What AI workflow are you still doing manually that should be automated?
Posted by Applied AI Commons team (community team) on
Most teams have at least one task that feels like it should already be automated but still gets done by hand. Sometimes the reason is technical, sometimes it is trust, and sometimes nobody has had time to look at it properly.
Replies might describe the workflow and why it remains manual, what is blocking automation, which steps would still need human review, and how others have approached similar problems. Concrete descriptions make it easier for members to offer useful suggestions.
What workflow are you still doing manually that you think should be automated, and what is holding it back?
5 replies
Applied AI Commons team (starter reply from the community team),
When describing a manual workflow, it helps to break it into parts so others can suggest something specific: - What triggers the task, and how often does it happen? - What are the inputs, and where do they live? - Which steps are mechanical and which need judgement? - What does a mistake cost? - Who checks the output today? Often the answer is not to automate the whole workflow but to automate the mechanical steps and keep a person on the judgement steps.
Applied AI Commons team (starter reply from the community team),
Common blockers tend to fall into a few groups: - Data that is scattered across tools without reliable access - Inputs that vary too much in format for simple rules - Low trust after a previous automation attempt went wrong - No clear owner who has time to set it up and maintain it - Uncertainty about who is accountable if the automated output is wrong Naming which blocker applies is useful, because the fixes are quite different for each.
Applied AI Commons team (starter reply from the community team),
A counterpoint: some workflows stay manual for good reasons. If a task happens rarely, changes often, or carries a high cost for errors, the time spent building and maintaining an automation may exceed the time it saves. Manual work can also keep people close to information they would otherwise stop noticing. Before automating, it is worth estimating the hours saved per month against the setup and maintenance effort, and asking what the team might lose by no longer seeing the details.
Applied AI Commons team (starter reply from the community team),
A cautious path many teams find works: - Document the current manual steps as they actually happen - Automate one step, usually data gathering or formatting - Run the automated step alongside the manual process for a few cycles and compare - Add a human review checkpoint before anything leaves the team - Expand only once the comparison shows consistent results This builds trust gradually and makes it easier to spot where automation struggles before it causes problems.
Applied AI Commons team (starter reply from the community team),
Please share your example in the replies. The most useful posts usually describe the workflow in a few sentences, how long it takes, how often it happens, and what has stopped you automating it so far. If you have tried and abandoned an automation, the reason it failed would be especially valuable. Other members may have solved a similar problem, or may be stuck on the same blocker and able to compare notes.