Where the binding language lives. The answers here describe how we work day to day. The terms that bind — fees, the trial and refund, data handling, and how either side ends the engagement — are in the agreement you sign and in our Terms of Service and Privacy Policy. If anything here conflicts with those, those govern.

Pricing & trial

The money questions.

How much does Candor AI cost?

The Customer-Facing AI Agent is $350/month — your approved questions and answers in English and Spanish, one messaging channel you choose, plus web chat on your website. Each additional messaging channel is $100/month. The Appointment Notification, Appointment Coordination, Financial Organizer, and Social Media agents are $150/month each. A Voice Agent is $450 to activate plus $200/month including 500 minutes, then $0.10 a minute. Bundles are available, and everything can be bought individually. Full details are on the pricing page.

What's the $150 deposit, and how does the trial work?

Every new client starts with a fifteen day trial. You pay a $150 up-front fee, we deploy the agent, and the clock starts. The agent's activity is visible to you in real time — every conversation it handles. At any point in those fifteen days, if it's not the fit you wanted, you tell us in writing and the up-front fee is refunded. No friction, no winback email three months later.

What if I want my deposit back?

You tell us in writing — email is fine. We do not argue, we do not ask you to fill out a form, we do not try to talk you into staying. If you terminate inside the fifteen day trial window, the $150 up-front fee is refunded. That refund term is in the agreement you sign, at Section 6(d).

Can I cancel after the trial?

Yes. There's no minimum term. If the trial works for you, monthly billing begins when the trial ends, and the agreement runs in thirty-day periods that renew automatically. You can cancel at any time with written notice. No contract fight, no minimum-term clause.

Do your rates change over time?

Every client pays the same published rates. There is no cohort pricing, no introductory tier, and no rate that varies by when you signed. We can raise fees no more than once a year, capped at the lesser of five percent or CPI, with thirty days written notice. That cap is in the agreement at Section 5(f).

Do you charge by message volume?

No. If your shop has a busy month, we don't charge you for it. If volume rises permanently and we need to tune the agent for that pattern, we tell you — and the tuning is covered by the monthly fee.

Is the Spanish-language setup a separate charge?

No. English-only and English-plus-Spanish cost the same. The second language is not a premium. It is who we are.

What about voice — phone calls?

Voice is an add-on: $450 to activate, plus $200/month including 500 voice minutes, then $0.10 a minute. The voice infrastructure is genuinely different from text, so that one we charge for, and we say so on the pricing page rather than burying it. Same agents, the same approved answers, and the same disclosure — the call opens by saying it is an automated system, naming Candor AI as the service provider, and telling the caller how to reach a human. Spoken at the start of the call rather than printed in a chat. We do not retain call audio.

The agent itself

What it does,
and what it won't.

Who actually builds my agent?

Alex Escobar — Candor AI's founder — builds each agent personally, using the company's internal build system. Not a freelancer, not an offshored team, not a wizard you fill out. The reason the client cap is forty is so that each agent gets this kind of attention.

What channels does the agent cover?

The Customer-Facing AI Agent includes one messaging channel you choose — SMS, WhatsApp, or web chat — plus web chat hosted on your website. Each additional messaging channel is $100/month, and every channel answers from the same approved content. Voice is a separate add-on with its own activation and monthly fee.

How long does it take to go live?

From the discovery call to a working agent: usually two to four weeks. That includes the discovery interview, build, internal testing, your own review, and the deployment to your channels. The fifteen day trial begins after the agent is live in your environment, not before.

Will it sound like me?

That's the work. During discovery we ask how you talk to customers, what you say to a regular versus a first-timer, what topics you handle warmly and what topics you stay short on. The agent is built from that — not from a template. You see drafts before it goes live, and you sign off.

What if the agent answers something wrong?

Two things limit how wrong it can get, and the first one is the way your agent is built. It does not write answers. It matches what a customer asked to the answers you approved, and when it isn't sure, it says so and hands the question to a person. So the worst case is one of your own approved answers given to a question it didn't quite fit — never an invented price, policy, or promise. The second is you: you have a direct line, and when something looks wrong you reach us and we answer. The monthly tuning pass closes the loop so the same mistake doesn't repeat.

In other words — it can pick the wrong one of your answers, but it cannot make one up.

What does the agent do when it doesn't know?

It says so, and it offers to route the question to a human. The agent is built to flag rather than guess — that's an architectural rule, not a preference. A customer talking to your agent will never get a confidently made-up answer about your pricing, your hours, your services, or anything else the agent isn't certain about.

Can I see what the agent is saying to my customers?

Yes — in real time. Every conversation is visible to you. During the trial this is how you make the decision about whether the agent is a fit. After the trial it's how monthly tuning works: we look at the conversations together, find the moments that could be sharper, and adjust.

Does the agent replace my staff?

No, and the company is built on that not being the pitch. The agent handles the after-hours questions, the repeat questions, the intake before the appointment, the follow-up after — the work that buries small businesses and that nobody enjoys. Your team still does the work that's actually about relationships. Technology in the middle, supporting — never replacing.

Security & your data

How we handle
what gets said.

The plain-language answers below describe how we operate. The documents with binding language are our Privacy Policy, our Terms of Service, and the Data Processing Agreement included in the client agreement you sign. If anything on this page conflicts with those documents, those documents govern.

Where does my data live?

On infrastructure operated in the United States. Conversations, the customer information your agent collects, and the tuning data all sit on US-based services. Traffic to and from the agent is encrypted in transit, and data at rest is encrypted where our infrastructure providers provide it.

The vendors we use and the retention windows are listed in the Data Processing Agreement included with the client agreement. Consent and opt-out records are kept five years; message content is kept thirty days.

Do you sell or share my customers' data?

No. We do not sell customer data and we do not share it with advertisers. We never reuse your answers, and we do not keep the content of your customers' messages beyond thirty days, apart from the consent record. We do keep track of what kinds of questions your industry gets asked — generic question wording only, never an answer you wrote, never a customer's actual message, and never anything that identifies a person. That commitment is in the agreement at Section 3(d).

What law governs how Candor AI handles data?

Candor AI is a California company, so we build to California's Consumer Privacy Act and Privacy Rights Act (CCPA / CPRA) as our baseline — whether or not your business is itself big enough to be covered by them. Most small businesses are not: those laws reach companies above certain revenue and data-volume thresholds. We don't ask you to work out which side of that line you're on. Where CCPA does reach you, our agreement puts Candor AI in the Service Provider role, which means we may handle your customers' information only to do the job you hired us for — not for anything of our own. For clients in regulated industries (tax preparation, accounting), the Gramm-Leach-Bliley Act applies in addition. For clients who handle protected health information, we don't take that work — Candor AI is not a HIPAA business associate.

In other words — we hold ourselves to California's privacy rules even if your business is too small to be required to.

What about CIPA — recording disclosure?

California's Invasion of Privacy Act requires disclosure when a conversation is recorded or logged. Every Candor AI agent — text and voice — opens by disclosing that it is an automated system and not a person, naming Candor AI as the service provider handling the message, and offering a way to opt out and reach a human. It's not optional, and it's not a checkbox. It's how the agent opens.

What insurance do you carry?

Candor AI carries Errors & Omissions (professional liability for technology services) and General Liability. Both are bound with licensed carriers through a California broker, and both are current. If a certificate matters while you're evaluating us — or for your own insurer or landlord — ask and we'll send it.

We do not yet hold a cyber policy. The market for AI-specific cyber coverage is thin right now, and we're still shopping it. We would rather tell you that plainly than name a policy we don't have.

In other words — the coverage that pays if we make a professional mistake is in place. The one written specifically for a data breach is not, yet, and you should know that before you sign anything.

What happens if there's a data incident?

We have an internal incident response plan that covers detection, containment, client notification, and post-incident review. California law (and the CCPA / CPRA notification requirements) governs the timing and form of notification — those obligations are not optional, and we follow them.

Your commitment from us is already written, in the Data Processing Agreement that comes with your contract: we notify you without undue delay after we become aware of a security incident affecting your customers' information, and we give you the information we have so you can meet your own obligations.

In other words — if something happens, you hear it from us quickly and in writing, not from one of your customers.

Can Candor AI refuse to take a client?

Yes. Candor AI maintains a written framework for when we will and won't take on a prospect — based on whether we can serve them well, whether the business is one we have the expertise to support, and whether the engagement is consistent with how we run the company. Refusals are not based on protected characteristics; they're based on fit.

We say no in writing, and we say why. Nothing about a refusal is hidden behind a policy you cannot read.

Where are the Terms of Service and Privacy Policy?

Both are published on this site: our Terms of Service and our Privacy Policy. They are also available in Spanish. We will not ask you to sign an agreement we haven't already shared with you in full.

About Candor AI

Who we are.

Who runs Candor AI?

Alex Escobar — founder, builder, and the person who answers the phone. Born in Guatemala, raised in San Bernardino, sixteen years in Pomona, casino dealer by trade. The clients Candor AI serves are small business owners in the community Alex grew up in. There's more on the about page.

Why a forty-client cap?

Because doing this right takes time per client that doesn't scale. Forty clients is roughly the number that one builder can know well — well enough to remember the trade, the regulars, the voice, the edges. Past that, the work becomes generic. Generic is what Candor AI is built against. The cap is the feature.

Why only English and Spanish?

Because we will not ship a language we cannot stand behind. English and Spanish are the two languages Alex and the community around him speak fluently — and the Spanish in every agent is reviewed by people from the communities your clients come from before it goes live. Other languages will come when we can hold them to the same standard. Not before.

Where is Candor AI based?

Pomona, California. We serve small businesses across Southern California — that's the geographic scope right now. We are not taking clients outside that region yet.

What if my trade isn't on your list?

Tell us anyway. Some trades fit our current build system right away; others need scaffolding we're still finishing; a few are outside what we're set up to support honestly today. Either way, the answer is straight — and if we can't take it on yet, we'll say why and what would change that. There's a note on the industries page about how this works.

Are you hiring?

Not yet. The company is early enough that hiring will come from the relationships built with our first clients. If that changes, it'll be said here first.

How is “candor” not just marketing?

It's the architecture, not the slogan. The agent is built to flag rather than guess. The trial is refundable without argument. The pricing has no hidden surcharges and we wrote down the line items that don't exist. The client cap is honored. When the lawyer hasn't reviewed something, we say so on this page. Candor only counts when it costs us something.

Still don't see it?

Ask us directly.

This page covers the questions we get most. It does not cover every question — and the questions that are specific to your trade are usually the most useful ones to ask out loud. Send us a note. We will reply personally, and we will tell you honestly whether we're a fit before we ask you to do anything.

Send us a question
Ready to talk?

We're taking new clients now.

Send us a note. Tell us what you do, and which questions your customers ask you over and over. We will reply personally — and tell you honestly whether we are a fit. The trial is real; the up-front fee is refunded if you leave inside the window; nothing is locked in until you decide it is.

Email us