Ada AI Pricing Explained: What Ada CX Actually Costs

Created time

Jul 10, 2026 04:07 PM

Title length (<60)

Author

Mike Heap

Last optimised

Ecomm?

How does Ada pricing actually work?

⚡

TL;DR: Ada publishes no price and bills for every conversation its AI handles, whether or not that conversation resolves. So the meter doesn't move with how well the AI does, and the whole quote arrives only after a sales call, once your volume clears Ada's minimum.

Ada publishes no pricing. The pricing page is a sales-consultation booking form: no free tier, no self-serve plan, no monthly option.

Its single concrete fact is the qualifier I opened with: you should be handling at least 300,000 customer service conversations a year before you're a fit. Everything else is quoted after a call.

Almost every other article gets this backwards. Ada bills by the conversation, and it says so itself.

In Ada's own post on AI agent pricing models, Ada argues for conversation-based pricing and against resolution-based, on the grounds that charging per resolution "punishes success": the better your AI gets, the more the vendor bills you.

So the "$1-$3.50 per resolution" figure the whole internet repeats about Ada describes a model Ada publicly argues against. It has no citable origin, and I'm not going to launder it here.

Independent contract data backs that up, and I trust it more than any vendor's own framing. Vendr aggregates real purchase data and describes Ada's pricing as "a combination of conversation volume, feature tier, and deployment complexity," with overage running roughly $0.05-$0.50 per conversation above your contracted allowance.

On top of that sits an annual platform fee, which Vendr puts at $30,000-$70,000 a year for its smallest deployment band. The lowest Ada contract in Vendr's data, counting everything, was $36,026. Add a one-time professional-services implementation, and here's what an Ada invoice actually looks like:

Breakdown of an Ada invoice into three components: an annual platform fee from around $30,000 a year, a per-conversation meter of about $0.05 to $0.50 with no published rate, and a one-time implementation of $15,000 to $100,000-plus.

Component	Price	How it's billed	The catch
Annual platform fee	~$30,000+/year	Annual commit, negotiated	The entry point observed across real contracts; no self-serve, no monthly
Automated conversations	~$0.05-$0.50 each	Per conversation, as overage above a contracted allowance	Ada publishes no rate card. The band comes from Vendr's observed contracts
Implementation / professional services	~$15,000-$100,000+	One-time, scaled by deployment complexity	Lands on top in year one. CSM-led delivery, no self-serve path
Voice	Not publicly priced	Per minute (sales-gated)	Ada has voice AI, but no citable per-minute rate exists
Channel / other add-ons	Not publicly priced	Sales-gated	Bundled into the negotiated contract

One point carries through the rest of this post: because Ada bills for conversations handled regardless of whether they resolve, your contract value doesn't move with your resolution rate. That works in Ada's favor, since you don't get billed more as the AI improves. The flip side is that you're committing to a meter reading before you can see the meter.

I wrote this post because of one contrast. You can read My AskAI's price, $0.10 a ticket, published on the page, without talking to anyone. With Ada, the price is a conversation you have to earn your way into.

What does Ada's platform fee cost, and what are you actually paying for?

⚡

TL;DR: The platform fee starts around $30,000 a year and buys the platform itself: the reasoning engine, the channels, the console, before a single conversation is automated. It's an annual commitment with no monthly option, no self-serve tier and no trial.

Before a single conversation is automated, you're paying the platform fee, which Vendr puts at $30,000-$70,000 a year for the smallest deployments (the lowest Ada contract in its data, all in, was $36,026). What that fee buys, as far as I can tell, is the platform itself: the reasoning engine, the channels, the console. The conversation meter sits on top of it.

Vendr's contract data stacks up by deployment size. A "Small" deployment runs about $30,000-$70,000 in platform fee plus $15,000-$30,000 of implementation; "Mid" is roughly $70,000-$150,000 plus $25,000-$50,000; "Enterprise" is $150,000-$300,000+ plus $50,000-$100,000+.

Across 112 recorded purchases the median all-in was $73,500 a year (the range runs from that $36,026 floor up to $318,400). That's a wide spread. With no list price, where you land depends on how hard you negotiate and how complex your deployment is.

Underneath all of it we're looking at annual commitment, sales-negotiated, with no monthly billing, no self-serve tier and no trial. You sign for a year before you've run a single ticket through it.

For an enterprise procurement team, I know that's business as usual. For anyone who wanted to test-drive before committing budget, it's a wall.

What counts as a "resolution" when Ada bills you?

⚡

TL;DR: Ada bills by the conversation and argues publicly against resolution-based pricing, yet it publishes no billing-time definition of what a conversation is. That undefined unit is the biggest variable in your invoice.

For Ada, the obvious question (what counts as a resolution) is the wrong one.

Ada bills by the conversation. It advocates that model and, in its own words, criticizes the entire resolution-based approach. So the question worth asking is "what counts as a conversation," and that's the one Ada leaves open.

Read Ada's blog closely, like I did, and you find it doing something clever. It points out, correctly, how inconsistently the word "resolution" is defined across vendors:

"one vendor might define it as there being no customer response for 5 minutes; another might consider any interaction that didn't escalate to a human agent as resolved"

Ada uses that inconsistency as an argument for its own model. But it then declines to publish a billing-time definition of its own unit either.

So I'm left asking: what closes a conversation? A session, or a rolling 24-hour window? Do two messages across chat and email merge into one conversation, or count as two?

There's no published rate, no tiers, and no definition. The biggest variable in your invoice is a word Ada bills on but never pins down.

Compare that to the vendors who do define the unit they charge for:

Vendor	What you're billed on	Is the unit defined at billing time?
Ada	Automated conversation	No. Argues for conversation-based pricing, publishes no definition of a conversation
Zendesk	Automated resolution	Yes. A 72-hour quiet period with no reopen and no human touch
HubSpot Breeze	Resolution	Yes. Resolved and not handed to a human for 72 hours
Intercom Fin	Outcome / resolution	Yes. An end-to-end AI resolution, or a configured Procedure handoff
My AskAI	Ticket	Yes. Per ticket, whether or not it resolves, with no vendor-defined "doesn't count" rule

My AskAI is the cleanest contrast in the category here: we charge per ticket, flat, at $0.10.

Table comparing the billable unit each vendor charges on and whether it is defined at billing time: Ada bills on the conversation and does not define it; Zendesk, HubSpot Breeze, Intercom Fin and My AskAI each bill on a defined unit.

There's no vendor-controlled definition of "resolution" to negotiate, because we don't bill on resolution. We count a conversation resolved when the AI handled it without escalating to a human, and we make escalation easy. We don't pretend to know an issue was truly solved.

Here's the knock-on effect at budgeting time: because the bill doesn't move with your resolution rate, your cost per resolved ticket falls as the AI improves. Most of what pushes a resolution rate up is your own work: better knowledge, tighter guidance, connected tools. A per-ticket meter lets you keep that upside, and the vendor doesn't charge you more for it.

What does Ada cost at typical volumes?

⚡

TL;DR: At Ada's own minimum of 25,000 conversations a month, the median $73,500 contract works out to roughly $0.25 per conversation and about $0.35 per resolved conversation at a 70% rate. Below 300,000 conversations a year Ada declines the deal, so two of the three volumes everyone else quotes can't be filled in good faith.

Every pricing post is supposed to nail the three-volume worked example: 1,000 / 10,000 / 50,000, with the effective cost per resolved conversation at each. I'll run it. But for Ada, two of those three rows can't be filled in good faith.

Table of Ada cost at four monthly volumes: at 1,000 and 10,000 conversations a month Ada declines the deal, below its 300,000-a-year bar; at 25,000 a month the median contract is about $73,500 a year, roughly $0.35 per resolved conversation; at 50,000 a month roughly $70,000 to $150,000 a year in platform fees.

First, the assumptions, all traceable. There's no published rate card, so I modeled from Vendr's observed contracts: a median annual contract of $73,500 across 112 purchases (range $36,026-$318,400), Vendr's per-band platform and implementation figures, and overage of $0.10-$0.50 per conversation above the contracted allowance.

Ada's own qualification bar is 300,000 conversations a year, which is 25,000 a month. If you sit the median customer right at that bar, the $73,500 contract works out to roughly $0.245 per conversation all-in. That's a rate I derived from real contracts, and Ada publishes no number of its own.

For the resolution rate I use the ~70% field median from our own benchmark corpus (195 rated deployments across roughly 55 vendors; it's an aggregate, directional and not apples-to-apples, so treat it as a reference point). A conservative 50% gives me a labeled sensitivity floor.

Volume	Annual conversations	Will Ada sell at this volume?	Observed annual cost	Effective $/conversation	$/resolved @70%
1,000/mo	12,000	No. 4% of Ada's stated minimum	Ada declines the deal	n/a	n/a
10,000/mo	120,000	No. 40% of the minimum	Ada declines the deal	n/a	n/a
25,000/mo	300,000	Yes. Exactly at Ada's bar	~$73,500 (median, n=112)	~$0.25	~$0.35
50,000/mo	600,000	Yes	~$70,000-$150,000 (Vendr "Mid" band)	~$0.12-$0.25	~$0.17-$0.36

At 1,000 and 10,000 conversations a month

Every other article on this SERP will quote you an Ada price at 1,000 tickets. Ada's own pricing page says it won't take that deal.

A thousand conversations a month is 12,000 a year, four percent of Ada's stated minimum. Ten thousand a month is 120,000 a year, still only forty percent of the bar.

At both volumes, the only fair cell reads "Ada would decline." I'm leaving those cells empty on purpose, because filling them with an invented number is exactly the mistake the rest of the internet makes.

Video preview — I Let AI Agents Resolve 10,000 Support Tickets, Here's How Much It Cost

At 25,000 conversations a month

This is the first row where Ada is actually purchasable. You're standing exactly on the 300,000-a-year bar. At the median $73,500 contract that's about $6,125 a month: an effective ~$0.25 per conversation, and ~$0.35 per resolved conversation at a 70% rate.

Drop the resolution rate to a conservative 50% and the contract doesn't change at all, because the bill doesn't move with your resolution rate by design. So only the denominator moves, and the effective cost per resolved conversation rises to about $0.49. Implementation ($15,000-$100,000+) lands on top in year one.

At 50,000 conversations a month

At 600,000 conversations a year you're comfortably above the bar and inside Vendr's "Mid" band (which it defines as 25,000 to 100,000 conversations a month): roughly $70,000-$150,000 a year in platform fees, plus $25,000-$50,000 of implementation in year one.

Spread across 600,000 conversations that's an effective ~$0.12-$0.25 each, or ~$0.17-$0.36 per resolved conversation at a 70% rate. Notice which way that moves. Ada gets cheaper per conversation as you scale, because the platform fee is the fixed part.

This is the volume Ada is built for, and I think the unit economics hold up here. What shuts out everyone below is the entry point and the bar.

What doesn't the Ada pricing page tell you?

⚡

TL;DR: Four line items decide the deal and none appear on the booking form: the 300,000-conversation floor, an 8 to 16 week professional-services implementation, the helpdesk bill you still pay underneath Ada, and no way to test the resolution rate that drives your invoice before you sign.

The booking form is quiet about four line items that actually decide the deal.

Ada's pricing page is a sales-consultation booking form. Its one concrete fact is the qualifier: a fit for companies with at least 300,000 annual conversations.

The 300,000-conversation floor is itself a price signal

The qualifier is a price in disguise: a great fit above 300,000 conversations a year. Below that line, you aren't a customer, and no amount of interest changes that. For most teams reading a pricing post, the most important fact about Ada's cost is that they'll never be quoted one, and that's most of the folks I hear from.

Professional-services-led implementation, measured in months

Ada deployments are CSM-heavy and run roughly 8-16 weeks (Vendr puts implementation at $15,000-$100,000+, depending on complexity). Workflow changes often route back through customer success, and there's no self-serve console for them. That's a real cost in both money and calendar time, and it's nowhere on the pricing page.

You still pay for your helpdesk

Ada is not a ticketing system. It layers above Zendesk, Salesforce, Genesys, Oracle or NICE and connects by API, so Ada's bill sits on top of whatever you already spend on the helpdesk underneath. When you're modeling total cost, Ada is a line you add on top of your existing helpdesk spend.

You can't test the thing that determines your bill

This one compounds. I don't think you should sign an Ada contract without a way to test first, and Ada gives you none: no trial, no sandbox, no way to simulate against your historical tickets. You can't measure the conversation volume or the resolution rate that drives your invoice before you commit to a year.

The per-conversation price is negotiable but knowable. Your own volume and rate are not, and Ada gives you no way to read them in advance. On the buyer side, the absence of a published price makes your costs hard to forecast.

What do real Ada customers actually pay (and complain about)?

⚡

TL;DR: There's no verifiable public genre of Ada billing complaints, and we won't invent one. When every customer is above 300,000 conversations a year on a negotiated contract under NDA, the invoice never gets screenshotted into a subreddit.

I went looking for the genre that every other pricing post leans on: the Reddit and G2 threads where customers vent about the invoice. For Ada, there isn't one.

So I'm not going to hand you a quote. No paraphrase, no "sources suggest," nothing invented.

There isn't a public, verifiable Ada billing complaint to point you at, and inventing one would be worse than saying so.

This is the one place I'll happily give Ada the win. There's no public invoice-complaint genre for Ada the way there is for the self-serve vendors, and that's a structural consequence of who Ada sells to.

When every customer is above 300,000 conversations a year, on a negotiated annual contract under NDA, the invoice never becomes a public thing that gets screenshotted into a subreddit.

Ada's public negatives, where they exist, are about the end-user experience: conversation loops, lost context, hard escalation. Those complaints come from a different person than the one worried about the bill.

The real risk with Ada, then, is the structural cost one. You won't find people complaining that Ada's bill surprised them, because the people who'd complain never got a quote, and the people who did signed an NDA. The much-repeated "someone paid $300k+" line is pure word-on-the-street: it has no verifiable source and traces back through a competitor's blog I won't cite, so it stays out.

Does Ada have a free trial, and what are the contract terms?

⚡

TL;DR: No free trial, no free tier, no self-serve plan. The only routes in are a sales consultation or an on-demand demo, both ending in an annual, sales-negotiated contract.

No free trial, no free tier, no self-serve plan. The only two doors into Ada are a sales-consultation booking and an on-demand product demo, and what's behind them is an annual, sales-negotiated contract.

If your buying process depends on proving value before you commit budget, Ada's answer to "can I test it first?" is no.

"Let me test it before I commit" is the most reasonable request a support leader can make of an AI vendor, and Ada structurally can't grant it, since there's nothing to test without a signed contract. For contrast, the way we handle that objection at My AskAI is a 30-day free trial with every feature unlocked, unlimited tickets, and no card required (you do sign up with a business email). The philosophies are opposite: Ada asks you to trust the model and negotiate the price; we ask you to run it on your real tickets first and read the price off the page.

How does Ada pricing compare to alternatives at the same volume?

⚡

TL;DR: At 25,000 conversations a month, the one volume where every vendor here is purchasable, Ada runs about $0.35 per resolved conversation against Intercom Fin's $0.99 and My AskAI's $0.14. Ada's conversation-based model stops charging you more as the AI improves, but the $30,000 floor and 300,000-conversation bar shut out everyone below.

Every comparison you'll read pits Ada against the field at 10,000 tickets a month. That comparison is misleading, because Ada won't sell at 10,000, so you'd be quoting a price for a deal Ada refuses. The only volume where every one of these vendors is purchasable is 25,000 conversations a month, Ada's own stated minimum, so that's where I've run it.

Vendor	Pricing model	Cost at 25k/mo	Effective $/resolved @70%	Notes
Ada	Conversation-based (resolution-independent)	~$6,125/mo ($73,500/yr median, n=112)	~$0.35	Bill flat vs resolution rate; implementation $15k-$100k+ extra in year one; sells only above 300k conv/yr
Intercom Fin	Per-outcome, $0.99	~$17,325/mo (standalone, no Intercom seats)	~$0.99	Bill rises as the AI resolves more
My AskAI	Per-ticket flat, $0.10	~$2,500/mo	~$0.14	Resolution-independent; cost per resolved ticket falls as the rate climbs
Decagon	Per-resolution (enterprise, non-public)	Directional only	~$2-$3/resolution	No public worked example; sales-quoted
Sierra	Per agreed outcome (enterprise, non-public)	Directional only	n/a	$150k-$386k a year; directional only

At its own minimum volume, Ada is cheaper per resolved conversation than Intercom Fin: roughly $0.35 against $0.99.

Ranking of cost per resolved conversation at 25,000 conversations a month: My AskAI about $0.14, Ada about $0.35, Intercom Fin about $0.99. Decagon and Sierra are enterprise, sales-quoted and directional only, so they are not plotted.

The conversation-based model stops charging you more for getting better, and at 25,000 conversations a month that's a real advantage over a per-outcome vendor. Ada's cost problem is the $30k floor and the 300k bar that shut out everyone below it, plus the fact that you can't measure your own numbers before you sign. The unit economics at scale are fine.

Decagon and Sierra sit in the same enterprise, sales-quoted world as Ada, and I'm not going to put a firm monthly figure in their cells: there's no public worked example to build one from, and false precision would undercut everything else here. They're directional: Decagon is roughly per-resolution in the $2-$3 range, and Sierra prices per agreed outcome on six-figure annual contracts.

And to name where Ada is simply the right call: if you need voice, chat, email and social all running on one reasoning engine, or you're routing through a contact center like Genesys, NICE or Oracle, or you operate at the enterprise scale Ada was built for, Ada does things My AskAI doesn't sell. We don't do voice, and we're not a contact-center platform. My concern here is narrower: whether you can compute your bill before you commit, and I'm not claiming My AskAI is unconditionally cheaper or better.

Is Ada actually worth the money?

⚡

TL;DR: Ada is a strong product priced for one specific buyer: an enterprise above 300,000 conversations a year with a CX team to run it. If you're below that bar, need to test before you buy, or want to read a price off the page, Ada isn't built for you.

Ada is a strong product priced for a specific buyer. Its own case studies report resolution rates ranging from 45% to 84% at real enterprise scale. You're deciding whether you're the buyer Ada's built for, and whether you can live with committing to a bill you can't fully compute in advance.

✅

Ada is worth it if:

You're above 300,000 conversations a year, and Ada will tell you itself if you're not.

You need voice, chat, email and social handled on one reasoning engine.

You're on Salesforce, Genesys, Oracle or NICE and want a CX automation layer above them.

You have a dedicated CX-automation owner to run and tune the platform.

Procurement prefers a single enterprise contract over a metered self-serve tool.

❌

Ada isn't worth it if:

You're under roughly 300,000 conversations a year (Ada declines the deal).

You need to test on your real tickets before you buy, but there's no trial or sandbox.

Your budget is sensitive to a conversation volume and resolution rate you can't yet measure.

You want to read a price without booking a sales call.

You want to keep your existing helpdesk spend flat, but Ada's cost is additive on top of it.

The founder view I'll leave you with: at scale, Ada's per-conversation economics are competitive, and the model that doesn't charge you more as it improves is a real feature. My problem was never the price.

The critique is that Ada asks you to commit to a bill whose size you can't compute in advance, on a unit it hasn't defined, above a volume most teams won't reach. If that's your world (enterprise scale, a CX team to run it, procurement that likes one contract), Ada earns its place.

We've watched that trade play out. RecruitCRM turned down Fin at $0.99 a resolution and climbed from 35% to 68% resolution on a flat per-ticket bill, and Apartment List now resolves 76% of its Zendesk tickets, saving roughly 101 hours a month. There are more of those in our case studies.

If it isn't, you want a model you can read on the page and test before you pay. You can compare the full feature picture in our Ada guide and the cheaper-at-your-volume options in the Ada alternatives round-up, or just read our pricing without talking to anyone.

FAQs

How much does Ada CX cost?

Ada publishes no public price. Independent contract data from Vendr puts the platform floor at roughly $30,000 a year and the median all-in contract at $73,500 a year (across 112 recorded purchases), with a conversation-based meter of about $0.05-$0.50 per conversation on top. Everything is quoted after a sales call, and Ada only sells to companies above 300,000 conversations a year.

What is the average cost of an AI agent?

There's no single average, because the category spans a very wide range depending on the pricing model. It runs from about $0.10 per ticket (My AskAI's flat per-ticket rate) to $0.99 per outcome (Intercom Fin) up to six-figure annual contracts for enterprise vendors like Ada and Sierra. The pricing model tells you more than the sticker does: per-ticket stays flat as your AI improves, while per-resolution and per-outcome bills rise with success.

How much does an AI chat bot cost?

For self-serve and mid-market tools, expect roughly $0.10-$0.65 per ticket or per conversation, often on a published rate you can read without a sales call. Enterprise platforms like Ada are sales-gated with annual contracts starting around $30,000 a year plus usage and implementation. The gap between those two worlds is really about who the vendor is built to sell to, something I come back to a lot on calls.

Does Ada publish its pricing?

No. The Ada pricing page is a sales-consultation booking form. Its only concrete detail is a qualifier that Ada is a fit for companies with at least 300,000 annual customer service conversations. There's no rate card, no tiers, and no self-serve price.

What counts as a resolution in Ada?

Ada doesn't bill on resolution at all. It advocates conversation-based pricing and argues against the resolution-based model on the grounds that it "punishes success." The twist is that Ada then publishes no billing-time definition of its own unit, the conversation, either. So the biggest variable in your invoice is a word Ada bills on but never pins down.

Does Ada charge for conversations the AI doesn't resolve?

Yes. Because Ada is conversation-based, every automated conversation is billable regardless of whether it ends in a resolution (per Vendr's contract data and Ada's own model blog). That's the flip side of the design: the bill doesn't rise as resolutions rise, but you do pay for conversations that don't resolve.

Is there a minimum spend or contract minimum with Ada?

Effectively yes, on two axes: a platform floor of roughly $30,000 a year, and a qualification bar of 300,000 conversations a year below which Ada positions itself as not a fit. Contracts are annual and sales-negotiated, with multi-year deals attracting discounts.

Does Ada have a free trial?

No. No free trial, no free tier, and no self-serve plan. The only routes in are a sales consultation or an on-demand demo, ending in an annual contract. If testing before you commit matters to you, that's worth weighing: some alternatives (My AskAI included, with a 30-day all-features trial) let you run the AI on your real tickets before paying.

Do I still pay for Zendesk or Salesforce if I use Ada?

Yes. Ada isn't a ticketing system of its own. It layers above your existing Zendesk, Salesforce, Genesys or Oracle stack via API, so its cost sits on top of what you already spend on the helpdesk underneath. Budget for Ada as a line you add on top of your existing spend.

How does Ada pricing compare to Intercom Fin?

At 25,000 conversations a month, Ada works out cheaper per resolved conversation than Fin, roughly $0.35 against Fin's $0.99 per outcome, because Ada's conversation-based bill stays flat while Fin's rises with every resolution. The trade-off is Ada's $30,000 floor and 300,000-conversation bar, where Fin is purchasable at far lower volumes. I've put the full side-by-side in the comparison table above.

Are there extra charges for voice or multilingual support?

Ada offers voice AI and supports many languages, but there's no citable public per-minute or per-language rate. Both are folded into the negotiated contract and sales-gated. I'd treat any specific voice or multilingual number you see quoted elsewhere with caution; Ada doesn't publish one.

Can I negotiate Ada's price down?

Yes. With no list price, what you pay is largely a function of how hard you negotiate, and Vendr reports average savings of around 17% on Ada contracts. Multi-year commitments typically unlock a further 15-30% off. The absence of a published price cuts both ways: it means the number is movable, but also that two similar companies can pay very different amounts.

Ada AI Pricing Explained: The Real Cost Behind "Talk to Sales"

How does Ada pricing actually work?

What does Ada's platform fee cost, and what are you actually paying for?

What counts as a "resolution" when Ada bills you?

What does Ada cost at typical volumes?