Let’s rewind to last week. How many hours did you spend doing manual entry of data, from one medium to another? Whether that looked like reformatting slides, updating an Excel sheet, or working within the dungeons of your LMS.
Now think about how many of those tasks required any actual judgment. Any creativity. Any of the things that made you want to work in learning in the first place.
Our role in L&D has expanded. Most of us are simultaneously the administrator, the content creator, the LMS manager, the project coordinator, and the client-facing relationship person. With so many hats to juggle, the honest case for automation becomes simple: there are tasks on your plate that a computer can do exactly as well as you can, and doing them yourself is a waste of your time.
What automation actually is
The word gets mixed up with AI agents, machine learning, chatbots, and every other technology trend of the last several years. That confusion makes automation sound harder and scarier than it is.
An automation is just a set of rules that runs without you watching. Think of dominoes. Each piece falls and pushes the next. The first piece is the trigger, the event that starts the sequence. Every piece after that is an action. There’s no guessing, judgment, or improvisation. If A happens, do B. Then C. Then D.
An AI agent, by contrast, can reason through ambiguity. It makes decisions, interprets context, adapts to the unexpected. That’s useful in some situations, and genuinely complex to set up responsibly. A basic automation does none of that. It follows the exact steps you defined, in the exact order, every single time.
We like predictability. With automations, you know exactly what it will do, because you told it exactly what to do. For most of the administrative work L&D teams carry out, that’s exactly what you need.
When a task genuinely does need judgement in the loop, you are into agent territory. Anatomy of a lean AI agent breaks down what that actually involves.
How to know what to automate
Not every task is worth automating, and trying to automate the wrong things wastes more time than it saves.
Here’s a quick test: could you write the complete instructions for this task on one sticky note, hand it to a colleague, and expect them to do it correctly? This is a sticky note, so you really don’t have space for context, judgment calls, or “it depends.” If you can, then it’s probably a good candidate for automation.
Tasks that pass this test tend to share a few traits:
They happen often.
They follow the same steps each time.
The output is easy to check.
Getting it wrong is recoverable.
Tasks that fail the test, things that require reading a room, interpreting feedback, weighing trade-offs, making a creative call, those should stay with you. Automation can handle the mechanical work so you have the time and attention to do those things well. If you want to know more about delegating work to machines, check our human-AI delegation framework.
You don't need to learn a new platform to start
Most L&D professionals assume automation means signing up for Make.com, Zapier, or some tool they don’t have budget for and don’t have time to learn. Some automations do live there. But a lot of the work that’s weighing you down can be handled by tools you’re already using.
PowerPoint has a built-in macro editor. Google Sheets can run scripts. Your email client probably has filters and rules you’ve never touched. Moodle, Rise, Storyline, and most authoring tools have export functions that can be automated or scripted.
The ceiling for what you can build gets higher as you learn more. But the floor is accessible right now.
Three automations we've actually built
These aren’t hypothetical. They are things we built at Artha to solve our repetitive problems.
The bullet point formatting macro
One of our clients had a detailed style guide that governed how bullet points ended. Sometimes the rule was no punctuation at the end of any bullet. Other times it was commas after each bullet, with a period on the last one. The style guide was still in flux, internal sign-off was ongoing, and we had to be ready to switch between formats at any point in the project.
We were doing this by hand. On decks with 60-plus slides, that meant going through every bullet on every slide and manually checking or changing the ending character. It was slow and error-prone.
We built a PowerPoint macro that ran both versions on demand. One button applied “no punctuation” formatting across every bullet in the deck. Another applied the comma-then-period rule. When the client confirmed their preference, the deck was reformatted in about three seconds.
The macro took five minutes to build with vibe coding, and saved us for every slide deck we touched for that client afterward.
The certificate automation
For every cohort of our AI Accelerator program, participants who complete the program receive a digital certificate. Early on, we generated these by hand: open the template, enter the participant’s name, update the cohort details, save, export, and send. For a cohort of 20 people, that process took one to two hours.
We then set up an automation in Make.com. When a participant was marked complete, the automation pulled their information from a spreadsheet, populated the certificate template, sent the certificate to their email with a lovely completion email that we configured, and logged the send. The whole sequence ran without us touching it.
Two hours of manual work per cohort became zero. Participants received it immediately after completing the program.
The Moodle content extractor
A client came to us with a content problem that sounds simple until you try to solve it. They had years of course content built inside Moodle, their LMS. Their organization had changed direction, and they needed to review, update, and potentially repurpose all of it. But accessing the content was painful: navigating through Moodle’s interface, manually copying text from each section, and pasting it somewhere reviewable. Their SMEs couldn’t work in Moodle directly, and copying dozens of courses by hand was going to take weeks.
We built an extractor that automated the export process. It pulled course content out of Moodle, organized it by section, and produced a clean Word document for each course. The documents were structured so that SMEs could read and mark up the content directly, and we could search or index across all of it at once.
What would have been weeks of tedious extraction became a process that ran overnight.
Where to start
Look at your task list from last week. Find one item you did by hand that required no judgment, that followed the same steps it always follows, and that you’ll do again next week. Write the steps on a sticky note. That’s your first automation.
Before you add anything new, it is worth checking what you already pay for. Auditing your L&D AI tech stack includes a scorecard you can run tool by tool.
