Productivity: automating your repetitive tasks

You resize the same photo to the same dimensions three times a week. You clean up the same address list after every export. You stick the same three PDFs together before every client email. None of these jobs is hard — that is precisely the problem: they are too ordinary to be noticed, and frequent enough to eat up half a day a month without leaving the faintest trace in your diary.

The rule of three: spotting what is worth the effort

Before you go looking for a tool, time the job. A four-minute task done twice a week costs close to seven hours a year. The same task at fifteen minutes, once every working day, comes to more than sixty hours. The arithmetic decides on its own what is worth lingering over.

After that, two questions are enough to settle it:

• Is the task deterministic? Same inputs, same settings, same expected result. If you are making a judgement call every time (“this one wants a tighter crop”), the task cannot be automated — it can be standardised, which is not the same thing.
• How many items at a time? Two files once a week do not justify writing a single line of code. Two hundred files every morning do.

Between those two lies the territory where most office work actually happens: too frequent to be done on autopilot, too variable to deserve a script. That is where the gain sits, and it has less to do with speed of execution than with getting the same result every time.

Three families of task, three treatments

Files that always have to look the same

Visuals for a website, attachments to a quote, deliverables handed to a client: the aim is not to go fast, it is to apply exactly the same settings on every pass.

Take the images on a blog. A JPEG straight out of a camera is commonly 4000 × 3000 px and 5 to 8 MB. Brought down to 1600 px wide at a JPEG quality of 80, it lands somewhere between 250 and 400 KB, with no visible difference on screen. That is a setting you fix once and never argue about again; running the file through an image compression tool then becomes purely mechanical. For recurring PDF bundles — quotes, appendices, terms and conditions — the difficulty is not the handling but the order of the parts: merging the PDFs in a sequence written down in black and white saves you re-reading the whole thing before every send.

Text data you copy across

CSV exports, recipient lists, inventories pasted from a spreadsheet: this data almost always arrives dirty. Duplicates, blank rows, inconsistent case, trailing spaces. Cleaning 800 rows by hand takes twenty minutes; putting them through a duplicate removal tool takes ten — provided you harmonise the case first, otherwise Smith@… and smith@… count as two separate entries and you will send the message twice.

The same logic applies on the publishing side. Rather than making up a web address by eye for every page, a slug generator applies the same transformation rule every time: accents stripped, apostrophes removed, spaces turned into hyphens, length kept under control. You end up with URLs that stay consistent over five years, which no amount of manual typing will guarantee.

The sums that come round every month

Pulling the VAT out of a gross figure, converting an invoice into a foreign currency, reading a Unix timestamp out of a log, fixing a meeting across three time zones: all ten-second operations that you get wrong one time in five when you do them in your head. A VAT calculator, a currency converter or a Unix timestamp converter save no time at all — they remove the arithmetic mistake, which normally costs a great deal more.

Write the recipe, not just the task

Genuinely automating an office task is not about the tool: it is about the settings sheet. As long as the values live in your head, nobody can take over, and you yourself will drift within three months. A sheet fits in five lines.

Keep that sheet next to the working folder, not in a notebook. The day somebody else picks up the workflow, they produce the same result without asking you a thing — and it is also what lets you compare two batches six months apart without wondering what changed in between.

The point at which a real script pays for itself

A web tool handles a one-off batch perfectly well, from any machine, including one you are not the administrator of. It hits its limit the moment the operation has to fire without you.

Move to a script when at least two of these conditions hold:

• more than a hundred files per pass;
• a daily rhythm, or an event-driven trigger (a file dropped into a watched folder, an incoming file);
• settings that have been strictly frozen for several months;
• a chain of operations run back to back: resize, rename, upload, notify.

The building blocks are well known and free: ffmpeg for audio and video, ImageMagick for images, PowerShell or a shell script for bulk renaming, the Shortcuts app on macOS or Task Scheduler on Windows for the trigger. Budget half a day of writing and debugging for a simple script, then maintenance every time a format, a path or a version changes. Below roughly twenty hours saved a year, the balance rarely tips the right way.

The traps that cancel out the gain

• Re-compressing a file that is already compressed. Every fresh JPEG pass degrades the image a little further, and it is irreversible. Always start again from the original, never from the published version.
• Converting a PNG to JPEG without thinking. The transparency disappears and the background turns white or black. A cut-out logo stays as PNG or WebP.
• Treating a scanned PDF as a text PDF. Converting it to an editable document will hand back nothing but a picture of the page, without a single selectable character. Run optical character recognition on it first.
• Forgetting the fonts. A document exported to PDF without embedded fonts displays perfectly on your own machine and shifts on the recipient's. Open the result in a different reader before you send it.
• iPhone photos. The HEIC format is still refused by plenty of content management systems and web forms. Plan the conversion in advance rather than in a panic, and check the orientation: it depends on an EXIF metadata tag that some conversions lose along the way.
• Invisible duplicates. A non-breaking space, an initial capital, a missing accent: two rows that look identical are not identical to the machine. Normalise before you de-duplicate, or the clean-up simply gives you false confidence.

A workflow that stands up without you

Three habits are enough to make a routine repeatable:

1. A bookmark folder called “Routines”, holding the five or six tools in the workflow, arranged in the order you use them. Hunting for the tool often costs more time than using it.
2. Two directories, always: “source” and “delivered”. The original must never be overwritten by its processed version, or the first mistake becomes permanent.
3. One recipe file, a sheet per task, updated the same day you change a setting — not three weeks later.

One practical point before you set all this up. The tools that run inside the browser — image compression and resizing, de-duplication, slug generation, calculations, formatting — work on your own machine, so the file itself is never sent anywhere. Image compression and resizing do ask for an account and belong to the subscription; the others are open to anyone. The ones that need a server, such as PDF handling, video or HEIC conversion, give you one conversion a day without signing up; beyond that, the €7-a-year subscription lifts the cap. Work out which of the two categories your routine falls into before you formalise it: it changes the way you will group your batches. And wherever a tool takes a file at all, its own page states the maximum size accepted — read it there rather than trusting a figure from memory.