Docebo Shape : First impressions

Firstly, kudos to Docebo for giving everyone free trials of this new tool.

Secondly, kudos for a funny launch video:

What is “Shape”

Shape is the latest tool to offer AI auto conversion of content into learning content. This would appear to be going for the “do you really need an instructional designer for this” market. Obviously this is a debatable starting ground for a product, but so to is the starting point of “only instructional designers can create learning”, so hey ho. This seems to be entering some of the space of tools like Wildfire and perhaps the quiz area – like Quillionz which I have used a bit in the past.

My experiment

I recently needed to build a new learning module on an overhauled document. This doc effectively amounts to a policy or practice document with some specific “do this” points related to expected behaviours.

Therefore, I thought I would see what Docebo’s new AI tool can do with the raw content of the policy doc in comparison to what I came up with (in Articulate Rise 360).

When you upload content it goes through the below steps (after you say if you want a small, medium or large project):

Of these steps, the only manual intervention is to give the Shape (yes, each project/presentation is itself a “Shape”) a title. The system does auto suggest three titles but you can create your own.

The output

What you get is effectively a short video, the tool picks out key text and overlays that automatically over selected stock images with a selected audio track (about 15 tracks included and you can upload your own).

This can be previewed in browser (all I have done so far) or published elsewhere.

Concerns

One concern that should probably be held is what happens to the data, how much the AI is improving through saving anything that may be your copyright, etc.

There are some predictable issues with the AI – for example, use of “interest” in the context of ‘an interest in something’ leads to a background graphic around interest rates. A lot of the images are also stock image rubbish but that was probably predictable.

The stock images that are used as backgrounds vary in quality which is a little odd as you would have thought they would all be of similar size to avoid scaling issues, etc. I certainly saw one or two that looked pixelated.

Some of the background choices were not great for contrast and being able to see the text.

The music was very ‘meh’.

I found the default speed a little fast for reading but it does at least force a little concentration 😉

Overall, the model is questionable given the distraction of the transitions and images in relation to cognitive load and redundancy.

The good

The output looks mostly professional and is in line with modern short adverts, for example this kind of thing could easily be done in Shape (note images are included although you have to upload your own videos if you want to use them – at least in the free trial version):

You can edit the Shape to change colours, images, etc to deal with some of the issues I raise under concerns about contrast (although still probably not great for accessibility?).

Perhaps most importantly, the AI does a pretty good job of spotting the key elements from the source material although there was some weird stuff toward the end.

The “medium” solution I requested came back as just over 3 minutes which suggests this is going for decent “short and punchy” rather than trying to be too clever.

Overall

Is it worth it? Well, for basic advertisements this seems great, it would be an easy way to create content for campaigns but I’m not sure if micro learning itself in this format is hugely helpful. That said, if we compare this with what was possible a few years back then the ease with which we can now create content is hugely impressive.

Docebo have a track record of improving their products and I know they have some really good people on their team so hopefully Shape can become a useful tool to Docebo’s LXP customers and beyond.