

Most people who use AI tools at work spend more time fixing than they realise.
A 2026 Glean study found the average knowledge worker spends almost an entire workday every week fixing AI outputs. For business teams rolling out AI tools across the organisation, this adds up quickly.
A 2026 Glean study found the average knowledge worker spends almost an entire workday every week 'botsitting' – checking, fixing, and rebuilding AI outputs.
A lot of AI fix-up time comes from two things: outputs that weren't self-checked, and wasted time rebuilding prompts you've already written before.
The solution? Add a self-audit instruction to any prompt where the output is high-stakes and quality is a must. And build a reusable template for anything you do more than once a month. Neither takes more than a few minutes to set up, and we’ve got the prompts right here for you.
If you want hands-on practice with techniques like these, we offer AI training for business teams at all levels — from first-time AI users to teams building workflows from scratch.
Technique 1: Ask the AI to audit its own output before you see it
Most AI tools will do a better job if you tell them what "good" looks like before they start and show them how to check their own work at the end.
This prompt does both:
Before you start: I need this output ready to use, not a first draft I have to fix. Here's the task: [describe what you need].
When you're done, check your own output against these three things: (1) Have I made any claims I'm not actually sure about? If so, either verify them or flag them. (2) Have I used any filler phrases or generic language that could be cut without losing anything? If so, cut it. (3) Is the format and tone right for [describe your context and audience]? If not, adjust it.
Only show me something you'd be comfortable putting your name on.
We ran this against a plain task prompt using a messy internal document - a comparison of Sydney printing suppliers that needed turning into a clean report.
The output from the plain prompt was usable, but the self-audit version was noticeably tighter. Here's the opening from the self-audited output:
Seven providers were reviewed across CBD and Inner West Sydney. For a small run with quality finishes (foiling, spot UV), Fast Print Services (Alexandria) or Snap/Kwik Kopy (CBD) are the strongest options. For lowest cost with acceptable quality, Print On Sydney offers the best pricing online. Officeworks is the only option not recommended for a client-facing product.
Four sentences without any unnecessary padding or scene-setting. The plain prompt version was twice as long without adding anything useful. The self audit chopped exactly the kind of thing you'd waste five minutes cutting after the fact.
The self-audit step adds a few seconds on the AI's end and removes substantial editing time on yours. Use it for anything going to a client or anywhere else you'd feel embarrassed if the output was sloppy.
Technique 2: Build a reusable prompt template for tasks you do every week
Find a task you do more than once a fortnight. Think follow up emails and summaries after meetings, or project updates for a manager or client.
Stop prompting from scratch. Take five minutes once and use this prompt to build a solid template. It’ll save a few minutes every time you use it.
Here's the base prompt for generating your own template:
I have a task I do regularly: [describe the task and frequency, e.g. "weekly project update to my manager", "post-training follow-up email to a client", "inbound sales enquiry response"].
Here's the context you need to do it well: [paste the relevant background — your role, your audience, the tone you're going for, any format preferences, and any constraints like word count or things to avoid].
Write me a reusable prompt template I can use each time, with [VARIABLE] markers for the parts that change between uses. Keep the template under 150 words.
We asked Claude to generate a weekly task update template for a consultant reporting directly to a founder, pulling updates from our internal CRM. Here's the template it produced:
Weekly Update — [DATE]
Pull my current tasks from the [CRM] and generate a weekly update for [MANAGER]. Format as a concise bullet list grouped by status: In Progress / Completed / Blocked / Upcoming.
For each task include: task name, one-line status update, any blockers or next steps.
Additional context for this week: [ADD ANY NOTES — e.g. "I was out Thursday", "launched X project", "waiting on client Y"]
Highlight: [ONE KEY WIN OR PRIORITY TO FLAG]
Keep the tone direct and professional. No fluff.
Clear, direct and easy to adapt. The same template then produced a complete, accurate summary of a full month of CRM updates in a few seconds. The update was formatted effectively, grouped by status and ready to send.
It’s worth noting that if you use an AI tool with scheduling capabilities, you can take this further. Set it up to run automatically each week, and the AI pulls your latest tasks, formats the update and has it waiting for you on Monday morning.
Once you have a template you're happy with, save it somewhere easy to access. A sticky note in Notion, a pinned message to yourself in Teams or a text expander shortcut all work well.
Get prompting!
Both of these take a few minutes to set up and nothing to maintain after that. Pick the prompt that fits a task you’ve got in front of you and start there. If you’d like a more general framework for creating great prompts yourself, you can check out this article on how we teach prompt engineering at the AI Training Company.


