AI Agent Cost (2026): $29/Seat SaaS vs $3,500 Once

If you’ve Googled “AI agent pricing” and walked away more confused than before, I get it. A monthly seat, an AI outcome, a phone minute, and a custom build are four different units. Comparing headline prices without normalizing them produces bad math.

Short answer: As of October 15, 2026, AI agent software starts at $29 per Intercom Essential seat billed annually ($39 monthly) or $32.50 for 50 Tidio Lyro conversations, then usage raises the bill. Workflow Automation and AI Front Door Look Up are $3,500 once. Desk is $7,500. The receptionist is $8,000. Compare 24 months at your real volume.

Buying model	Published starting point	What changes the bill
Tidio Lyro SaaS	$32.50/mo for 50 AI conversations	Monthly AI conversation quota
Intercom + Fin	$29/seat/mo annual ($39 monthly) + outcomes	$0.99 standard outcome; $9.99 qualification
Workflow Automation	$3,500 once	One named ops job, plus direct provider usage
AI Front Door	Look Up $3,500 · Desk $7,500 · Floor $14,000 once	Which actions an AI app may take
AI Receptionist	$8,000 once	Call volume on your own phone account

The three ways to buy an AI agent#

There are three buying models: rent a platform, commission fully custom software, or pay once for a focused deployment you own. Start with the job, who owns the setup, and the usage bill that continues. The smallest advertised number is a weak place to start.

SaaS subscription. Companies like Tidio and Intercom sell you access to their platform. You connect it to your website, configure it in their dashboard, and pay monthly. You never own the underlying system. If they raise prices, change features, or kill a plan, you absorb it.

Custom development agency. A software agency builds something from scratch: your own code, your own infrastructure, your own model connections. You own it when they’re done. You also pay for discovery, architecture, build, QA, deployment, and documentation. That path fits a complex operation with unusual systems. It is oversized for a salon, a law firm, or a contractor with three crews.

One-time deployed agent. A builder scopes one workflow, connects it to the tools you already run, tests it against real scenarios, and hands it off. You own what’s built. No monthly license to the builder. Phone, model, and hosting bills stay on your accounts. If you are still choosing among the three, the build vs buy AI agent breakdown puts Tidio, Intercom Fin, and a one-time build on the same 24-month table.

How much does AI agent software cost?#

AI agent software currently starts at $29 per Intercom Essential seat on annual billing, or $32.50 a month for 50 Tidio Lyro conversations. Seats, billing cadence, AI conversations, and Fin outcomes set the working bill. Price the volume your business expects.

Tool	Starting price	AI features	Where it gets expensive
Tidio Lyro	$32.50/mo for 50 conversations, annual	Standalone AI agent quota	The free 50 conversations are lifetime-only; higher quotas cost more
Intercom Essential	$29/seat/mo annual ($39 monthly) + outcomes	Fin on every plan; usage is separate	$0.99 standard outcome; $9.99 qualification
Intercom Advanced	$85/seat/mo annual ($99 monthly) + outcomes	Workflows plus 20 Lite seats	Same Fin meter, on a higher seat

These are vendor-published prices I rechecked on October 15, 2026. According to Tidio’s pricing page, Lyro starts at $32.50 per month for 50 AI conversations, advertised with two months free (annual billing). Tidio’s Lyro limit article still says the free 50 Lyro conversations are a one-time quota unless you buy a monthly package. According to Intercom’s pricing page, Essential is $29 per seat per month on the annual toggle, Advanced is $85, and Expert is $132, with Fin from $0.99 per outcome. Intercom’s plan comparison, updated the week I checked, lists the same Essential seat at $39 per month if you bill monthly, and Advanced at $99 monthly. Its outcome billing documentation puts a resolution, procedure handoff, or disqualification at $0.99 and a qualification at $9.99, at most one outcome per conversation. Copilot unlimited usage is $29 per agent per month billed annually, or $35 monthly, on that same pricing page. The Pro add-on starts at $100 per month. I keep the Fin meter alone in Intercom Fin cost.

The bigger issue for most small businesses is what these tools will not do at the sticker price. Most SaaS AI agents are built for ecommerce or B2B support chat. After-hours call routing, a CRM note, appointment booking, and a lead that has to land in the system you already run usually means paid add-ons, or a different product. Thinking through what AI for small business actually needs to do changes the decision.

How much does it cost to build an AI agent?#

Building an AI agent has no defensible universal price. Scope, integrations, testing, and ownership change the quote. My focused deployments start at $3,500 once. A bespoke agency project needs a written scope before its price means anything.

If you want an agent built rather than rented, there are two routes: a fully custom software project, or a one-time deployment scoped to a known workflow. I publish my deployment prices below. For an agency project, ask for a line-item quote covering discovery, architecture, engineering, QA, deployment, documentation, and maintenance. I explain the scope drivers in my guide to the cost to build an AI agent.

A fully custom build covers discovery, architecture, engineering, QA, deployment, and handoff documentation. Without a defined scope, a published market range is too broad to help an owner budget honestly.

Custom development makes sense when you have a complex multi-location operation, compliance requirements that rule out third-party SaaS, or infrastructure that needs deep bespoke integration. If you run a law firm, salon, HVAC company, or solo consultancy, start with one named job.

AI agent deployment costs#

AI agent deployment, on my published menu, is a one-time fee for one named job: $3,500 for Workflow Automation or AI Front Door Look Up, $7,500 for Desk, $8,000 for the receptionist, and $14,000 for Floor. You own the setup. Phone, model, and CRM bills stay on your accounts.

Offer	Price	What it handles
Workflow Automation	$3,500 once	One repeating ops job in the tools you already pay for
AI Front Door Look Up	$3,500 once	Answers on hours, prices, and status from the AI app the customer already uses
AI Front Door Desk	$7,500 once	The customer books, pays, or requests a quote from that app
AI Front Door Floor	$14,000 once	Desk, plus owner questions across up to five systems
AI Receptionist	$8,000 once	24/7 phone answering, booking, CRM writes, escalation

MCP is the open standard ChatGPT, Claude, Grok, and other AI apps use to reach a business’s tools. AI Front Door is that connection. It is not a chatbot locked to one vendor’s app. Workflow Automation is the back-office job: the invoice chase, the Friday numbers, the form that should already be a CRM row. The same prices are on the pricing page.

What that price includes: a scoping session, the build, the connection to the system of record you named, testing against real conversation or job patterns, and a handoff so you can run it. I don’t charge per call, per message, or per seat. A second job is a second quote.

Infrastructure — Twilio for voice or SMS, an AI model API, and whatever your CRM requires — you pay directly. I quote that from expected usage because a flat estimate would be fake precision. According to Twilio’s U.S. voice pricing, the page says pricing is current as of August 2026, and those figures were unchanged when I rechecked on October 15, 2026: local inbound calls are $0.0085 per minute plus $1.15 per month for the number; outbound local is $0.0140 per minute. AI and other voice features add separate charges. It’s on your accounts, not mine.

This is what owning versus renting an AI agent looks like in practice: the billing relationship is with the underlying services, not with a platform that can change its pricing page next quarter. If the question is specifically a Telegram bot’s API bill, that channel math is in what a Telegram AI agent costs.

What’s the AI agent cost per month?#

AI agent cost per month is a SaaS question. Intercom Essential is $29 to $39 per seat plus outcomes. Tidio Lyro starts at $32.50 for 50 conversations. A one-time deploy has no monthly license to me. Twilio and the model provider still send their own invoices.

Buyers type “ai agent cost per month” because every SaaS quote is a recurring line. That’s the right unit for Intercom and Tidio. It is the wrong unit for a deployment you own.

On Intercom, the monthly number is seats plus Fin. One Essential seat billed annually is $29; billed monthly it is $39. Add 500 standard outcomes at $0.99 and you are at $524 to $534 before qualifications, extra seats, Copilot, Pro, or phone. On Tidio, $32.50 is the Lyro floor for 50 AI conversations on the public annual price. More conversations move the quota, and a Customer Service plan is a separate product if humans also reply.

On a deployment, the monthly line is infrastructure you already pay: a Twilio number, minutes, model tokens, maybe a small host. There is no license to me after handoff. If volume jumps, you see it in your own dashboards.

What’s the cost per task AI agent buyers actually pay?#

Cost per task, on a public price list, is Intercom’s $0.99 outcome, or $9.99 when Fin qualifies a lead. My deployments add no per-task fee. Volume still costs model tokens and phone minutes, billed by those providers.

“Cost per task AI agent” is a real search. Vendors like the phrase because ninety-nine cents sounds cheap. Multiply it by a busy month and it isn’t.

Intercom’s published Fin meter is the cleanest public example: $0.99 per successful standard outcome, $9.99 per qualification, at most one charge per conversation. I keep that meter-by-meter breakdown in cost per task AI agent. The decision is whether you want a fee that scales with every conversation, or a one-time build plus pass-through usage.

The 24-month math#

A 24-month comparison has to hold volume still. These verified floors are useful, and they are not complete quotes. Tidio and Intercom rise with usage. A deployed agent adds direct infrastructure and any second job you name later.

Option	Upfront	Monthly ongoing	24-month total
Tidio Lyro minimum (50 AI conversations/mo)	$0	$32.50/mo	$780
Intercom (1 Essential seat annual + 500 standard outcomes/mo)	$0	$524/mo	$12,576
Intercom (same volume, seat billed monthly)	$0	$534/mo	$12,816
Workflow Automation or AI Front Door Look Up	$3,500	Direct usage varies	$3,500 + usage
AI Front Door Desk	$7,500	Direct usage varies	$7,500 + usage
One-time AI Receptionist	$8,000	Direct usage varies	$8,000 + usage

At 500 standard outcomes per month, the Intercom arithmetic is $29 for one Essential seat billed annually plus $495 in outcomes, or $524 monthly. Over 24 months that is $12,576. Pay the seat monthly and the same volume is $39 + $495 = $534, or $12,816 over 24 months. Qualifications change the total sharply because the published rate is $9.99, not $0.99. Advanced at $85 per seat on annual billing adds $56 a month to the Essential seat before any outcomes. This is a pricing comparison, not a claim that the products deliver identical workflows.

Most buyers run this math after signing up for a subscription. If you want the receptionist-specific version — what you get at $8,000 versus answering services and front-desk staff — AI receptionist pricing breaks it down.

What you’re actually paying for in a deployment#

A deployment quote pays for workflow judgment as much as code: scoping the handoffs, connecting the systems, testing real edge cases, and documenting how you run it after launch. A quote that skips those pieces gets expensive when you rebuild it.

Scoping. Understanding the real workflow — where the work comes from, what has to be captured, which system needs the record, and where a human must step in. Usually one or two sessions, and the part that decides whether the agent works.

Build. Connecting the model to your tools. For a phone agent: Twilio for voice, a model API, and your CRM. For AI Front Door: the MCP connector the customer’s AI app already speaks, pointed at the calendar, prices, or payments you named. For Workflow Automation: the one ops job, inside Jobber, QuickBooks, HubSpot, or the sheet you already use.

Testing against real scenarios. Actual message patterns and job rows, including the off-script ones and the ones the original scope did not anticipate.

Handoff. A session where you learn how to update the answers, adjust escalation, read the logs, and see when something needs attention. You’re not dependent on me after that point.

None of this pays for someone else’s platform infrastructure. You’re paying for work scoped to your business.

To run the 24-month math with your own numbers, use the free subscription vs. own calculator. No signup.

When this isn’t the right move yet#

Do not buy an agent before the workflow is stable. A one-time deployment fits a repeatable process with a system of record and a clear moment a human takes over. If the offer or the records keep changing, fix that first.

• You have a defined, repeatable intake or ops workflow — calls, DMs, form fills, invoice chases
• You’re losing leads to slow follow-up or gaps in after-hours coverage
• You have a CRM or system of record where the result can be written
• You’ve operated long enough to know which questions come in and which answers move people forward

Wait if:

• You don’t have a system of record yet — captured data disappears when the conversation does
• Your workflow changes every few months — you’ll pay to redeploy faster than the savings accumulate
• You’re still testing whether the market wants what you sell — fix the offer before automating the intake

If you’re not sure where you land, start with a workflow audit, not a purchase. I run those through my workflow audit — 30 minutes to map what you have and whether automation is the right next move.

What to do next#

Run the 24-month math on your expected seats, conversations, outcomes, and phone minutes. Then match the leak to one offer. A lower starting price does not help if the product cannot finish the job.

If a customer should book, pay, or ask a live question from ChatGPT, Claude, Grok, or another AI app that supports MCP connectors, AI Front Door is the fit. Desk is $7,500 once. If the leak is one repeating ops job in tools you already pay for, Workflow Automation is $3,500 once. If the leak is the phone, the receptionist is $8,000 once, and AI receptionist pricing is the comparison. Send the workflow audit if you want the map in writing. If you would rather talk it through, book a 20-minute call.