A practical playbook for making locked-down enterprise AI smarter
Improve that corporate AI without waiting for the next IT rollout.
Brian Snyder is chief innovation officer at Axicom.
If you use ChatGPT or Claude in your personal life, you know the “magic.” It knows your kids’ names, your writing style and that one project you’ve been procrastinating on for six months. It feels like a partner.
Then you log into your corporate AI tool — the secure, locked-down, GDPR-compliant “Sidekick” or “Your Company GPT” — and it’s like 50 First Dates.
“Hello, stranger,” it says every single morning. It has no idea who you are, what your company stands for or that you’re currently in the middle of a $20M crisis management campaign.
It’s “dumb.” Not because the model is weak, but because it’s starved of context.
In a restricted corporate environment, the magic doesn’t happen automatically. You have to inject it yourself. Here’s how to make your corporate AI work like a genius partner.
It’s all about the context cards
The secret to a “magical” AI experience isn’t a better prompt library — it’s context cards. These are standalone documents (Word, PDF or text) that you feed into your corporate system at the start of every project — or better yet, upload to the “knowledge base” of your custom GPTs, Gems or Agents.
There are three pillars you need to build:
1. The personal context card
Your AI needs a bio. Stop re-explaining your role every morning. Create a document that outlines:
• Who you are: Your title, your core responsibilities and your goals.
• Your voice: Examples of your writing style, your preferences (e.g., “no fluff,” “direct tone”) and your typical “no-go” words.
• Your workflow: How you like to receive information (e.g., “always give me three options,” or “start with a 3-sentence executive summary”).
Pro tip: Have the AI interview you. Tell it: “I want to create a Personal Context Card so you can assist me better. Ask me 10 questions about my job, my voice and my preferences. Then synthesize my answers into a comprehensive document.”
2. The organization context card
Individual brilliance is great, but your AI needs to understand the “company POV.” This document should include:
• Mission and values: What does your company stand for?
• Standard operating procedures: How does your team approach a media plan or an internal memo?
• The “no-fly” zone: What terms or executives are restricted? What are the brand guardrails?
When your AI understands the organizational context, it stops giving you “generic corporate slop” and starts giving you content that actually looks like it came from your agency.
3. The project context card
This is the most dynamic piece. For every major initiative (like a Truck Driving Championship or a $20M global launch), create a time-bound context card:
• The project charter: Goals, stakeholders and deadlines.
• Historical data: What did we do last year? What worked? What failed?
• The current vibe: What is the news cycle? What are the specific pain points we are trying to solve right now?
The “addendum” trick: Simulating memory
Corporate systems often “forget” conversations once a session ends. To fix this, use the addendum trick.
At the end of a productive session, tell the AI:
“We’ve made great progress. Summarize the key decisions we made, the new preferences you’ve learned about my style and any project updates. Format this as an ‘Update Note’ that I can add to my context documents for next time.”
Download that summary, paste it into your context card and upload it to your next session. You’ve just manually built a memory.
From efficiency to effectiveness
Most people use AI for efficiency — saving ten minutes on a draft. That’s fine, but it’s small thinking.
When you master context injection, you move to effectiveness. You start using AI to do work you couldn’t do before. You can ask it to:
• Predict how your “Agency POV” would react to a new cultural trend.
• Stress-test a communications plan against the specific biases of your stakeholders.
• Connect the dots between three different projects that have separate context cards.
Don’t wait for perfect
Your IT department might not give you a “magical” autonomous assistant for another year. Don’t wait. By building and injecting context today, you turn a dumb machine into a partner.
The most important thing you can do to get ahead right now is to be early. Start building your cards.
Learn more from Ragan’s Center for AI Strategy.
