Regal Review (2026)
Event-triggered calling with AI agents and human handoff, sold only to contact centres clearing 25 agents and 100,000 minutes a month.
Some links on this page can be affiliate links. We may earn a commission at no cost to you. This never affects our rankings. How we stop it.
Regal in Depth
Regal's pricing page publishes one useful sentence, and it is not a price. It is a floor: a good proxy is having at least 25 agents and at least 100,000 minutes of calls a month.
That sentence is two weeks old. Through 22 July it read 75 agents or 150,000 calls a month. Three things changed at once: the agent count fell by two thirds, the volume unit changed from calls to minutes, and the or became an and.
| Read on | The floor |
|---|---|
| Sep 2025 through Jul 2026 | at least 75 agents or 150,000 calls a month |
| 6 Aug 2026 | at least 25 agents and at least 100k minutes of calls a month |
The old bar could be cleared on either axis. The new one requires both, so a thirty-agent centre that previously qualified on headcount alone now has to clear a volume test as well. The floor came down and got harder to pass at the same time.
The only rate Regal publishes is an average, and it is on a blog
There is no rate card. The single per-minute figure anywhere on the site sits in a blog post about the cost of voice agents, and it says Regal's agents generally come in at $0.20 a minute, based on data from millions of calls handled.
Read what that is. Not a list price, not a floor, not a ceiling. It is a realized average across the vendor's book of business, which is a description of other people's invoices rather than an offer to you. It is more than most vendors in this category publish and it is not a price.
This site does not record $0.20 as Regal's per-minute rate for that reason, and does not build a cost stack from it. A stack needs a rate whose exclusions are stated, and Regal states only that one language model is included in what it calls base AI voice agent minute pricing. Every other model is flagged as more expensive, with no surcharge published for any of them.
Three meters, one of them acknowledged
Minutes are the meter everyone discusses. Two more exist.
Concurrency is purchasable: every agent handles up to fifty simultaneous calls, the order form defines how many are included, and more can be bought at a price that is not published. Past fifty the documentation's advice is not to buy more but to duplicate the agent and round-robin between the copies, which is a workaround presented as a feature.
And there is a platform fee, or there was. In 2022 the same pricing page answered the question directly: a platform fee based on the number of cumulative contacts, then usage on top. That answer has been deleted, along with one confirming a per-user charge and another promising integrations cost nothing extra. The only three disclosures Regal ever made about the shape of its bill are all gone from the live page.
The meter can run while the phone is ringing
Regal's documentation is unusually candid here, and it presents the choice as a trade rather than hiding it. On power or preview dialling the agent is on the call from the moment it starts to dial, including ring time, and the docs list incurring consumption on ring time as a con.
Progressive dialling avoids that charge. The cost is that answering-machine detection gets three seconds to decide, which lowers accuracy, and the agent can no longer leave a personalised voicemail. So the buyer picks between paying for unanswered ringing and losing the voicemail that makes outbound work.
On the specific billing consequence of one configuration, the docs say to ask your customer success manager. That is an honest deferral and it is also the answer to how much of this is knowable in advance.
The throughput manufactures a payroll
The concurrency documentation contains the most useful piece of arithmetic Regal publishes, and no competitor publishes anything like it. Fifty concurrent AI calls, a five-minute average call, a forty percent transfer rate and a thirty-minute human close require 120 licensed human agents to absorb the handoffs.
That is the vendor's own formula and it checks out. It is also the real cost model: one AI agent running at its default ceiling generates a headcount requirement an order of magnitude larger than its own meter. Regal deserves credit for publishing it, and any buyer should run it before the pilot rather than after.
What we had wrong, and what the contract says
This site recorded Regal as able to hand off a payment mid-call over a Stripe link. That was wrong. The word Stripe appears nowhere across the 135 pages of Regal's developer documentation, no payment processor appears in its integrations directory, and there is no payment action among the roughly eighteen the platform publishes. The collections marketing references payment flows; the product ships no rail. The entry has been corrected to say so.
The subscription agreement is worth reading before the pricing page. Fees are non-cancellable and non-refundable, the term auto-renews unless cancelled thirty days out, overages are billed by name to a card on file at a rate published nowhere, and on termination the customer loses access to their data with no obligation on Regal to return it.
One clause deserves its own sentence. The agreement requires the customer to provide a case study and a review on G2, Capterra or another platform Regal designates, within the first renewal term.
G2 is the third-ranking result for this product's pricing query. Nine results, six with a stake, no independent page and no forum. The most prominent apparently-independent source is one the vendor's own contract obliges its customers to feed. That is the sharpest illustration this site has found of why the byline audit exists.
Setting Regal Up
Demo, order form, annual-shaped contract. No free tier, no trial, no self-serve path, and the pricing form asks for a mobile number before it will talk to you.
- Check the floor before anything else. Twenty-five agents and 100,000 minutes a month, both, and it moved twice in the last fortnight.
- Ask for the Regal Voice Pricing schedule that the order form binds you to. It is not on the site and the link 404s.
- Get the model surcharges quantified. Only GPT-4o Mini is stated as included; every other model is described as more expensive with no number attached.
- Decide the dialler mode with billing in mind, because power and preview dialling bill the agent through ring time and progressive dialling costs you voicemail quality.
- Run the 120-agent calculation from Regal's own concurrency doc against your actual headcount before you size the AI side.
- Read the publicity clause and decide whether a contractual obligation to post a public review is one you want.
What the AI actually does
1 of 4 can act without a person in the loop. Those are the ones to scope carefully.
Phone, SMS, chat and WebRTC agentsHolds the whole conversation with your customerNot publishedActs
This is the deepest end of the autonomy scale anywhere in this directory. Regal sells agents that answer and place calls, send texts and run chat — build once, deploy anywhere — against named jobs: lead qualification, appointment setting, collections, bookings and support.
Its own line for support is that the agents understand customer intent and take action. There is no draft-and-approve version of this product.
SimulationsTests the agent before it speaks to anyonePlatform featureScores
The most useful thing on the page and one almost nobody else in the category lists. If an agent is going to talk to your customers unattended, being able to rehearse it against scenarios is the difference between a deployment and a gamble.
Regal CopilotBuilds the agent for youPlatform featureDrafts
Sold as from idea to deployment, in days. An AI that builds the AI; the output is a configuration a human ships.
Branded caller ID and spam remediationKeeps your numbers from being labelledPlatform featureRoutes
Not AI, and listed here because it is the thing that decides whether any of the above works. This site has said repeatedly that aggressive dialing gets numbers flagged and that the flag destroys the answer rate you bought the tool for. Regal is one of the few vendors treating that as a product surface rather than the buyer's problem.
Regal is the clearest case on this site of a vendor whose entire product acts. Everywhere else the interesting question is which tier unlocks autonomy. Here there is no drafting mode to fall back to: the agent picks up the phone, and the buying decision is whether you are willing to put a synthetic voice in front of customers at all.
Strengths and Weaknesses
- Publishes a floor at all, which tells an unqualified buyer not to waste a sales cycle. That is more than most enterprise vendors here manage.
- The concurrency formula is published and lets you compute the human headcount the AI throughput will require, which nobody else in this category does.
- The dialler documentation states the ring-time billing trade openly rather than leaving it to be discovered on an invoice.
- Warm transfer is properly built: the agent holds the contact, waits for the human, passes a summary, then drops, with recording continuous throughout.
- Native telephony through Twilio means no separate carrier relationship to manage.
- A stated $0.20 average is more than most enterprise vendors disclose, even though it is not a rate.
- No rate card anywhere. The only figure is a realized average on a blog post.
- The floor computes to roughly $240,000 a year at the vendor's own numbers, and it is stated nowhere.
- The order form binds customers to a pricing schedule that returns 404.
- Three disclosures about the bill's structure were deleted from the pricing page since 2022, including the platform fee and the per-user charge.
- Model surcharges exist and are unquantified, as are concurrency pricing and the overage rate the contract bills by name.
- The contract requires customers to post a public review, on a platform that ranks third for this product's pricing query.
Regal Pricing
No rate card is published. The only per-minute figure Regal states anywhere is $0.20, on a blog post, described as what its agents generally come in at across millions of calls — a realized average rather than a price. Against the pricing page's floor of 100,000 minutes a month that computes to $20,000 a month, or $240,000 a year, which the vendor never states. Source: Regal pricing page and cost blog, checked 2026-08-06 (vendor-authored).
No price is published.We print Custom rather than an estimate, a competitor's number, or a range we inferred. The absence is itself information: the price is a negotiation, and you will not know yours until you are in one.
Nothing purchasable is published, so the pricing section is a list of what to extract in the first call. Six things, and every one of them is a number Regal has or had and does not show.
- The per-minute rate for your volume, and whether $0.20 is above or below it
- What the base minute includes, beyond the one language model named as included
- The surcharge for the model you actually intend to use
- The platform fee, which the 2022 page said was based on cumulative contacts
- Whether human agent seats are metered separately from AI minutes
- The overage rate, which the subscription agreement bills by name and prices nowhere
And ask whether a call that transfers from an AI agent to a human is metered on both. The transfer documentation covers cold and warm mechanics in detail and says nothing about billing, which on a product built around handoff is the single most consequential silence on the site.
The Verdict
A serious enterprise product with a genuinely useful published floor and the most honest concurrency arithmetic in the category, sold with no rate card, three meters of which one is disclosed, an order form binding customers to a pricing schedule that 404s, and a contract clause requiring them to post the public review that ranks third for its own pricing query.
- You clear 25 agents and 100,000 minutes a month, which is the real qualifier
- Your leads decay fast enough that automated speed to contact changes outcomes
- You have humans to absorb the transfers the concurrency maths predicts
- You can extract a rate card, the model surcharges and the overage rate before signing
- You are below the floor, where you are not the customer regardless of fit
- You need to model the cost before entering a sales process
- You want in-call payments, which this does not ship despite the marketing
- A contractual obligation to publish a review is one you would rather not carry
Can it take a payment?
Full tracker →Marketing pages for collections reference payment flows, but Regal publishes no payment rail, no processor integration and no payment action. The only route is a Custom Action posting to an endpoint the customer builds and hosts, which is not a capability the product ships.