Practical AI Use Cases: Small Workflows, Big Time Savings

Practical AI Use Cases: Small Workflows, Big Time Savings

AI is increasingly becoming useful not because it replaces human expertise, but because it can take care of many of the repetitive, time-consuming tasks that surround our high-value work.

Some of the most useful applications are surprisingly simple.

Here are a few practical AI workflows that I have personally found useful.

1. Codex to Organize a Folder

A common problem in clinical and academic work is the accumulation of poorly organized files—PDFs, reports, presentations, manuscripts, patient documents and reference material scattered across folders.

Codex can be used to inspect a folder and reorganize files according to rules that you provide.

For example, a folder containing hundreds of files could be organized into:

– Patient-related documents

– Clinical trial papers

– Guidelines

– Imaging/pathology reports

– Research articles

– Administrative documents

– Other miscellaneous files

The interesting part is that the organization does not necessarily have to depend on the filename. The files can be examined based on their actual contents.

I used Codex to organize my Downloads folder into sections. I simply gave it instructions and pointed it towards the folder.

I also asked it to create a “Probably Deletable” folder for files that appeared unnecessary. I then manually inspected that folder before deleting anything.

That last step is important: let AI organize; you make the final decision.

2. Codex to Add an Index to a PDF

Another surprisingly useful application is adding an index to a large PDF.

Imagine having a 300-page book, guideline or document without a useful index.

Instead of manually going through the document, Codex can analyse it and create an index containing:

– Major sections

– Important topics

– Chapters

– Keywords

– Corresponding page numbers

I recently used this for a book that did not originally have an index.

I uploaded the file to Codex and asked it to create one.

The result was a document that was much easier to navigate.

This can be particularly useful for large guidelines, clinical trial protocols, textbooks and other reference documents that you repeatedly need to consult.

3. Quick Take: Think About Enfortumab Vedotin in Oral Cavity Cancer

One of my favourite uses of AI is as a rapid evidence synthesizer.

For example, I might suddenly wonder:

«“Quick take: What is the current evidence for enfortumab vedotin in oral cavity cancers?”»

Rather than immediately doing a full literature review, I can use AI to get a rapid overview of the topic, identify the important concepts, understand why the question might be biologically interesting and identify areas that deserve deeper reading.

The value here is speed.

I can do this during a tea break or in between patients and get a quick grasp of a new topic.

Later in the evening, I can go back and ask AI:

«“Now teach me this topic in detail.”»

This creates an interesting learning workflow:

Curiosity → Quick Take → Identify gaps → Deep dive → Verify with literature

Of course, the AI-generated summary is a starting point, not a substitute for checking the primary literature.

4. Gemini to Summarize Voice Calls and Extract Action Points

Voice calls often contain a lot of useful information, but important points can easily get lost

A transcription and summarization workflow can convert a conversation into:

Key discussion points

– What was discussed

– Important decisions

– Issues raised

– Information that needs verification

Action points

– Task: What needs to be done

– Person: Who is responsible

– Deadline: When it needs to be completed

Instead of going back through a 30-minute conversation, you can start with a concise list of decisions and pending actions.

I have found this particularly useful for summarizing points after on-call ward rounds.

One important caveat: make sure the other person knows that the call is being recorded/transcribed and that this is appropriate under the applicable privacy, institutional and legal requirements.

5. Codex to De-identify PDFs

Another interesting application is automated de-identification.

Suppose you have a folder containing clinical PDFs with identifiers such as:

– Patient names

– Hospital numbers

– Dates of birth

– Addresses

– Phone numbers

– Other direct identifiers

A scripted Codex workflow can be used to identify potential identifiers and generate de-identified copies while retaining clinically relevant information.

For example:

Original

«Mr. ABC, UHID 123456, 54-year-old male, admitted on 12/04/2026…»

De-identified

«Patient-001, 54-year-old male, admitted in April 2026…

This could potentially make preparation of research datasets much faster.

However, this is one area where we should not blindly trust automation.

Identifiers can appear in unexpected places—headers, footers, tables, scanned images, filenames and even document metadata.

Therefore, a safer workflow is:

Automated de-identification → automated verification → human review → final dataset

And of course, patient/client data should only be processed using tools and workflows that are appropriate for the relevant privacy, institutional and regulatory requirements.

6. AI for Real-Life Problems

AI is not useful only for professional work.

Some of the most satisfying applications are mundane problems where you simply don’t know what to do next.

The leaking bidet

I once needed to replace a leaking bidet.

In a place like Mumbai, the combination of Zepto + ChatGPT can work surprisingly well.

ChatGPT helped me understand what I needed, take me through the replacement step-by-step and troubleshoot the initial problems.

Not exactly a groundbreaking medical application—but extremely useful.

Fogging while driving in Kerala

I am not used to driving frequently.

So when I had to drive on a rainy night in Kerala, I suddenly had a problem: the windows started fogging up.

I called my dad, but explaining the exact AC/ventilation settings over the phone was difficult.

I asked ChatGPT instead.

It immediately gave me step-by-step instructions on how to clear the fog and what settings to use to prevent it from recurring.

Again, nothing revolutionary.

But it solved a real problem at the exact moment I needed help.

The Bigger Picture

These examples illustrate what I think is one of the most useful ways to approach AI.

The most valuable applications may not always be the flashy ones.

They are often the small workflows that eliminate friction from everyday work:

Organize → Index → Summarize → Extract → De-identify → Learn → Troubleshoot

AI does not have to replace the expert.

It can simply take care of some of the digital housekeeping surrounding the expert’s work.

And that can give us more time for the things that actually require human judgment, experience and creativity.

What are some practical ways you have been using AI in your own work or everyday life? Feel free to share them in the comments.