“Schedule your social posts and streamline the publishing” they said, “just record, we’ll take care of the rest” they said…
What you end up instead is not-quite getting the whole promise. But hey 80% is done! Yes, 80% of the feature is done, and look it helps. And I don’t mind doing the 20% myself in ONE tool or ONE task. But when those 20% compound, it gets exhausting. Not just doing those 20%, but remembering all the little nowhere mentioned 20% steps. So my ‘automated’ pipeline ends up having 11 manual steps over 4 tools.
Concrete example - buffer.com promises to publish your YouTube shorts automatically. But there are some important YouTube shorts features, like linking the original video, that are missing in the integration - so now I have to do those parts manually.
On the generation side, Riverside can generate the shorts, but when it comes to caption-generation with the emphasis I want, or nice overlays and b-roll flexibility - I need to do that myself. Or find another tool, like Submagic, that does THAT one thing well, oh wait, except it’s again 80% - because producing shorts there screwed up the audio and I still had to do a round manually after creating them there.
So a “simple” task of short generation becomes - > open one tool, do things, (get used to things the tool changed), move to another tool, do some more things there (consider all the nuances of that tool), export and schedule the short, but don’t quite let the tool publish it because before publishing it you have to do some more steps.
And so the promise of “well, just spend a couple of hours creating shorts, schedule and forget about them” is a lie. “Just record and we’ll take care of the rest” is a lie. The products are not done to 100%, the integrations are not done to 100% and it’s never the 100% I needed.
And so I have to keep a mental(or not) checklist of all the things I have to do manually. And before you say - well you could use AI to automate that last 20%, that last 20 is usually the hard part, and hence not done by the tool themselves, so I’d spend 80% of MY time, they(the product creators) didn’t spend doing that. sigh
The tools are brittle, each tool has a WAI/WAD gap, and I, the user of those tools, become the one who provides resilience.
The worst part is remembering these little but important steps and checklists.
On top of this, every tool changes, and with that every integration changes. So if YouTube releases a feature tomorrow, buffer will integrate it whenever, if ever, and now I have to keep track of that as well.
This is my individual level. You know this repeats in every organization on every team level. This is how we end up with “tabs and tabs of tools” and then a new tool comes and promised to “replace all the tools with one” - but ends up just being yet-another tool. Or an “integration platform sooooo flexible, it’ll allow you to do ANYTHING and integrate ANYTHING” - sure.
So what’s the solution?..
I don’t have one so I’m asking for collective wisdom here. But my thoughts are following:
Let’s first separate each problem, because I think there are several - some of that 20% is pure “toil” - the seams of integration that I have to account for and some of those 20% are exactly what you have to do to stand out. Because the more these tools do, the more people can use them and so if you don’t do those 20%, you’ll just end up doing what everyone else is - average.
The tool itself is 80% done - there are some features present, some absent
find another tool that covers more of your needs (usually this means a “lower level” tool - e.g. riverside → DaVinci (a more professional editing tool))
find another tool that only does the absent part - and deal with the seams (again another 20% from integration)
The integration is 80% done - since an API or an easy integration is missing the tool couldn’t effectively integrate with another
Integrate it yourself - BUT, if it would have been easy, the creators would do it so you’ll have to invest 80% of your time into this little seam fix
The differentiating 20% - when everyone has access to those features, you stand out by adding your “20% spice” into the mix, otherwise your content starts looking just as average as the rest
no automation here, cause that’s the YOU part, that’s what makes you stand out and this is the only thing you want to have a say in. But KNOWING where this 20% is, is exactly the checklist issue
The checklists: the need to hold the manual steps, or steps in-between, or the facts about limitations in your head and them constantly changing with the changes in the tools or your needs
The checklist itself is the consequence of the three above, so theoretically, if we’d solve the top 3, no checklists would be needed. But you know that’s not going to happen. The WAI/WAD gap is never going to be closed.
A common set of advice is
Stop keeping the checklist in your head. Fine. I’ve written it down. And then the checklist changes every time - what’s the point putting it on paper anyway? And yes, especially in the social media, I’ve noticed it does change quite often.
Less tools, less headache. Well, yes, technically, but if I don’t use the tools, I’d need to not only do the 20%, but actually the 80% they did solve - so a learning curve of more advanced tools, or building it from scratch - no, thanks.
AI will help?
When it comes to the “seams” and closing that 20% which is not there, because of missing APIs, I hate to admit this, but, what, the only way is the “computer use” and UIPath-like automations and playwright? It’s ugly. And damn it, it’s gonna add another 20% of toil because of all the exceptions, and changes in the UI.
And you know what?
With current AI-assisted development, and tools and products created using AI - there will be more and more of these 20%s, and us, humans, would just need to handle MORE, not less…
—
I want a system that manages this all. I want a system that closes these little gaps. The one that orchestrates the rest and adjusts to their changes. The one that knows which one of those 20% is toil and which ones are “taste” and pulls me in on taste, and resolves the toil on it’s own.
I want a system/tool that only focuses on the exceptions and on the learning from changes. One that has adaptability and flexibility towards change. Any agent needs to be given “circumstances” (in this case - YouTube has feature X in state Y, and buffer integrated it in state Z) and automates a task in those circumstances. When circumstances change, this should be detected by the system/agents and it should re-evaluate whether the previous checklist makes sense or not and what needs to be done to accommodate this change.
I want my agents resilient and learning. Is it feasible?