AI Learning Blog: Prompting, Machine Learning & AI at Work

The five-part prompt we teach in week one, and the step almost everyone skips

Most first prompts fail for the same reason: they ask for an answer without saying who it's for. In week one of the masterclass we use a five-part structure (role, context, task, format, check), and the "check" part is the one learners skip. Here's the template, three before-and-after examples from real student work, and why asking the model to list its assumptions saves you a second round of edits.

AR

Ananya Raghavan
Lead instructor, Prompt Engineering Masterclass
9 Oct 2026, 8 min read

AI for beginners

Can you learn AI without coding? What our August beginners actually built

We asked learners in our August beginner batch whether they wrote any code during the course. Most didn't, and nearly all of them still built something they use every week: a meeting-notes summariser, a lesson planner for a Class 8 science teacher, a lead tracker for a small travel agency. This post walks through what they made, where they got stuck, and the exact point at which Python started to matter for the few who went further.

VI

Vikram Iyer
Head of learner success
6 Oct 2026, 6 min read

Machine learning

Train-test split, explained with a cricket scorecard

If you've ever wondered why a model that scores 98% in testing falls apart in real use, a cricket scorecard explains it better than a textbook. Picture judging a batter only on the innings you watched them practise in the nets. Using IPL match data, we show what a train-test split is, why you shuffle before splitting, how much data to hold back, and the one leak that quietly inflates beginner scores.

MK

Dr. Meera Krishnan
Machine learning lead, PhD in statistics
2 Oct 2026, 10 min read

Responsible AI

Why your AI summary got the numbers wrong, and a four-minute check that catches it

AI assistants are good at sounding sure. They're less good at copying a revenue figure from page 14 of a PDF. Going through learner assignments, we keep seeing the same three slips: rounded numbers, merged table columns and percentages that appear nowhere in the source. Our three-step check (ask for page references, spot-test two figures, compare the totals) takes about four minutes and catches most of them before they reach your manager.

RM

Rahul Menon
Responsible AI trainer, former data-protection consultant
29 Sep 2026, 7 min read
Fact-checked by Dr. Meera Krishnan

Automation

I automated our invoice emails with a no-code AI agent. Here's every step

Every month our finance team sent the same "your invoice is attached" email around sixty times. I rebuilt that job as a no-code agent: it picks up new invoices from a shared folder, drafts a polite email in the client's preferred language and waits for a person to approve it before anything goes out. Screenshots of every step are included, plus the two settings that stopped it from emailing the wrong client during testing.

SP

Sneha Pillai
Automation engineer, builds our AI Agents course projects
25 Sep 2026, 12 min read

Data skills

Cleaning a 40,000-row sales sheet with Excel and an AI assistant

A shop owner from Madurai sent us a sales sheet with 40,000 rows, three different date formats and "Chennai" spelt five different ways. Working with an AI assistant alongside plain Excel, we cleaned it in a single afternoon. The post shares the prompts we used, the formulas the assistant suggested, the one suggestion we threw out, and a simple way to confirm no rows were lost along the way.

SP

Sneha Pillai
Automation engineer
22 Sep 2026, 9 min read

AI at work

What to put in your company's AI usage policy (a one-page template for HR)

Banning AI tools rarely works; people just use them on their phones instead. The HR teams we train get better results from a one-page policy that answers four questions: which tools are approved, what data must never be pasted in, who reviews AI-written content before it's sent, and how staff report a mistake. Below is the template we hand out in workshops, with notes on adapting it for a 20-person company.

FS

Farah Siddiqui
Corporate training lead
18 Sep 2026, 8 min read
Reviewed by Rahul Menon

Machine learning

Overfitting, explained by the student who memorised the answer key

Every class has one student who memorises last year's answer key, tops the mock test, then struggles in the board exam. That's overfitting. Using a small house-price dataset from Coimbatore listings, we show what it looks like on a chart, why a more complex model isn't always a better one, and three practical fixes: more data, simpler features and cross-validation. No maths beyond Class 10 level.

MK

Dr. Meera Krishnan
Machine learning lead
15 Sep 2026, 6 min read

Generative AI

Brand-consistent social images from AI tools, without a full-time designer

Generating one nice image is easy. Generating thirty that look like they came from the same brand is the hard part. I share the style sheet I give image tools (colours, lighting, camera angle and a list of what to leave out) and show a full month of Instagram posts made with it. There's also a short section on licensing, and an honest note on when you should still pay a designer.

AR

Ananya Raghavan
Lead instructor, Prompt Engineering Masterclass
11 Sep 2026, 7 min read

Careers

Seven questions to ask before you pay for an AI certificate

Before you pay for an AI certificate, ask what it actually proves. Can you see a sample final project? Is the work reviewed by a person or auto-graded? Will a hiring manager recognise the name? I answer these questions from learners every week, and this checklist covers the seven I'd ask first, including the one most course pages quietly avoid answering. It applies to our own certificate too.

VI

Vikram Iyer
Head of learner success
8 Sep 2026, 5 min read

The TrainLearnAI blog

Latest articles

The five-part prompt we teach in week one, and the step almost everyone skips

Can you learn AI without coding? What our August beginners actually built

Train-test split, explained with a cricket scorecard

Why your AI summary got the numbers wrong, and a four-minute check that catches it

I automated our invoice emails with a no-code AI agent. Here's every step

Cleaning a 40,000-row sales sheet with Excel and an AI assistant

What to put in your company's AI usage policy (a one-page template for HR)

Overfitting, explained by the student who memorised the answer key

Brand-consistent social images from AI tools, without a full-time designer

Seven questions to ask before you pay for an AI certificate

Submit your article

Thanks, . Your article is in review.