This Week: How to Build AI Assistants That Actually Work for You
Dear Reader,
Last week, I introduced the PAST Framework, a simple structure to help you write better prompts and get clearer, more useful responses from tools like ChatGPT.
But what if you’re ready to go beyond just writing prompts? PAST gets you better responses, but what about when you need the same task done dozens of times?
That’s where this week’s framework comes in - SHAPE - your blueprint for designing AI assistants that behave with purpose.
Quick Recap: What PAST Solves
PAST helps you write one-off prompts that work.
It ensures your prompt includes:
- Persona: who the AI should be
- Action: what you want it to do
- Structure: how it should deliver it
- Tone: how it should sound
It’s great for getting better results, faster. But it doesn’t scale on its own.
Next Level: Designing with SHAPE
SHAPE is the system I use when building custom GPTs or AI assistants—whether I’m helping a consultant turn blog posts into newsletters, or helping a team automate onboarding across multiple platforms.
It stands for:
- Scope
- Helpfulness
- Authority
- Process
- Edge Cases
Let’s walk through each part.
S — Scope
What exactly should this assistant handle (and what should it avoid)?
Example: This assistant converts raw meeting notes into client-ready summaries. It does not generate strategy, take meeting bookings, or handle CRM updates.
Setting clear boundaries will help prevent confusion and minimise hallucinations.
H — Helpfulness
How should the assistant behave? What’s the tone, level of initiative, pacing?
Example: Be proactive, brief, and friendly. Offer a suggestion when appropriate, but don’t make decisions without input.
This sets expectations and creates consistency across interactions.
A — Authority
What expertise does the assistant have? When should it lead, and when should it defer?
Example: Confidently suggest optimisations for email subject lines, but always defer to the user when tone or brand voice is uncertain.
This is about keeping it useful without letting it overstep.
P — Process
What step-by-step approach should the assistant follow when completing a task?
Example: 1) Review notes. 2) Identify main points. 3) Write summary. 4) Suggest 2 improvements. 5) Return output in markdown format.
This is your assistant’s operating system - a step-by-step breakdown of the process it should take to achieve its purpose.
E — Edge Cases
What should the assistant do if something unexpected happens?
Example: If the input is unclear, ask for clarification. If a task falls outside scope, suggest a manual review and explain why.
This doesn’t make your assistant any smarter or trustworthy. Planning for edge cases keeps you and your personal experience in control.
Real Example: PAST vs SHAPE in Action
Let’s say your original prompt is:
“Can you help me write some LinkedIn posts?”
That’s not much to go on. So we upgrade it using PAST:
PAST Prompt Version
- Persona: You are a B2B content strategist.
- Action: Write 3 LinkedIn post ideas based on this blog
- Structure: Each idea should include a hook, a theme, and a CTA
- Tone: British English, confident but non-salesy.
Great. That works - now let’s turn this into a reusable assistant using SHAPE:
SHAPE Assistant Design
- Scope: Repurpose blogs into short-form LinkedIn content. Don’t write new long-form content from scratch.
- Helpfulness: Write drafts, offer multiple angles, and suggest improvement tips. Keep tone warm and practical.
- Authority: Confident with formatting, hesitant with emotional tone—ask for approval when unsure.
- Process: 1) Summarise blog 2) Identify 3 post angles 3) Write 3 drafts with hooks, body, CTA.
- Edge Cases: If input is unclear or tone seems off-brand, pause and ask for clarification.
Why This Matters (Especially Now)
We’re entering the 'agentic AI' era where your assistants don’t just generate content, but take actions, make decisions, and collaborate across tools.
To scale with AI, you need more than good prompts. You need clarity of purpose and behaviour.
PAST helps you ask for what you want.
SHAPE helps you build something that delivers it every time.
Together, they let you stop copy/pasting and start building systems.
What If AI Actually Saved You Time Instead of Creating More Work?
Choose one task you repeat weekly and we’ll design an AI assistant that does it exactly how you want - every single time.
What you get:
- Custom SHAPE profile (no more trial and error)
- Working assistant ready to use immediately
- Template to build more assistants for other tasks
Until next time,
Jim
Signal Over Noise is written by Jim Christian. Subscribe at newsletter.jimchristian.net.
Join Our Signal Over Noise Community on Facebook
I've created a dedicated space where we can dive deeper into the topics covered in this newsletter, share insights, and connect with other signal-seekers. The Signal Over Noise Facebook page is where I'll be posting additional analysis, sparking discussions on emerging AI trends, and building a community of professionals focused on finding what truly matters in this rapidly evolving landscape.
Why join?
- Get bite-sized insights between newsletter editions
- Participate in focused discussions with fellow readers
- Share your own signal-finding perspectives
- Connect with others navigating the AI space thoughtfully
Ready for AI That Actually Works Together?
Stop switching between disconnected AI tools. In a 90-minute AI Action Plan Session, I'll show you how to set up the kind of orchestrated workflows I use daily—where your AI can read your files, update your systems, and execute complex tasks across all your tools. Let's design your unified AI workflow.
