How I Got Consistent AI Responses Across My Team

And the simple system I built that fixed it

I still remember the day one of my team members sent me a draft of our marketing communication.

Right there, in the first paragraph, it said: "We are numero uno in Artificial Intelligence."

I sat with it for a moment. Took a breath.

This wasn't bad writing. This person was trying. They'd used AI, done their best, and genuinely thought this was good. But it was wrong on every level — the tone, the claim, the language, all of it. Not us. Not even close to us.

I called them over. Explained what we stand for, how we talk, what we never say. They nodded, went back, rewrote it.

Next time? Similar problem. Different words. Same gap.

And I realised — this wasn't a them problem. It was a me problem.

Everything that makes our communication ours lived in my head. Not written down anywhere. Not in any system. Just... in my head.

HBR recently named this "workslop" — AI content that looks polished but misses the mark. A Workday study found that 37% of time saved through AI gets eaten back by rework. I wasn't surprised. I'd lived it.

The problem isn't the tool. It's that the context — the voice, the standards, what good looks like — has never left the manager's head.

That's when I explored Projects in ChatGPT — now available in Claude and Perplexity too. A permanent workspace with consistent memory. Not a chat that forgets everything when you close it. A place that holds your context, always.

I call it a shared context system.

After the "numero uno" moment, I didn't sit my team member down again.

I built a system instead.

I created a ChatGPT Project for our team communications and put five things inside it:

1. Past communications that worked — emails that landed well, proposals that won, messages that sounded exactly like us. So AI knows what good looks like with real examples, not vague instructions.

   
   

2. What we never say — specific phrases, claims, and tones that are off-limits. Not "be professional." More like — don't make claims we can't back up, don't use corporate superlatives, don't write like a press release.

   
   

3. Our brand persona — who we are, who we're talking to, what we care about. Short and plain. Three paragraphs, not thirty pages.

   
   

4. Communication guidelines by type — how a client email differs from a proposal differs from a LinkedIn post. Same brand, different contexts.

   
   

5. A bad example with notes — one piece of communication that missed, and exactly why. Sometimes showing what wrong looks like is more useful than showing what right looks like.

   
   

First drafts came back closer to right. Iterations reduced. The corrections I gave actually stuck — because the context was there to hold them.

My team wasn't the problem. The missing shared context was.

It's not just a manager problem. It's especially a founder problem.

A founder I know runs a B2B SaaS company, team of 20, growing fast.

He called me recently, frustrated. "Yaar, every time my team builds a sales deck, it comes back wrong. Wrong tone, wrong focus, sometimes wrong facts. I spend more time fixing it than it would take to build it myself."

I asked him one question — "Where does your team start when they build a deck?"

He paused. "They just... start."

That was the problem. Every deck was being built from zero. His knowledge — which clients respond to what language, what goes on slide one, what a winning deck feels like — existed only in his head.

I told him — stop explaining after the fact. Build it in before they start.

He created a ChatGPT Project for his sales decks and put five things inside it:

1. Three past decks that won deals — so the team knows what good looks like in practice, not theory.

   
   

2. Client persona notes — who the typical buyer is, what they care about, what objections they usually raise.

   
   

3. Communication guidelines — the tone, the language, what the company never says in a sales context.

   
   

4. A slide-by-slide structure — what goes on each slide, what the opening must establish, how to close.

   
   

5. One example of a deck that didn't work — with notes on exactly why.

   
   

Two weeks later he messaged me. First drafts were coming back usable. His review time halved. His team stopped guessing and started building from a foundation.

He didn't hire better people. He didn't run a training. He just took what was in his head and built it into the system.

What building this system does to you

Here's something I didn't expect.

Writing those five things was harder than I thought. Not because writing is hard — but because it forced me to articulate things I'd never put into words before. Standards that lived in my gut had to become actual sentences.

A manager who can articulate their standards clearly has thought deeply about them. A manager who can't — hasn't yet.

The system doesn't just fix your team's problem. It forces you to fix your own clarity first.

And it compounds. Every update — new feedback, new examples, new guidelines — makes the system smarter. First drafts keep getting better. The gap between what you expect and what you receive keeps closing.

You're not solving today's problem. You're building something that improves every week.

What you can do this week

Start small.

1. Pick one recurring communication — client emails, proposals, sales decks, status reports.

   
   

2. Find 3 past examples that represent what good looks like. Yours, not templates from the internet.

   
   

3. Write 10 plain guidelines. Not "be professional." Something specific — "Don't use superlatives like 'numero uno' unless you can back it up with data."

   
   

4. Create a ChatGPT or Claude Project. Put the examples and guidelines in. Name it clearly.

   
   

5. Have your team use it for the next three pieces of work. Then review. See what changed.

   
   

The first time feels like effort. By the third project, you'll wonder why you ever tried to explain your standards one correction at a time.

The problem was never that AI can't write well.

The problem was that the only person who knew what "well" meant was you.

Build the system. Get your thinking out of your head. Let it work for you — every time, for every project, with every person on your team.

That's how shared thinking becomes shared quality. And that's how quality compounds.

Stay grounded. Stay curious.

-----

This is part of my ongoing work helping managers build an AI mindset—practical ways to think, decide, and work in the age of AI. It's not about mastering tools, but understanding how to work intelligently with intelligence. Looking forward to your feedbacks