Making a business video with AI: what AI does well, what it doesn't, and your options

The three families of ‘AI video’ tools

‘AI video’ covers very different tools. The question that matters is not which software to use, but who appears on screen.

Avatars and generated video. A synthetic presenter speaks your text, often in many languages, with no filming at all. That suits training and procedures, where the content matters more than the person.

Stock footage. Each sentence of your text gets an archive clip, music and a voice. Handy for recycling a blog post, less so for showing your product or your workshop.

AI-assisted editing. The footage is real: you film, and AI does the repetitive work. Editors such as Descript or CapCut belong here; so does Marqo, which also writes the script and publishes.

Trust: the real criterion for a local SME

A local SME often sells a relationship: the garage owner who knows your car, the accountant who picks up the phone, the restaurant owner who greets you. Your customers may well meet you in person, and a video showing the real person and the real premises prepares that meeting.

A good avatar is still an avatar: it cannot show your workshop or answer the question ‘who is behind this?’. Stock footage, for its part, shows offices and smiles that are not yours. For internal training an avatar may be enough; for the social media of a business that lives on its reputation, a real person remains the safest choice.

What AI does well today

• The first draft of the script: a clear structure (hook, two or three points, call to action) from your topic. It saves you the blank page, not the proofreading.
• Captions: a synchronised transcript within minutes, keywords highlighted. You still need to check names and figures.
• Repetitive editing: removing silences, assembling takes, lowering the music under speech, adding zooms. Time-consuming by hand, routine for AI.
• Translation and dubbing: text adapted, the filmed person's voice dubbed, captions translated. In Switzerland, one shoot can serve French, German and Italian (see the guide to multilingual video).

What still needs a person

• Expertise. AI knows neither your prices, nor your lead times, nor last week's customer story. Without those details, your script sounds like your competitor's.
• The face and the voice. Trust comes from the person who stands behind the business on camera. You don't need to be an actor: a teleprompter lets you read while looking into the lens.
• The final check. Names, figures, prices, technical terms, the meaning of translations: have every language you don't speak checked by a native speaker.
• Choosing the topics. AI suggests; you decide what matters to your customers.

Consent, voice and data: what to settle

• Consent from the people filmed. The image and voice of an identifiable person are personal data under the Swiss Federal Act on Data Protection (FADP), which requires you to inform the person concerned. Get written consent from your staff: which networks, for how long, and what happens if they leave.
• Dubbing and avatars. Dubbing someone means having them say their text in a language they never recorded it in; an avatar based on an employee can even ‘speak’ without them. Cover this in the agreement, and never clone a voice without consent.
• Labelling. YouTube requires creators to disclose realistic AI-generated or altered content, such as a real person made to say something they didn't, but not an AI-assisted script, captions or cloning one's own voice for a dub. Other networks have their own rules.
• Where the files go. Ask where the videos are hosted and which providers process the voice and the text. At Marqo: hosting in Switzerland with Infomaniak, dubbing using ElevenLabs technology, AI providers listed in the privacy policy.

Your first AI-assisted video in 7 steps

1. Pick a customer question: the one you hear most often at the counter or on the phone.
2. Have a 30- to 60-second first draft written, then add your facts: a price, a lead time, a real example.
3. Read it aloud and cut anything you wouldn't naturally say.
4. Prepare the shoot: smartphone held vertically, light from the front (a window), a quiet room and, if possible, a wireless microphone. Sound matters as much as the picture.
5. Film with a teleprompter, section by section; redo a take when you stumble.
6. Let AI edit, then check the names, figures, captions and pacing.
7. Publish and keep going: presence comes from regularity (see the guide to social media videos for SMEs).

What does it cost?

Prices for avatar and editing tools change often: check their pricing pages. Here are published Swiss benchmarks, and Marqo's prices.

These prices don't compare like with like: an agency delivers a film shot by professionals; with software you film yourself, and the shoot costs you time rather than an invoice. The detailed calculation is in the guide to business video costs.

Where Marqo fits

Marqo belongs to the third family. AI writes the script from your content strategy. You film a real person with the free Prompteur Marqo (iPad or Android tablet) and Camera Marqo (iPhone or Android) apps. Marqo then edits automatically: silences removed, takes assembled, zooms, 13 animated caption styles, AI-suggested cards, music that dips under speech, ambiences, split screens and an end card. The video can be dubbed into up to 32 languages (Pro and Business plans) and scheduled on Instagram, Facebook, LinkedIn, TikTok and YouTube (see the video studio in detail).

Its limits: no avatars, so someone has to be filmed; no paid ad campaign management; no publishing to X; German versions are in Standard German, not Swiss German.

Frequently asked questions