How to Run AI Through a Human Workflow: Ask → Verify → Decide

Every consultant I know has the same problem with AI: they love what it drafts, but they don't trust it enough to hit "send."

The fear is real. AI hallucinates statistics, invents case studies, and sometimes generates confident-sounding nonsense that could torpedo your credibility with a client. As a business architect who's built Target Operating Models for few years, I can't afford to present fabricated ROI data or misaligned strategic recommendations just because ChatGPT thought it sounded good.

But here's what I've learned: AI isn't the problem. Treating it like an autopilot is.

The solution isn't to avoid AI—it's to build it into your workflow as a tool, not a decision-maker. This means creating a repeatable three-step system that separates what AI does well (drafting) from what humans must own (verification and judgment).

The Three-Step Framework

Step 1: Ask (AI Drafts)

This is where AI shines. Use it to generate the first pass of any deliverable—whether that's a pain points summary, a capability assessment, or a Cost-Benefit Analysis.

What you do:

• Upload your raw inputs (current state notes, quantified pain points, workshop transcripts)
• Give AI a clear instruction: "Draft a 1,200-word analysis of how these operational pain points map to business capabilities. Use a professional consulting tone."
• Let it produce the skeleton

Why this works: AI tools like Claude 3.5 Sonnet excel at synthesis and can process large volumes of information quickly. They turn unstructured notes into coherent narratives faster than any human can. But this draft is just that—a draft.

Step 2: Verify (Human Validates)

This is the non-negotiable step most people skip. You must fact-check every claim, statistic, and strategic assertion before it goes into your final output.

What you do:

• Cross-reference any data points AI mentions against your source documents
• Check that the logic flows correctly (e.g., "Does this pain point actually map to this capability, or did AI make a leap?")
• Use tools like Perplexity Pro to validate external claims with real-time sources
• Flag any statements that "sound right" but you can't verify—these are hallucination red flags

Pro tip: Create a "verification checklist" specific to your domain. For my TOM work, I check:

• Are the capability definitions aligned with industry standards?
• Do the quantitative benefits match my original calculations?
• Are the strategic opportunities actually tied to the client's North Star objectives?

Step 3: Decide (Human Owns the Output)

This is where you exercise judgment. AI gave you a draft. You verified the facts. Now you decide: What stays? What goes? What needs your unique perspective?

What you do:

• Rewrite sections where AI's tone doesn't match your voice or the client's culture
• Add the context AI can't know (e.g., "Based on my conversation with the CFO last week, this cost estimate needs to account for...")
• Insert your point of view—the "so what?" that turns analysis into insight
• Make the final call on whether this content is ready to represent your professional brand

Why you can't skip this: Thought leadership content that doesn't sound AI-generated requires your authentic voice and domain expertise layered on top of AI's draft.

Why This Framework Solves the Trust Problem

The genius of "Ask → Verify → Decide" is that it removes the binary choice between "use AI" or "don't use AI." Instead, it embeds AI into a controlled process where:

1. AI accelerates drafting (saving you 60-70% of writing time)
2. You maintain quality control (protecting your credibility)
3. Human judgment remains the final arbiter (ensuring strategic alignment)

This isn't about "AI vs. Human." It's about designing a workflow where both contribute their strengths. AI handles volume and speed. You handle accuracy and wisdom.

The Bottom Line

AI hallucinations are real, but they're only dangerous if you treat AI output as finished work.

By running AI through a human workflow—Ask → Verify → Decide—you get the productivity gains without the credibility risk. AI becomes your research assistant and first-draft writer, while you remain the architect, validator, and decision-maker.

For consultants like us who work with complex frameworks like TOMs, capability models, and strategic alignment, this isn't optional—it's essential.

Start with one deliverable this week. Use AI to draft it, verify every claim yourself, and then decide what's ready for your client. You'll find that the workflow doesn't slow you down—it makes AI finally usable at the quality level your reputation demands.