Your AI sounds generic.

In June 2025, Shopify’s CEO Tobi Lutke tweeted that prompt engineering was dead. The post got 1.9 million views. By 2026, 82% of IT leaders said prompts alone were insufficient.

He was not saying AI stopped working. He was saying the skill had evolved.

Prompt engineering was the skill of writing a clever question and hoping AI understood what you meant. Context engineering is the skill of giving AI everything it needs before it answers. The shift sounds subtle. The output difference is not.

As someone who teaches AI on LinkedIn Learning, I can tell you what the gap is: you’re giving the AI one of five fields and expecting it to supply the other four. In this issue I’m handing you the full template.

five puzzle pieces snapping together

The Concept Shift

Prompt engineering is writing a clever question.

Context engineering is building the right environment so the question gets answered well.

The difference in output is not marginal; it’s the gap between something you rewrite and something you send. Harvard Business Review named AI-generated-but-generic content “workslop” in September 2025. MIT found that 95% of enterprise AI pilots produce zero measurable ROI, and the primary culprit is missing context, not a bad model.

The 5-Field Template | Works in Claude or ChatGPT

Before your next serious AI request, fill in all five fields:

Role: Who you are and what you're working on.
(Example: "I'm a director preparing a mid-year review for a high-potential analyst.")

Goal: What "done" looks like — specifically.
(Not "write a summary." Write: "I need a 3-paragraph executive summary, 
one paragraph per initiative, for a director-level audience.")

Constraints: What it cannot do. Tone, length, what to leave out.
(Example: "No HR jargon. Max 3 sentences. Not punitive in tone.")

Format: Email / bullets / exec summary / table / paragraph.
(Example: "One paragraph per section. No bullet points — she reads on mobile.")

Example: One real example of what good looks like.
(Paste a previous version, a comparable document, or a tone reference.)

The more complex the request, the more of these you need. You don’t need all five for an email draft. You need all five for a performance review, a board update, or a vendor negotiation.

Before / After | Quarterly Status Update

Before:

Write a summary of last quarter's performance.

After:

I'm a VP preparing a quarterly summary for my director. 
Format: 3 wins, 2 blockers, next quarter's priorities. 
One paragraph each. No bullets — she reads on mobile. 
Here's last quarter's for tone reference: [paste it]

The difference: you filled in Role, Goal, Format, and Example. Four fields instead of one. The output requires no rewriting.

AI For You infographic: context engineering

Before / After | Performance Review Comment

Kay handing a full briefing folder to the AI

Before:

Write a performance review comment for someone who missed their Q2 targets.

After:

I'm a director writing a mid-year review comment for a high-potential analyst 
who missed Q2 revenue targets due to a territory realignment outside her control. 
Tone: direct and developmental, not punitive. 3 sentences max. No HR jargon. 
Here's the comment I wrote for her last year: [paste it]

The difference: Role, Goal, Constraints, Format, and Example, all five. The comment you get back is one you’d sign your name to.

Before / After | Vendor Negotiation Prep

Before:

Help me prep for a difficult conversation with a vendor.

After:

I'm a procurement lead preparing for a contract renegotiation with a vendor 
who has missed two SLAs this quarter. Goal: open the renegotiation conversation 
without triggering a defensive reaction. Constraints: I can't threaten to cancel yet — 
legal hasn't approved that path. Format: give me 3 opening lines and the one question 
I should ask first.

The difference: every field is filled. You get something usable in the next meeting, not a generic negotiation framework off the internet.

You do not need all five every time

Quick email draft? Two fields is probably enough. Strategic document going to your leadership team? Use all five.

The rule: the more the output matters, the more context you give. Most people only fill in the goal. The other four fields are where AI stops being generic and starts being useful.

The debate-ready framing

a plain panel next to a vivid detailed one

When someone on your team says “I tried AI and it gave me garbage output,” this is almost always the context engineering problem. The model did not fail. The setup did.

The shift Tobi Lutke was pointing at is real. The skill is not writing a clever question. It is building the right environment for the question to be answered well. That is context engineering. That is what separates the professionals getting consistent value from AI from the ones who tried it once and moved on.

This week, pick one thing you regularly ask AI to write. Apply all five fields before you send. Notice the difference.

Come tell me on LinkedIn what you applied it to. I read every message.

VP @ Oracle | LinkedIn Top Voice for AI | LinkedIn Learning Instructor | Views are my own

AI For You is a weekly newsletter for corporate women (and yes, the guys reading this too, welcome) navigating careers in the AI era. Real tools, real context, no tech background required.