Use case · Founder Video for AI startups
Founder video production for AI startups
Founder video for an AI startup has one job that general video production doesn't: to show that the product works on real inputs, and to explain how you know it works.
I research the topic, prepare the questions, direct a remote recording, and edit the video with real product runs, screen shares and simple diagrams. You check every technical claim before anything is published.
The general service, with full scope, process and deliverables, is on Founder Video Production. This page is about what changes when the product is AI.
Who this is for
- Founders of AI, ML, agent, LLM-infrastructure and developer-tool companies
- Technical founders still close to sales, who explain the product better on a call than the website does
- Teams whose buyers are engineers, data leads or technical evaluators who want proof before they book a demo
- Companies past launch day, with real users and real deployments to talk about
Why AI products are harder to explain on video
The output changes between runs. A model can give a different answer to the same input tomorrow. A video that only shows the best run looks staged to anyone who has used the product, and a video that ignores variation raises the first question a technical buyer asks.
Buyers are wary of claims. Every company now says it uses AI. Technical buyers assume a demo is cherry-picked until they see otherwise, so "it's accurate" means nothing without how you measure it.
The interesting part is invisible. The decisions that make your product better, like how you evaluate outputs, handle edge cases or keep data safe, happen in code and pipelines. Someone has to turn them into something a buyer can see. More on that in how technical founders explain a complex product publicly.
The stack moves fast. Model names, providers and benchmarks change every few months. A video that names them without a date gets old quickly and can end up saying something that is no longer true.
Security questions come early. Where data goes, what is stored and who can see it often decide a deal before features do.
All five are solved in preparation and in what you choose to show on screen.
Six videos that work for AI startups
| Video | The buyer question it answers | What it has to show | Where it gets used |
|---|---|---|---|
| Real input, real output | "Does it actually work on something like my data?" | An unedited run on a realistic input, with the output on screen. If outputs vary, show two runs | Homepage, product page, sales follow-ups |
| How we measure quality | "How do you know it's right?" | The evaluation method in plain words: what you test, how often, what a failure looks like | Product page, technical evaluators, AI-answer pages |
| A decision we made | "Why did you build it this way?" | One architecture or product decision, the alternative you rejected, and what it cost you | YouTube, LinkedIn, comparison pages |
| Where it breaks | "What happens at the edges?" | One known limitation and how the product handles it, or when it is the wrong tool | Sales calls, docs, FAQ |
| The security answer | "Can we get this through review?" | Where data goes, what is stored, which controls exist, said by the founder | Sales follow-ups, security pages |
| The category explainer | "How should I think about this space?" | The founder's view of the problem and the approaches to it, including ones that aren't yours | YouTube, search pages, newsletters |
Most AI companies need the first two before anything else. A founder explaining a decision is more useful once the buyer has seen the product work. Whether YouTube itself is worth the effort is covered in is YouTube worth it for B2B SaaS founders.
How I propose to keep the technical claims accurate
This is the approach I propose for AI startup projects. The exact steps are agreed in the scope before recording.
- A claims list before recording. Every number or claim we plan to say goes on a list with its source: an eval result, a customer metric, a benchmark you ran. If there is no source, we don't say it.
- Real runs, labelled. Demos are recorded on the real product. If a segment is sped up, cut, or re-run, the video says so on screen.
- Dates on fast-changing facts. Model names, providers and benchmark results get a date in the video or the description.
- Your technical review. You, or the engineer you choose, review the edit for accuracy before anything is published. I fix what you mark.
- Raw files kept. You keep the raw recordings, so a claim can always be checked against what was said.
My job is to ask the question a technical buyer would ask next, keep asking for the example, and make the answer clear. You are the one who confirms it is correct.
What one recording session can give you
A remote session of 60 to 90 minutes, with questions sent a few days before, can often produce more than one video when the plan allows it. For an AI startup, a common shape is one real-input walkthrough, one evaluation explainer and a few short clips cut from both, plus a written page built from the same recording. Your time is usually 2 to 2.5 hours per recording batch, including the brief and the review.
Founder video or a launch video?
They are different products, and many AI startups need both at different times.
| Launch video | Founder video | |
|---|---|---|
| Job | Get attention on launch day | Help a buyer understand and trust the product before a call |
| Length | Usually 30 to 90 seconds | 2 to 5 minutes for explainers, longer on YouTube |
| Who is on screen | Often animation, product UI or a short founder clip | The founder, with screen shares, diagrams and real runs |
| Made by | Launch-video studios, often with scripts and voiceover | A producer who interviews the founder and edits for clarity |
| Shelf life | The launch week | As long as the product works that way |
If you need a polished animated piece for Product Hunt next week, a launch studio is the better choice. If buyers keep asking the same questions on calls, founder video is.
What this is not
- Not animated launch ads, complex animation or motion-graphics-heavy production
- Not studio or on-location filming
- Not paid promotion, and no guaranteed views, subscribers or leads
Price and ways to buy.
Each project gets a fixed quote, set by the number and length of videos, the clips and the revision rounds. There is no published per-video price.
Try it first · 21 days
Founder Video Sprint
€750 / 21 days
One edited founder video of up to about 3 minutes with captions, one written companion piece of up to about 600 words, and one LinkedIn post. Founder time is about 3 hours in total.
Standalone project
Fixed quote
For example, a real-input walkthrough and an evaluation explainer for a launch, with clips.
Over 90 days
Founder Presence
€3,000 / 90 days
Short founder videos can be part of the agreed production scope, alongside written content. This is an engagement price, not a video price.
Market prices for founder video are compared in how much founder video production costs for B2B SaaS.
Proof.
Questions AI founders ask.
What if the demo fails during recording?
The approach I propose is to treat it as useful material: show it and explain what happened, or re-run it and say on screen that it was re-run, without presenting a cut as one clean run.
Do we need a script?
No. You get questions a few days before, and the conversation follows them. Founders explain AI products better when they answer real questions than when they read.
How do you handle accuracy numbers?
In the proposed approach, only with a source you can show: your own eval results, a customer metric you are allowed to share, or a published benchmark. Each one goes on the claims list before recording.
Our product changes every month. Will the video date?
Parts of it will. That's why model names and benchmarks get dates, and why the most useful videos explain decisions and methods, which change more slowly than features.
Who owns the footage?
You do. You receive the raw recordings with the final files.
The fit call · 30 minutes
30 minutes. We look at what you need and what you already have.
Then I tell you whether a project, the Sprint or a 90-day engagement fits, and you get a fixed quote before committing.