# 5 AI Productivity Skills | Field Office

> AI is the most useful when it fits the way you work. Here are five skills we’re experimenting with at Field Office, each with an example to try and a file you can download and adapt to your own workflow.

Source: https://www.field-office.ca/blog/five-ai-skills-for-better-everyday-work

Content Production

# 5 AI Productivity Skills

10 November 20267 minute read

AI is the most useful when it fits the way you work. With _skills_, you can give your AI clear, reusable instructions for recurring tasks, so it can write in your voice, ask better questions, or take action with the tools you already use.

Here are five skills and ideas we’re experimenting with at Field Office, with examples you can try and files you can adapt to your own workflow.

## 1\. My Voice: Make it sound like you

You ask for a client update. The facts are right, but the text is too formal, or too enthusiastic, or way longer than anything you’d actually send.

_My Voice_ starts with the person you’re writing to and the channel you’re using. A message to a longtime teammate shouldn’t sound like a first email to a prospective client. This skill looks at a small, relevant sample of your own writing and uses that evidence to guide the draft.

It pays attention to concrete choices: how direct you are, how much context you include, how you make requests, and whether you use a greeting. It also removes common AI habits like filler openings, inflated language, and repetitive conclusions.

The meaning still has to stay intact. Names, dates, numbers, uncertainty, and commitments shouldn’t change to make a sentence sound smoother. If you say the mobile layout needs checking before it goes to the client, the draft must preserve that condition.

Try: “Draft a reply in my voice. The layout is approved, but mobile needs a final check before we send it. Use my recent messages in this thread for tone.”

If relevant examples are unavailable, your AI should say so. And a good draft is still a draft: this skill doesn’t authorize sending or publishing it.

Read full My Voice skill

[Download .md](https://www.field-office.ca/skill-guides/my-voice.md)[Open Markdown file for My Voice](https://www.field-office.ca/skill-guides/my-voice.md)

## 2\. Grill Me: Ask questions before committing

AI is quick to suggest a solution. That speed is useful when the goal is clear, but less useful when the questions that determine the goal are still unanswered.

_Grill Me_ directs your AI to interview you about a plan before building it. The version we’ve provided here is focused on software design, but you can adapt it for other kinds of planning. It starts with what you want to build, then works through the important decisions in order, asking one to three focused questions at a time.

For a dashboard, “Which charts should we use?” isn’t the first question. Start with who needs the dashboard, which decision it supports, and what the underlying data actually tells you. “A project needs attention” might mean a missed deadline, an unanswered client request, or an exhausted budget. Those are different problems.

After each round, the AI summarizes what’s settled and what remains open. It should challenge easy answers, expose dependencies, and keep the conversation out of implementation mode until the design is understood.

You can adapt the same approach to business decisions like pricing, demand, or delivery capacity. The principle is the same: find the assumptions before they become expensive commitments.

Try: “Grill me on this plan before suggesting a build: a dashboard that tells us which client projects need attention.”

Read full Grill Me skill

[Download .md](https://www.field-office.ca/skill-guides/grill-me.md)[Open Markdown file for Grill Me](https://www.field-office.ca/skill-guides/grill-me.md)

## 3\. Take the Wheel: Finish the permitted work

Your AI can give you a six-step answer and still leave all six steps for you to do. Or, you can let it get things done.

_Take the Wheel_ is built around a straightforward preference: when your AI has the tools and permission to complete the task, it should complete the task. Instructions are the fallback, not the default deliverable.

The skill tells the AI to check for an appropriate app integration first. If a direct connection can’t do the work, browser or computer tools may be available. The AI should use what’s actually connected rather than assume it has access to every system.

A concrete outcome helps. “Help with the blog” leaves a lot open. “Save one unpublished article in our blog CMS and give me a working review link” identifies the result, destination, and stopping point.

That doesn’t remove approval requirements. Sending a message, publishing a page, making a payment, or changing access can have consequences. The AI should complete the permitted parts, pause where required, and describe anything still blocked.

The final report should be specific: what changed, where to find it, and what remains. Saved is different from published. A prepared draft is different from a sent message. A link is only useful if it opens the result you asked for.

Read full Take the Wheel skill

[Download .md](https://www.field-office.ca/skill-guides/take-the-wheel.md)[Open Markdown file for Take the Wheel](https://www.field-office.ca/skill-guides/take-the-wheel.md)

## 4\. Workflow Insights: Stop making the same fixes

Repeated corrections are evidence. If you keep redirecting an AI, there may be a process problem to fix.

The _Workflow Insights_ skill reviews your recent conversations and coding sessions over a defined period, fourteen days by default. It samples recent work and looks for repeated user corrections, recurring tool errors, long debugging loops, and unnecessary duplication.

The aim is to turn a pattern into a practical recommendation. If a mistake can be caught automatically, a test or validation rule may be more reliable than another instruction to “be careful.” If the issue depends on judgment, a clearer instruction or a focused feedback note may fit better.

Findings need evidence, and proposed changes stay separate from applied changes. The review shouldn’t silently rewrite the workflow it is investigating. Broader coverage and scheduled reviews are potential extensions to this skill, but they aren’t included in this version.

The same method can inform business operations, but that requires the right records, access, and instructions. It’s not an automatic audit of every app your business uses.

Try: “Review the last two weeks of accessible project sessions. Find the repeated corrections and recommend one improvement for each. Don’t apply changes yet.”

Read full Workflow Insights skill

[Download .md](https://www.field-office.ca/skill-guides/workflow-insights.md)[Open Markdown file for Workflow Insights](https://www.field-office.ca/skill-guides/workflow-insights.md)

## 5\. Onboarding Relationships: Fit setup to the business

A hairdresser, an accountant, and a manufacturer don’t onboard people in the same way. A new customer might need an appointment profile. A new accounting client might need a document checklist. A new supplier might need linked contacts, operating requirements, and purchasing records.

_Onboarding Relationships_ is built around those differences. Download it and adapt it to the relationships your business actually onboards.

It starts with the business context already available in the conversation. Who is being onboarded? What must exist before work begins? Which tools hold the information? It asks only for what’s missing rather than repeating an intake questionnaire you’ve already answered.

Next, it checks available integrations before promising to create anything. Reading a customer record doesn’t necessarily mean your AI can create one. Access to a CRM doesn’t imply access to the scheduling or purchasing system.

It then shows a reviewable plan, checks for duplicate records, and identifies which steps need approval or manual work. Its final report returns useful URLs, confirmed results, errors, and remaining decisions.

The goal is to make setup possible from chat where the connected tools allow it, while keeping the limits visible. Start with one relationship type and a real process your business repeats.

Read full Onboarding Relationships skill

[Download .md](https://www.field-office.ca/skill-guides/onboard-relationships.md)[Open Markdown file for Onboarding Relationships](https://www.field-office.ca/skill-guides/onboard-relationships.md)

## Start with the friction you already know

These skills address different stages of work. _My Voice_ helps you communicate. _Grill Me_ tests the plan. _Take the Wheel_ carries out the permitted task. _Workflow Insights_ helps you improve the process. _Onboarding Relationships_ turns a business-specific setup process into a reusable workflow.

You don’t need to adopt all five at once. Pick one task you repeat, define the result, and compare the outcome with how you work today. Did you need fewer corrections? Were the important questions answered earlier? Did the AI finish the work and make the remaining steps clear?

That’s a more useful starting point than adding another prompt to a collection. Make one recurring task work better, then build from there.

Good luck!
