AI works best when it removes work, not judgment
Over the course of this year, we’ve talked about the “easy button” and how many corporate leaders are blindly trusting AI to run their businesses.
Since we already covered AI scams, the legal consequences of using AI, and what happens when people are held accountable for their AI slop, I want to try something different.
I want to try to defend the AI “easy button.”
Despite my months and months of ranting about easy buttons, AI easy buttons absolutely exist. The problem is that companies keep pressuring AI, over and over, to make decisions that require experience, context, and accountability.
The best AI easy buttons don’t replace judgment. They eliminate the tedious work surrounding it.
Turn meetings into to-do lists
This, personally, has been one of the most transformative areas of my consulting business. See, maybe it’s my ADHD brain, but I can either have a great conversation with someone OR take great notes. If I try to do both, I do a poor job at both.
For me, AI transcription and summarization may be the most obvious legitimate easy button. Who would have thought we could have our cake and eat it too!
I personally use Fireflies.ai, but select your favorite note-taking app and have it start recording meetings and extract the following:
- Decisions made.
- Questions requiring follow-up.
- Assigned tasks and owners.
- Deadlines mentioned — this is my favorite!
- Unresolved debates. (I’ll come back to this one)
The person who owns the meeting (and sets up AI note-taking) reviews the AI’s output and sends it to the team. AI isn’t deciding what the business should do; it’s documenting what people already decided and attempting to hold them accountable.
Re: accountability: Ever had someone say they would do something and then, when confronted on that deliverable/action, they tell you, “no, that wasn’t me…” You can figure out a nice way to be like, “Yeah, so my AI notes back on XX/YY had you commit to this”. If they still fight you, then you can bring out voice recordings :)~
Measurable value:
- Time spent writing notes during the meeting.
- Percentage of tasks with an owner and due date.
- Missed or overdue follow-ups.
- Reduction in “what did we talk about last time” meetings.
Identifying data and patterns before analysis
Give AI access to exports from approved, non-sensitive datasets and ask it to categorize, summarize, identify efficiencies, or flag items that stand out.
SEO examples:
- Group thousands of Search Console queries by intent. (Better yet, your competitor’s keyword data!)
- Categorize declining pages by template or content type.
- Compare winners and losers after an algorithm update.
- Surface unusual changes requiring investigation.
- Convert messy exports into usable tables and graphics.
The boundary matters! AI can identify where something changed. It shouldn’t be trusted to determine why it changed without validation.
Measurable value:
- Number of rows reviewed.
- Analysis hours saved.
- Percentage of the dataset categorized.
- Additional opportunities or data anomalies discovered.
Transform existing work into new formats
Yes, I know we all want the easy button to take the prompt “take this idea and provide me a fully polished deliverable,” but that’s just not gonna happen.
Instead, leverage AI to recreate formats!
Examples:
- Convert podcast/webinar video into LinkedIn posts.
- Create executive summaries from detailed public reports.
- Draft meta descriptions from existing page copy.
- Convert documentation into FAQs or training materials.
AI handles transformation. A human protects the substance, voice, and accuracy of the end product.
You get it by now. You’re not asking it to think for you, just execute the grunt work you don’t actually want to do yourself. QA is the new superpower we need to adopt!
Measurable value:
- Production time savings. (Per asset creation)
- Number of assets created from one source. (This was always the goal since day 1!)
- Editing time. (But don’t let it blindly rewrite; make it give you recommendations or challenge the structure.)
Use AI as a first-pass editor
Most companies obsess over using AI to create more. But handing AI the final draft is like letting the ball boy call the game-winning play.
Let AI review the tape. Your experienced people should still make the call. (Anyone else getting excited for football!?)
Ask AI to review your work for:
- Missing information.
- Contradictions.
- Unanswered customer questions.
- Confusing language.
- Requirements that weren’t addressed.
This isn’t final approval. It’s an extra set of eyes that can spot missed opportunities before a credentialed or knowledgeable human signs off.
AI is far less dangerous when its job is to raise questions, identify gaps, and suggest opportunities — not write, approve, and submit the final answer.
Measurable value:
- Issues (big or small) caught before submitting your final work.
- Revision cycles and general review time.
- Post-publication corrections if content is online.
- Stakeholder feedback.
Eliminate repetitive technical work
This is one of the best use cases for AI. Use AI to produce small, testable pieces of technical work, such as:
- Spreadsheet formulas.
- Regex expressions.
- SQL queries.
- Schema markup.
- Redirect mapping.
- Basic scripts. (I especially love this for Apps Script within GSheets!)
- Looker Studio calculated fields.
The output either works or it doesn’t, which makes your QA pretty straightforward.
You still define the desired outcome, own QA of the result, and control the implementation. Nobody should deploy AI-generated code across a production site without reviewing it. (OK, I might have done this a time or two for SEOJobs.com, but let’s just say lessons were learned)
The safest easy buttons produce testable outputs. Not opinions that merely sound convincing.
Measurable value:
- Time required to complete the task.
- Manual steps eliminated.
- Error rate.
- Number of records processed.
- Cost compared with manual execution.
Delegate the labor, not the accountability
The true concept of an AI easy button was never the problem. The problem was deciding what we were willing to allow the button to control and what its impact would be.
Meeting notes, data organization, content transformation, editing/reviews, and repetitive technical tasks are all tedious, measurable, and relatively easy to QA.
Go ahead, press the button for all of these!
Strategy, expertise, prioritization, and final approval are 100% different. Those require context and someone willing to accept responsibility when the answer is wrong.
Use AI to remove time from the larger process. Use people to determine whether the work is accurate, useful, and worth acting on.
Don’t allow AI to push its own easy button. It shouldn’t decide what matters, simply what the options are, and the details surrounding them.
This post first appeared on the author’s website and is republished here with permission.
