You Asked AI to Do Eight Things. It Did Three.
Most people assume AI understood what they meant. It didn't. This free guide gives you 6 rules that make AI do exactly what you asked, every time.
Get the Free GuideGet Your Free Copy
Instant PDF download. No sales call required.
The difference IFP makes
Without IFP
"Update all 8 tabs in the dashboard."
3 tabs updated. 5 skipped with no explanation.
With IFP
"Update all 8 tabs in the dashboard."
All 8 tabs updated. Missing coverage flagged before execution.
You're Not Doing It Wrong. But There's a Gap.
When you give AI an instruction, it doesn't read every word the way you do. It matches patterns, estimates intent, and fills gaps with assumptions, often silently. The result is a response that looks complete but has quietly dropped half of what you asked for.
This isn't a flaw in the model. It's a gap in how instructions are parsed. AI systems process scope words like "each" and "all" differently from how a human colleague would. Without a structured protocol, scope narrows, tools get substituted, and outcomes drift from your original intent.
The Instruction Fidelity Protocol was developed to close that gap. It's six rules that force AI to parse your exact words before acting: not what it thinks you meant. Every word you write becomes binding. Every scope constraint is checked. Missing coverage gets flagged.
The 6 Rules That Close the Gap
A preview of what's inside the guide. Each rule has worked examples, anti-patterns, and a ready-to-use checklist.
Parse Before You Act
Every sentence is decomposed word-by-word before any action is taken. No summarising, no grouping.
Scope Words Are Binding
"Each", "all", and "every" enumerate the complete set: not just the items seen on a walkthrough path.
Use the Named Tool
If the user names a specific tool, that exact tool is used. No silent substitution without stating the fallback.
Define the Outcome First
Extract the verbs before the nouns. Write one outcome statement before creating any task list.
Verify Against Words Not Intent
Verification maps back to the user's exact sentences, not whether the code compiled or the API returned 200.
Missing Coverage Is Failure
Any user sentence with zero corresponding tasks is a gap. It must be covered before execution begins.
What's Inside the Download
Three resources. One objective: AI does exactly what you asked, every time.
The IFP Guide (7-page PDF)
All 6 rules explained in full with real examples from production AI systems. Separate sections for business owners (no code) and developers (implementation patterns).
Fillable IFP Checklist (Editable PDF)
A ready-to-use fillable PDF checklist. Run through the 6 checkpoints before every AI interaction. Takes 2 minutes. Prevents the half-done problem.
Claude Code IFP Skill (Free)
Install IFP directly into Claude Code as a reusable skill. It enforces the 6 rules automatically every time you give Claude a complex instruction. Includes SKILL.md and installation guide.
Ready to Make AI Listen Properly?
Download the free IFP guide and start applying the protocol in your next AI session.
Get Your Free Copy
Instant PDF download. No sales call required.
About Smart AI Solutions
Smart AI Solutions (Pty) Ltd is a Cape Town-based AI consultancy that helps South African businesses implement AI tools, automate workflows, and build internal AI capability. We work across industries with teams ranging from 3-person startups to enterprise departments.
The Instruction Fidelity Protocol was developed by Loxly Atkinson (CEO) after observing a consistent pattern across client engagements: teams were using AI correctly by most measures, but getting incomplete results because their instructions were being partially parsed. The protocol formalises the six-step correction that closed the gap in every case we tested.
This guide is the first time IFP has been published in a format anyone can apply immediately, regardless of which AI tool they're using.