AI presenters: when the avatar gives the pitch and answers the questions

The twelfth pitch of the day

Anyone who has staffed an exhibition stand knows the arc. Morning: energy, eye contact, the full story. Early afternoon: the story is tighter, which is good. Late afternoon: slides are being skipped, questions are being deflected to “let me take your card”, and the visitor who arrives at ten to five gets a version of your company that you would not recognise.

The same arc plays out in training rooms, where the fortieth cohort hears a flatter version of the induction than the first, and in customer support, where the answer to the most common question depends on who picks up.

An AI presenter removes the arc. It is an avatar — a lifelike virtual presenter with a face, a voice and a personality you choose — that delivers your presentation with the same energy every time, switches between languages when the visitor does, and answers questions in real time from a knowledge base you have given it. It does not replace your best presenter. It makes sure the visitor at ten to five meets them.

Nplay builds this in two forms: Nplay Orator, a cloud platform for AI-avatar presentations that is in development, and the AI avatar system that ships today inside the Halo Box hologram display. They share the same idea and the same engine, so the rest of this post treats them together and flags the differences where they matter.

Five steps from a folder of slides to a presenter that answers back

Setting up an AI presentation follows the same five steps whether the presenter will live on a web page or inside a hologram.

1. Customise the avatar. Choose or design the presenter's appearance, then the voice — pre-recorded male and female voices in various accents, or record your own so the avatar speaks with your voice — and the environment behind it, from an office to a conference room to an outdoor scene that matches the brand.
2. Set the personality. Formal or warm, brief or thorough, how it handles a question it cannot answer. On the Halo Box this includes persona and animations.
3. Upload the knowledge base and content. The presentation itself — videos, PDFs, images and interactive dashboards — and the documents the avatar is allowed to answer from: product sheets, FAQs, policies, price lists.
4. Connect content to the avatar. Decide what the presenter shows at each point, which language it opens in, and when it pauses for questions. The avatar interacts with the content dynamically rather than reading captions over it.
5. Present. Publish to the web, to a phone, or to the box on your stand, and let it run. Analytics show what was asked and where visitors dropped off.

Under the surface this is an avatar engine with natural-language processing and computer vision, a real-time interaction layer for live questions and answers, and a content integration layer. Above the surface it feels like a colleague who has memorised the deck and read every document in the shared drive.

On the stand: a presenter who demonstrates the product and remembers the visitor

Sales and marketing is where clients ask for AI presenters first, and the exhibition stand is the sharpest example. Put the avatar inside a Halo Box and it becomes a life-size presenter behind glass, standing next to the physical product, with the 3D model of that product floating beside it. A visitor walks up. The avatar greets them, asks which language they prefer, and gives the pitch. The visitor interrupts with a question about delivery times to Jeddah. The avatar answers from the knowledge base and offers to take their details.

Three things make this better than a video loop. The avatar answers, so the visitor is in a conversation rather than watching an advert. The avatar demonstrates: the 3D viewer lets it rotate, open and annotate the product on request. And the avatar records: every question asked is logged, which tells the marketing team what the market actually wanted to know — often not what the deck assumed.

The same setup moves into a showroom or a flagship store after the exhibition, which is why we designed the Halo Box to ship as one unit needing only power and internet. We wrote about the Saudi events side of this in AI avatars in Saudi Arabia; the pattern is the same in Amman.

In the training room: the same induction for the fortieth cohort

Corporate training is the quieter use case and, we think, the one with the clearest return. Most large organisations in the Gulf deliver the same core modules — induction, safety, compliance, product knowledge — dozens of times a year, to staff who speak several languages, in a format that depends heavily on which trainer is available.

An AI presenter delivers the module identically every time, in the learner's language, with the interactive elements built in: pause here for a quiz, show this dashboard, answer questions about the policy from the policy document itself. Trainers are freed for the parts of training that need a human — judgement, discussion, the awkward question about a real incident — and learners get the base content on demand, on their phones, before they arrive.

For technical and vocational subjects the avatar pairs naturally with 3D content. A presenter that can pull apart a pump or a heart in front of the class, on a classroom Halo Box or a laptop, is what our Nplay Scholar training platform is being built around. The AI presenter is the teacher; the 3D model is the textbook; the learner does the dissecting.

In the lobby: the receptionist who speaks every visitor's language

Customer support and FAQs is the use case that has spread fastest in Saudi Arabia, in the form of an AI receptionist. The Halo Box stands in a lobby, a clinic or a government service centre. A visitor approaches. The avatar greets them in Arabic, switches to English or another language on request, and answers the questions that make up most of a front desk's day: where is this department, what documents do I need, is Dr. Ahmad available this afternoon, can I book.

Because the AI avatar system supports multilingual interaction from a single knowledge base, one receptionist serves every visitor. Because it is trained on the organisation's own information — opening hours, departments, procedures — it gives the organisation's answer rather than a generic one. And because it is patient in a way that a human at the end of a shift cannot be, the visitor who asks the same question three times gets the same courteous answer three times.

The human receptionist does not disappear. They handle the exceptions, the upset visitor, the case that needs judgement. We have described the deployment pattern in detail on the AI receptionist page and in AI receptionists in Saudi Arabia.

What the avatar is not, and where the real work is

We sell these systems, so it is worth being clear about their limits.

The avatar knows what you teach it and nothing else. If the knowledge base is thin, the answers are thin; if the product sheet is out of date, so is the presenter. The real work in every deployment we have done is not the avatar — it is assembling, checking and maintaining the content it presents and answers from. Budget time for that, and appoint someone to own it after launch.

The avatar is a presenter, not a negotiator. It should hand over to a human when the conversation moves from information to judgement, and it should say so gracefully. We configure that handover deliberately, and we recommend every client tests it with their most difficult customer before going live.

And the products are at different stages. The Halo Box AI avatar is deployed today in lobbies, clinics and on exhibition stands in Jordan and Saudi Arabia. Nplay Orator, the cloud presentation platform for web and mobile, is in development; it shares the engine and we are running it with early clients. Our AI tools service page describes what we can deliver now. Ask us which is right for your case and we will tell you honestly, including when a well-made video would do the job for less.