Case study · 8 min read

AI Training for 11 Departments: Delhi Electric Company

How a hands-on AI at Work session on 17 October 2026 gave Delhi Electric Company's employees department-specific prompts, practice files and live demos for procurement, tendering, accounts, HR, sales and more. With slides and photos from the session.

By Surendra Pandar, AI Trainer ·

Surendra Pandar presenting to Delhi Electric Company employees in their office during the AI at Work training
Organisation
Delhi Electric Company
Sector
Electrical EPC and projects
Audience
Employees across 11 departments, in four groups
Date
17 October 2026
Format
On-site, hands-on session of about 3.5 hours
Topic
AI at Work: choosing tools, prompting, department labs, Copilot, safety

In short

  • One session, 11 departments, four groups. Each group worked on its own exercise card and fictional practice file, so every employee practised on work that looked like theirs.
  • The room saw the same prompt run live in Claude, ChatGPT, Gemini and Copilot, learned the CRAFT prompt formula, and then spent about 70 minutes on department labs.
  • Every employee left with a cheat sheet, a department prompt book and a written plan of three AI tasks to use that week.

The challenge

An EPC company has very different work in every department. Buyers compare vendor quotes. Tendering executives read 180-page documents. Engineers write specifications from rough notes. Accounts reconcile vendor statements. HR drafts policies. Sales writes proposals. A single generic AI talk helps none of them.

The brief was practical: give every employee something they can use on Monday morning, in their own department, with the tools they already have. Free versions of Claude, ChatGPT, Gemini and Copilot Chat were enough for every exercise.

Opening slide of the AI at Work session: the right tool, a prompt formula, and your own tasks
The promise at the start of the session: leave with the right tool, a prompt formula that works in every tool, and your own tasks tried on real work from your desk.

How the session was built

The deck is organised as seven short modules, with the speaker aiming for more typing than listening. The participants were not asked to believe anything. They were shown a story, a comparison, a formula, and then they did the work.

  • 01 A story you might recognise: Rahul, a fictional tendering executive, and his week with and without AI.
  • 02 Which AI for what: four tools, each best at something, with a live demo running one prompt in all four.
  • 03 Writing a good prompt: the CRAFT formula, and a ten-minute hands-on prompt.
  • 04 Your desk, your work: department labs in four groups.
  • 05 Copilot in Microsoft 365: a live demo of Outlook, Excel, Word and Teams.
  • 06 Make it a habit: save your best prompts and plan three tasks for the week.
  • 07 Smart and safe: where AI still gets it wrong, what never to paste, and who owns the output.
Group photo of Delhi Electric Company employees with trainer Surendra Pandar after the AI at Work session
Delhi Electric Company participants after the AI at Work session.

Opening with a story, not a tool

The session opens with Rahul. A 180-page tender lands on Monday. He reads it page by page, builds the eligibility checklist by hand, chases vendor emails on Thursday, and submits at 11 PM on Friday after almost missing a clause. The next slide replays the same week with AI: Claude produces the checklist, key dates and missing documents by 9:30 on Monday, Copilot summarises the vendor mails and drafts the compliance document in Word, and Rahul spends Tuesday to Friday on pricing strategy and client calls.

The slide that follows says it plainly: there is a Rahul in every department. Procurement compares three quotes in two minutes. Accounts reconcile in ten minutes, not half a day. HR gets a full job description in five minutes. Sales turns meeting notes into a follow-up mail instantly. Marketing plans a month of posts in one sitting. The hours chart in the deck is marked as illustrative numbers, which is how it was presented in the room.

Slide from the training showing Rahul's week with AI
“Same week, with AI”: same Rahul, same tender, same deadline.
Slide showing examples of AI time savings in procurement, accounts, HR, sales, store and marketing
“There is a Rahul in every department”: the bridge from one story to eleven departments.

Four tools, one rule

Rather than recommend a single tool, the session teaches a rule. Inside Outlook, Excel or Word: Copilot. Long document or careful writing: Claude. Image or social media post: ChatGPT. Fresh information from the web: Gemini.

Then the room sees it. The same prompt, a polite email to a cable vendor asking for a 5% discount on a 2 km order of 11kV XLPE cable in under 100 words, is run live in all four tools, and participants note the tone, length, detail and what each tool asks back.

Live demo slide: the same vendor discount email prompt run in Claude, ChatGPT, Gemini and Copilot
The live demo slide: same prompt, four answers. What to watch for: tone, length, detail, and what it asks you.

A longer walkthrough of each tool and the exact slide used in the room is in the guide: Which AI tool for which job.

CRAFT, and the two-prompt comparison

The deck frames AI as a brilliant new joinee on day one: smart and fast, and knowing nothing about your work. CRAFT is how you brief that joinee. Context, Role, Action, Format, Tone.

The comparison that lands best is the vendor quote example. “Compare these quotes” gives a generic answer you redo yourself. The CRAFT version names the company, the cable and the project, asks for a senior procurement manager's view on price, delivery, payment terms and warranty, and wants a table first, then three lines of recommendation. Participants then wrote their own CRAFT prompt for a real task from their desk in ten minutes, and asked one follow-up.

Slide showing a full CRAFT prompt for comparing three 11kV XLPE cable quotes
“CRAFT in action”: a complete prompt for the cable quotes example.
Slide comparing a one-line prompt and a CRAFT prompt
“Same task, two prompts”: what changes when the prompt carries context.

Department labs: four groups, 70 minutes

After a 15-minute tea break the room split by group. The format of the 70 minutes was fixed so that nobody waited: 20 minutes of short demos, 40 minutes of work on the department card and practice file, and 10 minutes to share one result. Each group got one live demo that matched its work.

Slide explaining how the 70-minute lab works: watch, do, share
The lab format: watch, do, share.
GroupDepartmentsLive demoTool used
A · Buying and StoresProcurement, Purchase, StoreCompare 3 vendor quotes from Excel into a table and a best-vendor pickClaude
B · Engineering and ProjectsTendering, Design Engineering, Project ManagementRead a tender in 15 minutes: checklist, key dates, documents neededClaude
C · Money and PeopleAccounts, HR, Front Desk and AdminReconcile a vendor account: ledger against vendor statement, mismatches flaggedCopilot in Excel
D · Sales and MarketingSales, MarketingA month of posts in one go from a short company briefChatGPT

Every group's practice files used fictional names and numbers: vendor quotes for 11kV cable, a PO delivery tracker, a stock register, a sample substation tender, a BOQ, vendor invoices, expense claims, an HR policy draft, sales meeting notes, a marketing brief and a pipeline sheet. Each department card listed two tasks and a bonus task, and ended with a “check before you trust” line, so the habit of verifying was built into the exercise.

What each department tried

DepartmentTasks on the card
ProcurementCompare vendor quotes, draft an RFQ, negotiation points
PurchaseFind late orders, delay follow-up email, rank vendors
StoreItems below reorder level, PO versus goods received, monthly stock summary
TenderingEligibility checklist and key dates, pre-bid queries, commercial risk scan
Design EngineeringDraft a panel specification from rough notes, check a BOQ, lessons-learned entry
Project ManagementWeekly client report, risk list, milestone plan
AccountsInvoices to a table, vendor reconciliation, expense summary
HRJob description, policy to circular (with a Hindi version), 30-day onboarding plan
Front Desk and AdminMeeting invite and agenda, visitor log summary, reply to a tough email
SalesMeeting notes to follow-up, proposal outline, competitor scan, forecast summary
Marketing30-day post calendar, LinkedIn post, a graphic for the post

Copilot in Microsoft 365, live

Because most employees already work in Outlook, Excel, Word and Teams, the Copilot module was shown live and tried later. The prompts on the screen were deliberately everyday: summarise a thread in five points with action items, draft a reply that is polite but firm on the delivery date, ask which vendor has the most late deliveries, add a column for days delayed, make a chart of delays by vendor, draft a compliance statement from the tender summary file, and recap a Teams meeting with owners and dates.

Slide: Copilot lives where you work, in Outlook, Excel, Word and Teams
“Copilot lives where you work”: Outlook, Excel, Word and Teams.
Live demo slide for Copilot in Outlook with three prompts
Outlook demo prompts used in the session.

Making it a habit, and making it safe

The session ends on the part the deck itself calls the one that matters most. Participants learn four ways AI still gets it wrong: it invents details, it can slip on calculations, it does not know the company unless told, and it may be out of date unless it searches the web. They get a before-you-paste list: no passwords, bank details or employee personal data; client drawings and prices stay out of personal accounts; company work goes in company accounts. And they are told that they own the output. Read it before sending, check every number, verify engineering values against IS standards.

Then comes the habit. Each person writes three AI tasks for the week, with the tool and the day, and saves their best CRAFT prompt in Claude Projects, ChatGPT Projects or GPTs, Gemini Gems or the Copilot Prompt Gallery. The deck's closing line for the day is the one people remember: AI is the assistant, you are still the engineer.

Slide listing what to keep out of AI tools and what is fine to use
“Before you paste”: the data rules taught in the safety module.
Hands-on slide asking participants to write down three AI tasks for the week
“My 3 AI tasks for this week”: tool, day and task, written down and photographed.
Slide summarising what participants take home: four tools, CRAFT, a prompt book and three tasks
What participants took home: four tools, CRAFT, a department prompt book and a plan for three tasks.

What the team received

  • A one-page cheat sheet: which AI for what, CRAFT, three lines that improve any answer, ten prompts anyone can use, and four rules before you paste.
  • A department prompt book with ready prompts for all 11 departments.
  • Exercise cards and fictional practice files grouped A to D.
  • A short feedback form, filled in by phone, so the next session can be improved.
Delhi Electric Company employees seated at their desks while the trainer presents
Participants following the live demo from their own desks.

Takeaway for other organisations

Training works when it is organised around the job, not the tool. A prompt book per department and practice files that mirror real work are what turn a one-time session into a weekly habit. If your organisation has several departments with different work, the same structure of story, tool rule, formula, department labs and safety carries over, and the files change to match yours.

If you would like a session like this for your team, book a short scoping call below.

“The AI training helped our employees use AI in everyday tasks, reduce repetitive work and improve productivity — a valuable step towards smarter workflows.”

Delhi Electric Company, Employee AI Training

Frequently asked questions

Can AI training be tailored to different departments in one company?

Yes. For Delhi Electric Company the session was split into four groups covering 11 departments, each with its own prompts, exercise card and practice files.

Does the training use our real company data?

No. Practice files use fictional names and numbers, and participants are taught what never to paste into an AI tool.

How long was the Delhi Electric Company training?

About 3.5 hours on site on 17 October 2026, including a 15-minute break and about 70 minutes of department labs.

Which AI tools did the employees use?

Claude, ChatGPT, Gemini and Microsoft Copilot. Free versions were enough for every exercise, and each department card named the tool that suited its tasks.

What did employees take away after the session?

A cheat sheet, a department prompt book with ready prompts, exercise cards and a written plan of three AI tasks to use in the following week.

Related reading