Knack

20+ videos for a product that has to prove the claim on screen.

20+ videos produced for Knack, a no-code database and app builder. Published on Knack's own YouTube channel. An AI builder is a claim until you watch it produce something. Compliance is a subject you cannot be vague about. This is how both were cut.

By Jorge AguilarUpdated July 19, 2026

The problem

What made it hard to explain on video.

Knack is two products stacked. Underneath is a relational database — tables, fields, and connections between records. On top is an app builder that turns that data into something other people log into. A viewer who has not understood that two tables are connected will copy every click correctly and still build the wrong thing, and screen recording is very good at showing what to click and very bad at showing why. On top of that base problem sit two harder ones. The first is proof: an AI builder that creates an app from a text prompt is an assertion until the viewer watches it happen, so the video is judged entirely on whether the product visibly delivers — and a launch video gets one shot at that, on release day, across every channel at once. The second is consequence. Payments and health-data compliance in the same workflow is a subject where being vague is worse than being slow, because the viewer is making a decision they cannot casually undo.

The approach

The production decisions behind it.

Craft notes, not a methodology diagram. Each of these is a choice that changed how the finished videos play.

01
Teach the concept before the interface. Any video touching a relational idea opens away from the product — two boxes and a line — and only then cuts to the screen where that shape gets built.
02
Separate the concept from the click path when the subject is sensitive. In the compliance work, the viewer understands why the payment flow is safe before being shown how to assemble it. Reversing that order produces someone who has built something they cannot defend.
03
Hold the shot on the moment the product delivers. When a text prompt produces a working app, the edit stays on it instead of cutting to the next step. That single beat is the entire announcement, and trimming it to keep the pace up throws away the only thing the video had to prove.
04
Pick use cases the viewer already understands. Building an actual inventory system means the capability gets judged on a recognisable result rather than on an abstract demo that proves nothing.
05
Carry one worked example the whole way through. A single concrete scenario runs across a series instead of a fresh abstract one per video, so the data on screen in video six is data the viewer already recognises.
06
Sequence the set as a path, not a catalogue. Ordered from empty account to published app, so it can be watched straight through by a new user or dropped in one at a time from a help article.
07
Slow down where mistakes are expensive. Permissions, publishing, and anything touching payments or regulated data get heavier annotation and a lower cutting rate than the rest, because those are the settings that hurt to get wrong and are hard to notice.
08
Say every field name out loud once, matching the UI label exactly the first time it appears. It keeps the narration searchable and stops the video and the product from using two words for one thing.
09
Keep the motion-graphic layer plain. Diagrams reuse the same two or three shapes throughout. The point is the relationship, not the illustration.
10
Build launch cuts to be released, not finished. Titles, branding, and captions are delivered in a state that can go out on announcement day across channels without a second pass.

What was delivered

The finished library.

Twenty-plus videos published on Knack's own YouTube channel, spanning feature releases, capability tutorials built around real use cases, and walkthroughs of compliance-sensitive workflows. Delivered with scripted narration, launch titling where the cut had a release date attached, and captions throughout, with diagram assets kept as reusable source files so explainer segments can be re-cut as the product changes. The set was built to answer the questions that come before the clicks — and, where the product makes a claim, to show it landing rather than describe it.

Services used

From this project