In my last post, I wrote about a simple idea:
AI is a tool in a rather large toolbox.
That analogy struck a chord because it cuts through the hype.
AI is not magic.
It is not a replacement for expertise.
It is not a shortcut to good thinking.
It is a tool.
And like any tool, its usefulness depends entirely on the person using it.
Give a master carpenter a power saw, and you get a beautifully crafted table.
Give the same saw to someone who has never measured a plank in their life, and you may end up with three uneven legs and a trip to the emergency room.
The same principle applies to AI.
Over the past few years, first as a freelancer and now as a marketing and communications professional, I have used AI almost daily.
I use it to build proposals, structure articles, generate ideas, analyze messaging, prototype systems, and create operational frameworks.
Sometimes it saves me hours.
Sometimes it helps me see angles I might have missed.
Sometimes it confidently produces something so gloriously wrong that I laugh out loud.
That last part is important.
Because AI does not know your business.
It does not know your customers.
It does not know your constraints.
It does not know the politics in your organization, the realities on the ground, or the difference between what should happen and what actually happens.
It can only predict what is most likely based on patterns in the data it has seen.
In other words, AI is an exceptionally fast guesser.
A very useful guesser, but a guesser nonetheless.
If you want better answers, you need to give it better information.
That is where most people stumble.
They ask vague questions and expect brilliant answers.
Then they conclude that AI is overrated.
That is a bit like handing a contractor a sticky note that says, “Build me a house,” and then acting shocked when the result is not exactly what you had in mind.
So let us talk about how to use AI properly in operations and business workflows.
What AI Can Do Extremely Well
AI shines when the task involves structure, pattern recognition, and rapid iteration.
It is especially useful for:
- Drafting documents
- Summarizing large amounts of information
- Creating process checklists
- Generating SOPs
- Building templates
- Brainstorming ideas
- Rewriting content for different audiences
- Analyzing messaging
- Organizing research
- Creating project plans
In operations, this means AI can help you:
- Write standard operating procedures
- Build onboarding documents
- Prepare meeting agendas
- Draft client proposals
- Generate status reports
- Create customer support scripts
- Structure training materials
- Develop process maps
Think of it as an extremely fast operations assistant with broad general knowledge and no ego.
It never complains.
It never asks for a raise.
It also occasionally hallucinates with the confidence of a man explaining cricket rules after two beers.
What AI Cannot Do
AI cannot:
- Read your mind
- Understand unwritten context
- Know internal politics
- Access real-time realities unless you provide them
- Distinguish between theoretical and practical constraints
- Make strategic decisions for you
- Accept accountability for bad outcomes
It does not know:
- Who will read the output
- What your stakeholders care about
- What language resonates with your customers
- Which legal or compliance constraints apply
- What has failed before
- What “good” looks like in your organization
If you do not provide that context, the AI fills in the blanks using probability.
Sometimes those assumptions are useful.
Sometimes they are spectacularly off the mark.
Garbage In, Polished Garbage Out
We have all heard “garbage in, garbage out.”
With AI, it is more dangerous because the garbage comes back looking polished and persuasive.
That is where people get fooled.
The formatting is clean.
The language is smooth.
The tone sounds authoritative.
But beneath the polished exterior, the logic may be weak, the assumptions may be wrong, and the recommendations may be impractical.
This is why human judgment remains non-negotiable.
AI can accelerate thinking.
It cannot replace thinking.
The Information AI Needs From You
If you want AI to produce useful operational outputs, provide the same information you would give a competent employee or consultant.
1. Objective
What are you trying to achieve?
Examples:
- Reduce onboarding time from 10 days to 5 days
- Create an SOP for handling customer complaints
- Build a proposal for a defense technology client
- Design a content approval workflow
2. Audience
Who will use the output?
Examples:
- New hires
- Senior management
- Customers
- Engineers
- Sales teams
3. Context
What is happening in your organization?
Examples:
- We are a 300-person company
- The audience is technical but time-constrained
- Our customers are enterprise buyers
- The process must comply with ISO requirements
4. Constraints
What limitations must be respected?
Examples:
- Budget limits
- Approval chains
- Regulatory requirements
- Deadlines
- Available tools
5. Desired Format
How should the output be structured?
Examples:
- Checklist
- SOP
- Slide outline
- Executive summary
- Table
6. Tone
How should it sound?
Examples:
- Formal
- Conversational
- Technical
- Persuasive
- Concise
7. Examples
Provide reference material if possible.
Examples are gold.
If you can say, “Use this structure,” the AI becomes significantly more accurate.
A Bad Prompt vs a Good Prompt
Bad Prompt
“Create an onboarding process.”
That is the equivalent of saying, “Cook something.”
Good Prompt
“Create a 30-day onboarding plan for a junior sales executive joining a B2B defense technology company. The employee is fresh out of school and unfamiliar with SAP. Include daily tasks for the first week, weekly milestones for the rest of the month, and recommended training resources. Present the output as a table.”
Now the AI has enough information to produce something useful.
AI as an Operations Multiplier
Used correctly, AI can compress hours of work into minutes.
For example:
- Drafting a policy that normally takes two hours may take 15 minutes.
- Building a meeting summary may take three minutes.
- Turning notes into a structured SOP may take ten minutes.
- Creating alternate versions for different audiences may take seconds.
That speed matters.
But speed without judgment is how organizations create impressive-looking documents that no one can actually use.
My Personal Workflow
Here is how I typically use AI.
Step 1: Define the Outcome
What does success look like?
Step 2: Provide Context
Who, what, why, and constraints.
Step 3: Generate a First Draft
Treat it as a starting point, not a final answer.
Step 4: Critique the Output
What is missing?
What feels generic?
What assumptions are wrong?
Step 5: Refine Iteratively
Continue until the output is genuinely useful.
Step 6: Apply Human Judgment
Reality always gets the final vote.
Where AI Delivers the Highest ROI
The biggest gains usually come from repetitive knowledge work.
Examples include:
- Report creation
- Proposal drafting
- Documentation
- Meeting notes
- Research synthesis
- Content repurposing
- Internal communications
If you find yourself doing the same type of thinking repeatedly, AI can probably help.
If the task requires deep domain judgment, AI can assist, but you still need to drive.
The Human Advantage
The real advantage today is not access to AI.
Everyone has access.
The advantage lies in:
- Clear thinking
- Precise instructions
- Domain expertise
- Critical evaluation
- Rapid iteration
In short, the winners are the people who know what they want and can recognize quality when they see it.
A Useful Mental Model
Think of AI as a very bright intern.
It can:
- Work quickly
- Generate ideas
- Draft documents
- Follow detailed instructions
But it can also:
- Misunderstand context
- Make incorrect assumptions
- Present flawed work confidently
You would not hand an intern a vague assignment and publish their first draft without review.
Do not do that with AI either.
The Operational Impact
When teams learn to use AI effectively, they become faster and more consistent.
Processes are documented more thoroughly.
Ideas are tested more quickly.
Communications become sharper.
Execution accelerates.
That does not eliminate the need for skilled people.
It increases the value of skilled people.
Final Thoughts
AI is not a crystal ball.
It is not an oracle.
It does not possess secret insight into your business.
It makes educated predictions based on the information available.
If you provide poor inputs, you get polished assumptions.
If you provide rich context, clear objectives, and thoughtful constraints, you get outputs that can materially improve your operations.
So the next time AI gives you a mediocre answer, do not ask whether the tool is broken.
Ask whether you gave it enough to work with.
Because the quality of the output still depends on the quality of the thinking behind it.
And that part, thankfully, is still your job.
