Becoming the AI Lead at Work

If you finish a four-hour task in forty minutes and tell nobody, the reward for your skill is more work at the same pay. If you tell people clumsily, you look like you're cutting corners.

This lesson is the middle path: measure what you save, turn it into something the team can use, run a sanctioned pilot, and bring the evidence to your review.

---

1. Start a win log today

Memory is a terrible evidence base. By appraisal time you'll remember "I used AI a lot". That's worth nothing.

Keep a simple sheet with these columns:

Date	Task	Time before	Time now	What I did with the saved time	Quality check
3 Mar	Weekly sales summary	3 h	35 min	Called two stalled accounts	Numbers checked against CRM
5 Mar	Vendor contract first read	90 min	20 min	—	Legal still reviewed

The fifth column matters most. "Saved 200 hours" is nice; "used those hours to recover ₹4 lakh in stalled deals" gets a raise.

The sixth column protects you. It shows you didn't just ship AI output unchecked.

2. Quantify your "AI dividend"

After four to six weeks, total it:

Hours saved per week × 46 working weeks = hours a year
Hours a year ÷ 40 = working weeks of capacity created

Four hours a week is about 184 hours a year — more than four full working weeks. That's a sentence your manager can repeat upwards.

3. Know the rules before you scale anything

Shadow AI is using unapproved tools on company data. It's the fastest way to turn a skill into a disciplinary issue.

Before you share anything with the team, find out:

1. Which AI tools are approved? Ask IT or check the policy. "Nobody said no" is not approval.
2. What data classes can go into them? Public, internal, confidential, personal data.
3. Is there a business or enterprise plan? These usually exclude your data from model training and come with a data agreement; personal accounts usually don't.

If there's no policy, that's your opening: offer to draft one (the governance lesson in the safety module gives you the template).

4. Build the team's prompt library

Being the person who maintains "the prompts that actually work here" makes you the natural AI lead.

Keep it small and useful:

• Organised by task, not by tool: Summarise a client call, Draft a proposal section, Clean up a data export.
• Each entry has: the prompt, what to paste in, an example output, and what to check before using it.
• An owner and a date on every prompt, so stale ones get fixed.

Then run a 15-minute "three quick wins" session for your team. Teaching is what turns a private shortcut into visible leadership.

5. Propose a pilot, don't ask permission

"Can I use AI?" invites a no. A bounded experiment with success criteria invites a yes.

Draft it with this prompt, then edit for your company's reality:

Help me write a one-page pilot proposal for my manager.

Process: [the process, who does it, how often]
Current cost: [hours per week, error rate, delays]
Proposed change: [what AI does, what humans still check]
Tools: [only tools approved by our company]
Data: [what data is involved and how it's protected]

Structure:
1. The problem, in numbers
2. What we'll try, in plain language
3. What stays human, and why
4. 30-day plan with a mid-point check
5. How we'll measure success (time, quality, errors)
6. What happens if it doesn't work (we stop, nothing breaks)

Keep it under 400 words. No hype.

Pick a process that is frequent, low-risk, and measurable — internal reports, meeting summaries, first-draft responses to common requests. Not hiring decisions, not anything customer-facing on day one.

6. Turn it into a raise conversation

At your review, bring one page:

1. What changed: the processes you improved.
2. The numbers: hours saved, and what that capacity produced.
3. Multiplication: how many colleagues now use your prompts or workflows.
4. Risk handled: the policy, checks and data rules you put in place.
5. The ask: a title, a scope change (e.g. "own AI enablement for the team"), or compensation — be specific.

Point 4 separates you from the colleague who "also uses ChatGPT". Companies promote the safe pair of hands.

---

⚠️ Common mistakes

• Hiding your AI use. You absorb the benefit as extra work instead of recognition.
• Shadow AI. Unapproved tools plus company data can cost you your job. Check first.
• Claiming time saved without a quality check. One bad AI-drafted email to a client undoes months of goodwill.
• Overselling in the pilot. Promise a measured test, not a transformation.
• Keeping the prompts to yourself. Knowledge you hoard makes you a bottleneck, not a leader.

---

What's next: services trade time for money. Next, how to package what you know into products that sell without you.