I’ve spent 12 years in the trenches—rebuilding SEO infrastructures, fixing broken go-to-market motions, and cleaning up analytics setups that were essentially firing random data into a void. If there is one thing I’ve learned, it’s that most companies don’t have a strategy problem; they have a documentation problem. They spend weeks on 100-slide decks that look beautiful in a boardroom but are analytics and attribution functionally useless when a product manager sits down to ship code on a Monday morning.
At Valdor Consulting, my approach is simple: if a document doesn’t force a concrete decision, it’s just overhead. I keep a short client list because I don't want to manage "slide theater." I want to manage outcomes.
Recently, everyone has been asking: "How do I use AI to write faster strategy memos?" The answer isn't "ask ChatGPT to write the whole thing." The answer is building a rigorous, repeatable, multi-model workflow that treats AI as a junior analyst, not a CEO. Here is how you do it without losing your soul—or your credibility.
The "Monday Morning" Test: Why Most Memos Fail
Before you even open an LLM, ask yourself one question: "What decision will this change on Monday?"
If you cannot map the content of your memo to a specific, binary outcome—like "Ship feature X," "Kill campaign Y," or "Adjust target ICP by 15%"—then stop writing. Vague recommendations are the hallmark of consultants who bill by the hour rather than the impact. When you use AI to generate filler, you are amplifying the noise, not the signal.
The Multi-Model Workflow: Structuring Your Thinking
I never use a single model for a critical product document. Relying on one tool creates a feedback loop of hallucinations and biased logic. Instead, I use a multi-model workflow to build structured documents that actually stand up to scrutiny.
1. The Architect Phase (ChatGPT Plus/o1)
I use a high-reasoning model like ChatGPT (o1 or GPT-4o) as the architect. The goal here isn't to write prose; it’s to build the skeleton. I feed it my raw, messy notes, user interview transcripts, and historical data points, then I give it a specific constraint:
ChatGPT for product teams- "Analyze these notes against a standard RICE framework." "Identify three logical inconsistencies in my proposed go-to-market timeline." "Create a table that maps the current technical SEO debt against potential revenue gains."
2. The Auditor Phase (Claude 3.5 Sonnet or similar)
Once the structure is built, I switch models. I treat this new model as the "skeptical editor." I feed it the draft created by the architect and ask it to play devil’s advocate. Why? Because different models have different "personality" weights. If the architect model is optimistic, the auditor model is cynical. This prevents the echo-chamber effect that ruins most AI-assisted work.
3. The Synthesis Phase (Suprmind)
For complex product strategies, I use Suprmind to manage the knowledge layer. It helps ensure that the decisions we are making today are grounded in the context of the decisions we made six months ago. By feeding our internal decision logs into the process, I make sure we aren't reinventing the wheel or contradicting past learnings.
The Critical Step: AI Fact-Checking
The biggest threat to AI-assisted strategy isn't bad grammar; it's the "hallucination of expertise." AI will sound incredibly confident while inventing data points. AI fact-checking is not an optional add-on; it is a mandatory part of the workflow.
Here is my protocol for verifying every memo:
Type of Claim Verification Method Quantitative Data Cross-reference against raw CRM/Analytics exports. Never trust an LLM to "calculate" trends from memory. Competitive Intel Check against actual site architecture and pricing pages. AI is often 6-12 months behind on UI/UX changes. Technical Feasibility Review against your engineering team's current sprint capacity. Do not let the AI suggest "low-effort" solutions if the underlying API doesn't support them.If you don't have the source data to back up the AI's claim, it does not belong in the memo.
Building a Repeatable Growth System
At Valdor Consulting, I don't just hand over a document; I hand over a system. A product decision memo should be the bridge between abstract strategy and executable code. When we use AI to help draft these, we aren't looking for "cool copy." We are looking for clarity.
When you combine technical SEO knowledge—understanding how search engines actually crawl and index content—with high-level product strategy, you create documentation that serves two purposes: internal decision-making and external authority building. If your product strategy is clear enough to be captured in a structured memo, it is almost certainly clear enough to be repurposed into the kind of high-value content that actually drives organic traffic.
The Trap of Attribution
I have a visceral hatred for attribution setups that nobody trusts. The same logic applies to AI-assisted memos. If your team doesn't trust the logic behind the decision, they won't execute on it. That is why I demand that every memo includes the "source trace."
Every time the AI pulls a data point, I force the documentation to link back to the raw source. This builds psychological safety. When an engineer asks, "Why are we pivoting to this feature?" the answer isn't "The AI suggested it." The answer is "The data from our Q3 churn analysis showed X, and our technical debt audit from last month confirmed Y."
Final Thoughts: Don't Be a Buzzword Factory
The industry is obsessed with the idea that AI will replace the need for deep work. It won't. It just makes the work faster for those who are already good at it and more dangerous for those who aren't.
If you are looking for a magic button that creates a strategy for you, you’ll find plenty of consultants in Belgrade and beyond who will sell you that dream. But if you want to ship, grow, and actually understand why you’re doing what you’re doing on a Monday morning, you have to do the heavy lifting yourself. AI is just the hoist.
Use ChatGPT to structure, use specialized tools like Suprmind to synthesize, and use your own brain to audit. If you can’t defend the decision without the AI, you haven’t made a decision—you’ve just made a guess.
Quick Checklist for Your Next Memo:
- Does this memo identify one clear "Monday morning" action? Did I use at least two different models to cross-verify the logic? Is every data point linked to a raw, verifiable source? Did I remove all fluff, buzzwords, and "consultant-speak"? Is the format structured (tables, clear headings) rather than dense, narrative prose?
Keep your lists short, your data clean, and your decisions binary. That’s how you actually move the needle.