There is a strange way to use AI badly.

You open ChatGPT to save time.

You ask for something.

The answer is almost right.

So you ask again.

Then you rewrite part of it.

Then you fix the formatting.

Then you copy the result into another tool.

Then you realise the AI misunderstood an instruction from five messages ago.

Twenty minutes later, you’ve technically “used AI.”

But you haven’t necessarily saved any time.

That is the part of AI productivity that gets missed.

Using AI more often does not automatically make you more productive.

The real improvement comes when AI removes a repeated piece of work and leaves behind something you can actually use.

Recent discussion around AI and productivity reflects this distinction. AI adoption is moving quickly, but wider productivity gains are not automatic because businesses still have to redesign workflows, train people and absorb implementation costs.

At the individual level, the same principle applies.

The question isn’t:

How many tasks can I give AI?

A better question is:

Where am I repeatedly spending time that doesn’t require my full attention?

That’s where AI starts becoming useful.

1. Stop Writing a New Prompt Every Time You Need Something

One of the easiest ways to waste time with AI is to treat every conversation as a fresh task.

You need a product description.

You write a prompt.

Tomorrow you need another product description.

You write another prompt.

Next week, same thing again.

This is exactly the pattern your source article identifies as one of the most expensive mistakes: solving the same problem repeatedly instead of turning it into a reusable system.

Turn repeated prompts into templates

Suppose you regularly write customer emails.

Instead of typing:

Write a follow-up email for this customer.

build something reusable:

Write a follow-up email for [CUSTOMER TYPE].

Goal: [GOAL]
Context: [CONTEXT]
Tone: [TONE]
Length: [LENGTH]
Required points: [POINTS]

Now you replace the variables.

That’s much faster.

The same method works for:

• product descriptions;
• social posts;
• meeting summaries;
• client reports;
• spreadsheet analysis;
• customer support;
• research briefs;
• blog outlines;
• sales emails.

The productivity gain doesn’t come from a brilliant prompt.

It comes from not rebuilding the prompt next Tuesday.

2. Tell AI What the Finished Work Should Look Like

A surprising amount of AI “productivity” disappears during cleanup.

The answer is good, but it’s in paragraphs.

You need bullets.

It gives you 1,000 words.

You needed 300.

It gives you Markdown.

You wanted a table.

The source material identifies this directly: when you fail to define output format, you create post-processing work for yourself.

So before asking AI to do the task, ask yourself:

What would a finished, usable answer look like?

Then describe that.

For example:

Summarise these meeting notes.

Output:

• 5 key decisions
• action owner beside each task
• deadline if mentioned
• unresolved questions at the end
• maximum 300 words

Now the output is closer to something you can use immediately.

OpenAI’s current prompting guidance explicitly recommends being specific about the desired outcome, length, format and style, and putting the instruction clearly at the start.

That sounds simple.

It is.

It also removes a ridiculous amount of avoidable reformatting.

3. Put the Important Instruction First

People often write prompts the same way they tell stories.

Background first.

Details next.

Actual request at the bottom.

Something like:

I’m working on a new email campaign for our software company. We serve small businesses and recently launched a new reporting feature. We’ve had some customer feedback about usability and our audience tends to prefer concise communication. Could you write a promotional email? Keep it under 150 words and don’t mention competitors.

The AI has to identify which part matters most.

You can make its job easier.

Write a promotional email under 150 words.
Do not mention competitors.

Product: reporting software for small businesses
New feature: automated reporting
Audience preference: concise communication
Goal: encourage feature adoption

Same information.

Less wandering.

Your source article recommends exactly this: put the core task and critical constraints first, then supply supporting context.

This is one of those changes that saves a few minutes at a time.

Which doesn’t sound impressive.

Until you do the same task fifty times.

4. Give AI the Context You Keep Repeating

There is another productivity trap.

You spend ten minutes explaining your company to AI.

Tomorrow you explain it again.

Then again in another chat.

Then again when using another model.

Your article calls this out as reusable context: recurring information should be stored once and inserted when needed rather than reconstructed every time.

Build reusable context blocks.

For example:

Company context

We sell accounting software to small businesses. Our customers are usually owner-managed companies with fewer than 25 employees. Our tone is clear and practical. Avoid exaggerated claims. Our main differentiator is simplified reporting for non-accountants.

Save that once.

Then use it wherever necessary.

You can do the same with:

• your writing style;
• customer personas;
• product information;
• brand rules;
• standard disclaimers;
• reporting structure;
• project requirements;
• recurring data definitions.

The interesting productivity improvement here is not faster writing.

It is fewer repeated decisions.

5. Use AI for the First Pass, Not Necessarily the Final Decision

AI becomes very useful when the cost of producing a first version is high but the cost of checking it is low.

Consider a messy set of meeting notes.

You could spend twenty minutes organising them.

Or ask AI:

Organise these notes into decisions, action items and unresolved questions.

Then spend three minutes checking the result.

That is a sensible productivity trade.

Other good examples include:

• summarising long documents;
• extracting action points;
• classifying customer feedback;
• restructuring messy notes;
• generating first-draft formulas;
• comparing several options;
• producing an initial research structure;
• turning raw information into a first draft.

The useful distinction is this:

Let AI reduce the cost of getting to version one.

You still decide whether version one is correct.

This also helps prevent the problem raised in your anti-AI writing material: when AI replaces the original thinking entirely, the result often becomes generic or detached from the user’s actual judgment. Your source recommends using AI to critique and refine rather than blindly rewriting everything.

That principle applies far beyond writing.

6. Test Important AI Workflows Before You Trust Them

Suppose you create a prompt that produces a perfect report.

You save it.

You build part of your workflow around it.

Then three days later it produces something completely different.

That does not necessarily mean the AI “broke.”

Generative models can produce variation across runs.

Your uploaded article recommends testing important prompts multiple times before relying on them for repeated work.

For everyday use, you don’t need to turn this into a laboratory.

But if the prompt matters, test it.

Try different examples.

Look for:

• missed constraints;
• inconsistent structure;
• invented information;
• strange edge cases;
• formatting failures.

If a prompt works four times and fails badly on the fifth, that failure is valuable.

You found it before a customer did.

7. Match the AI Tool to the Job

Another easy mistake is assuming every AI model should behave exactly the same.

They don’t.

Some handle long documents better.

Some are particularly strong at coding.

Some follow complex formatting instructions more reliably.

Some integrate with your existing software.

Some have access to files, browsing, workplace systems or other tools.

Your original article correctly argues that model-specific behaviour matters and that important workflows may need slightly different instructions depending on the AI being used.

Anthropic’s current documentation, for example, still recommends clear structural separation for complex prompts and notes that XML-style tags can help distinguish instructions, context and inputs when needed.

The productivity lesson is not:

Learn every AI model.

It’s:

Don’t spend half an hour forcing the wrong tool to do a job another tool handles easily.

If your current setup works, keep it.

Switch when the workflow gives you a reason.

8. Turn Repeated AI Tasks Into Workflows

This is where the biggest productivity gains usually appear.

A single prompt saves a few minutes.

A reusable workflow can save those same minutes every day.

Imagine that every Monday you prepare a weekly sales report.

Without a workflow:

1. Export the data.
2. Open the spreadsheet.
3. Calculate the changes.
4. Identify unusual results.
5. Write a summary.
6. Format an email.
7. Send it to the team.

Now suppose AI handles the repetitive middle:

Sales data → comparison → anomalies → draft commentary → review

You still approve the interpretation.

But the routine preparation disappears.

This is the difference between using AI and building AI into work.

The original article reaches essentially the same conclusion when it argues that reusable prompt libraries should become assets that improve over time instead of leaving users with a blank prompt window for every task.

9. Use AI to Reduce Switching Between Tools

Not all wasted time looks like work.

Sometimes it is:

copy,

paste,

open another tab,

find a file,

copy again,

reformat,

switch back.

Ten seconds here.

Thirty seconds there.

Repeated hundreds of times.

If your AI environment can access the tools where you already work, use that.

For example:

Read this document and create the action list.

Or:

Compare these two spreadsheets and highlight the differences.

Or:

Summarise these customer emails and group the problems by category.

The less information you manually ferry between systems, the more useful AI becomes.

This is also where privacy matters.

Do not casually paste sensitive customer records, confidential business information or restricted workplace material into unapproved AI tools. Recent reporting around workplace AI adoption has highlighted “shadow AI” use as a productivity benefit accompanied by significant data-security concerns when staff use unsanctioned systems.

Saving time isn’t much of a productivity improvement if it creates a compliance problem.

10. Ask AI to Find the Work You Should Stop Doing

This may be more valuable than asking it to do additional work.

At the end of a week, give AI a list of your recurring tasks.

Then ask:

Categorise these into:

1. work I must personally do,
2. work AI could assist with,
3. work that could be automated,
4. work that may not need to be done at all.

That fourth category matters.

Productivity isn’t the ability to complete more pointless tasks per hour.

Sometimes the correct optimization is deletion.

You can also ask:

Which three activities here appear repetitive but still require human judgment?

Or:

Which parts of this process create no direct value for the customer?

Now AI becomes less like an assistant and more like a process-review tool.

That’s a much more interesting use.

A Simple AI Productivity System

You do not need fifty AI hacks.

Start with one recurring task.

Then ask five questions.

1. Does this task happen repeatedly?

If no, automation may not be worth the setup.

2. Which part requires actual judgment?

Keep that part with you.

3. Which part is repetitive?

Give AI that part first.

4. What should the finished output look like?

Specify the format before asking.

5. Can I reuse this next time?

Save the prompt, context or workflow.

That’s the basic cycle:

identify → delegate → check → save → reuse

The first use may not save much time.

The fifth usually does.

Where AI Actually Improves Productivity

AI is particularly good at work involving transformation.

Turning:

• notes into summaries;
• data into commentary;
• documents into action items;
• research into outlines;
• raw ideas into structured drafts;
• customer feedback into categories;
• long explanations into short versions;
• recurring instructions into repeatable templates.

It becomes less reliable when the task depends on information it doesn’t have, requires precise judgment without review, or carries significant consequences if it gets something wrong.

That distinction is useful.

You don’t need AI doing your entire job.

You need it doing the parts of your job that never deserved an hour in the first place.

Frequently Asked Questions

How can AI improve productivity at work?

AI can increase productivity by reducing time spent on repetitive writing, summarisation, formatting, research organisation, data classification, first drafts and recurring workflows. The largest gains usually come from repeated processes rather than isolated one-off prompts.

What are the best ways to use ChatGPT for productivity?

Useful applications include summarising documents, creating templates, drafting routine communications, reorganising notes, analysing structured information, creating checklists and producing first drafts that you review.

Does using AI automatically make you more productive?

No. Poorly designed AI workflows can create additional rework, checking and formatting. Productivity improves when the AI output fits directly into the next step of your workflow.

Should I create reusable prompts?

Yes, particularly for repeated tasks. A reusable prompt reduces setup time and improves consistency because you do not have to reconstruct the same instructions repeatedly.

How detailed should an AI prompt be?

Detailed enough to remove ambiguity, but not filled with irrelevant information. State the task, important context, constraints and desired output format clearly. Current OpenAI guidance similarly recommends clear, specific prompts with sufficient context and iterative refinement.

Can AI replace workflow automation software?

Sometimes for small tasks, but not always. AI is good at interpreting unstructured information, while traditional automation is often better for predictable rule-based steps. The strongest workflows frequently combine both.

Tools & Resources

A general AI assistant such as ChatGPT or Claude can handle many of the productivity uses described above.

For repeated work, the more important question is whether your AI tool can fit into the rest of your workflow.

Look for capabilities such as:

• document analysis;
• file handling;
• spreadsheet support;
• integrations;
• reusable instructions;
• project context;
• automation;
• structured output.

But don’t start with the tool.

Start with the task you keep doing.

Then ask whether AI can remove one unnecessary step from it.

That is usually where productivity begins.