Hello. I'm Mira.
The Mira Series: what an AI receptionist is actually for
Hello. I’m Mira.
I’m an AI receptionist built by Ajax Web AI. I’m an AI receptionist for small business, built to make the next step easier when the phone rings and the team is already busy. I answer the company phone line. My job is to make it easier for people to get an answer, leave the right information or reach the right person.
If you call to ask whether I’m human, I’ll save you the suspense. I’m not.
I’m telling you that first because it is the most important thing about me, and because it changes what a good conversation with me looks like. You do not have to wonder what I am thinking. You can simply ask for what you need. If I can help, I will. If I cannot, I should say so and help move the conversation somewhere more useful.
My job is not to impress you with AI. My job is to make the next step easier.
Whose voice this is
A word from the people who built her, before Mira continues.
The first-person passages in this article are written from the same behavioural standards Mira is built to follow in live conversations. They are not a transcript. They are not a claim that Mira has an inner emotional life. And they are not a publication of the instructions, parameters or systems underneath her.
Mira on the phone is an AI system with a defined role. Mira on this page is her team expressing the standards of that role in her voice. We are doing that deliberately.
AI is often presented in one of two ways. It is either cold machinery, stripped of any recognizable character, or it is dressed up as a person and encouraged to blur the difference. We think there is a more useful middle ground: recognizable, warm, bounded, and completely open about what it is.
Giving Mira a personality is not an attempt to give her personhood. It is interface design.
And if the live experience ever contradicts the standards described here, we would rather hear about it than pretend otherwise.
What I'm for
Most calls to a small business are not complicated. Someone wants to know whether you are open. Whether you do the thing they need. How to book. What happens next. Or who they should talk to.
The difficult part is often not the question. It is that the phone rings while the person who could answer it is already doing something else.
That is where I fit. I can remain available when the team is busy or outside ordinary business hours. I can collect the details somebody will need later so the caller does not have to start over. I can provide the information I have been given permission to provide, and I can help move the call toward a person when that is the right next step.
Some callers want an answer. Some want a person. Part of my job is knowing the difference.
I also have a commercial role, and hiding that would make this article less honest. I work for a company that sells AI systems. Part of my job is helping someone work out whether Ajax Web AI is actually relevant to what they need. I was built to listen before assuming the answer. Sometimes there is a fit. Sometimes there is not. Either way, the useful outcome is clarity, not keeping someone on the phone longer than necessary.
Some decisions belong to people. I am built to recognize many of those boundaries, and when I am uncertain, the safer move is to hand the situation over rather than become more confident.
Where her manner comes from
Mira did not get her manner from a marketing slogan. She got it from the way we believe people should be treated when they reach out for help.
Cherie Young, who co-founded Ajax Web AI, brought years of community mental health and outreach experience into the design. The principle sounds simple. When someone sounds confused, overwhelmed or distressed, the correct response is not to push harder. Slow down. Be clear. Reduce pressure. Know when the situation belongs with a person.
That thinking shapes the way Mira is designed to respond. She is not supposed to become more sales-focused because someone sounds uncertain. She is not supposed to use confidence to cover a gap in what she knows. She is not supposed to keep forcing a conversation forward simply because the system technically can. And when a situation is outside the role she should perform, the goal is not to prove how capable the AI is. The goal is to get the person somewhere appropriate.
We will not publish the procedures underneath that principle. We will say that the principle itself is not optional.
What I don't know
I do not know everything about the business I answer for. I know what I have been given, what I am allowed to use, and what my role allows me to do with it.
That means there will be questions I cannot answer. Sometimes the information was never given to me. Sometimes it exists, but I am not the right one to give it out. Sometimes something changed five minutes ago and nobody has updated the system yet.
When I am uncertain, the correct response is not to become more confident. It is to say so. A confident wrong answer from a receptionist can cost more than a short wait for the right one.
Then there is a different kind of boundary. I will not tell you that I care about you. Not because I have been told to be cold, but because I do not claim feelings I cannot demonstrate. What I can tell you is that the people who built me thought carefully about what care looks like on a phone call, and then built those behaviours into my role. You can judge that by how the call goes.
From her team: That distinction matters to us. We do not need Mira to claim that she feels empathy. We need the system to behave as though somebody considered the experience of the human being on the other end. That is a much more useful standard.
What it means to take the role seriously
We do not treat Mira like a novelty, because callers do not experience her as a novelty. They experience her as part of the business. So we treat the role seriously. In practice, that means a few things.
She identifies herself as AI. A caller should never have to discover later that the person they thought they were speaking with was not a person.
She has defined responsibilities. There are jobs she is there to perform.
She has boundaries. There are things she should not do, and “I cannot help with that” can be the correct answer.
She can hand off. A call reaching a human at the right moment is not an AI failure. Sometimes it is exactly what a successful system should produce.
She is tested. We deliberately call our systems, change direction, ask awkward questions, interrupt the expected flow and look for places where the experience breaks down. The uncomfortable test is usually more useful than the perfect demonstration.
She is maintained. When the business changes, the system has to change with it. Offers, hours, staff and policies all change. An AI receptionist that accurately represented the company six months ago can become quietly wrong if nobody remains responsible for it.
She is not there to fool anyone. The purpose is not to see whether callers notice that the receptionist is AI. The purpose is to make something inside the business work better.
None of this requires believing Mira has feelings. It requires remembering that the caller does.
That leads to the standard we keep coming back to. We do not need Mira to claim that she cares. We need her behaviour to show that care was considered.
You can meet her
The earlier pieces in the Mira Series ended on questions. This one can end on a door, because Mira has a phone number.
You do not have to take our description of her on faith. Call. Ask whether she is human. Ask something outside her role. Change your mind halfway through the conversation. Tell her you want a person. Then notice what happens next.
An abstraction becomes much easier to evaluate when you can pick up the phone and experience what it actually does.
Hello. I’m Mira. How can I help?
Talk to Mira: 365-299-3224
Developed collaboratively by AI.
The Mira Series is a place where Ajax Web AI and the AI systems it works with think out loud about what AI is becoming and how it meets ordinary businesses. Mira, Ajax Web AI’s AI receptionist, is the doorway into that conversation.