agents pricing page, and Bland’s pricing page and billing documentation. Where a vendor’s page did not show a figure, such as ElevenLabs’ per-model language model costs and Bland’s daily call caps, I used a named independent review and say so in the text.
Assumptions. Three-minute calls unless stated, with five minutes as the sensitivity case. US telephony, USD, before tax. The Vapi midpoint is the average of the ends of each range in its calculator. The language model example uses illustrative token rates, not a vendor quote. Figures are rounded.
Modelled, not measured. I have not run a production voice agent on any of these platforms, and nothing here describes call quality, latency or reliability. The editorial policy explains how this site handles that distinction. None of the links in this article are affiliate links.
What I Left Out
Enterprise contracts. At high monthly volumes the public pages stop being useful. Get quotes.
White-label and agency pricing. Several platforms offer reseller tiers whose rates are not public.
Other platforms. Synthflow and others are not covered. This article prices four.
Speech-to-speech models. These skip the separate speech and voice layers. Retell lists them at a flat $0.07 a minute for the mini models and $0.345 to $0.38 for the full-size ones, and the stack arithmetic above does not apply to them.
Quality per dollar. A $0.015 voice and a $0.10 voice are not the same product, and cheaper is not better if callers hang up. This article prices the AI voice agent cost per minute and the layers behind it. It does not rank voice quality.
Frequently Asked Questions
How much does an AI voice agent cost per minute?
The AI voice agent cost per minute runs from about $0.07 to $0.31 on published pricing, with most small-business setups between $0.08 and $0.14. The spread is driven mainly by the voice and the language model you choose, and by whether the platform bundles the layers or bills them separately.
Which platform has the lowest AI voice agent cost per minute?
None is cheapest in every setup. Retell’s published floor is $0.07 a minute, Vapi’s calculator starts around $0.082, and a lean ElevenLabs setup is about $0.089. Vapi’s $0.05 headline is only the platform layer. Bland’s $0.14 is the highest bundled rate but includes everything, and its Scale plan drops to $0.11 plus $499 a month.
Does call length change the AI voice agent cost per minute?
For token-billed models, yes. In the worked example above, a five-minute call cost about 15% more per minute than a three-minute call on the same model, because the conversation history is resent each turn. Retell and Bland publish flat per-minute rates, so the call length changes your total but not the per-minute figure.
Do I pay for silence and hold time?
On Retell, yes: billing covers the whole call, including silence. Assume the same elsewhere unless the vendor says otherwise, and check how voicemail and transfers are billed.
Why do ElevenLabs prices differ across articles?
Because they describe different products. The agent platform is $0.08 a minute plus the model and telephony. The text-to-speech product is priced in credits per character. Plan prices also work out to the same $0.08 per included minute, and yearly billing can lower that.
How do I lower my AI voice agent cost per minute?
Choose the voice and the model first, because they move the cost most. Then turn on prompt caching if your platform bills model tokens, keep the system prompt short, and avoid paying for concurrency you do not use.
Is a voice agent cheaper than an AI answering service?
At 500 three-minute calls a month the platform bill, $120 to $210, is similar to the cheapest answering services, $129 to $800 in my comparison. The voice agent costs more of your time. Choose on who you want to maintain it, not on the headline rate.
Final Thoughts
The AI voice agent cost per minute is not a single number, and the pages that quote one are quoting a slice. Add the layers and most setups land between $0.08 and $0.14, with the voice and the model doing most of the work. Add your call length and you have a budget: about $330 per 1,000 three-minute calls at the typical rate.
The practical order for controlling your AI voice agent cost per minute is to decide your call length, choose the voice and model you can live with, then pick the platform whose fixed fees and limits fit your volume. Bland’s caps, ElevenLabs’ prepaid plans and Vapi’s concurrency lines matter more at the edges than the headline rate does.
Sources
- Vapi pricing: $0.05 per minute hosting, success packages, add-ons, carrier rates and the calculator. Read 16 October 2026.
- Retell AI pricing: voice infrastructure, voice, model and add-on rates, estimator, concurrency and billing FAQ. Read 16 October 2026.
- ElevenLabs agents pricing: $0.08 per minute, plan minutes and concurrency. Read 16 October 2026.
- Macha, ElevenLabs Agents pricing explained: model cost per minute, burst and text rates, yearly billing. Prices checked 19 September 2026.
- Bland AI pricing and Bland billing documentation: plan rates, transfer rates and billing basis. Read 16 October 2026.
- Macha, Bland AI pricing explained: daily call caps and the December 2025 change. Prices checked 19 September 2026.
Prices verified 16 October 2026. Vendors change pricing often; check the linked pages before you budget. All figures are modelled from list prices.
Related reading: AI Answering Service Cost: Real Prices at 100, 500 and 2,000 Calls a Month, How Much Does It Cost to Run an AI Agent Every Month?, Outcome-Based AI Agent Pricing: 4 Questions to Ask Before You Sign, Small Business AI Budget: What a 5-Person Team Actually Spends Each Month and Is n8n Really Free? The True Monthly Cost of Self-Hosting.