Bring AI skills to your team: a proposal template for your manager

Why a pilot, not a purchase

Your team is probably using AI already. Microsoft and LinkedIn's 2024 Work Trend Index reported that 75% of knowledge workers used AI at work, that 78% of AI users brought their own AI tools, and that only 39% of people using AI at work had received AI training from their company. The same report found 66% of leaders said they wouldn't hire someone without AI skills.

So the question for your manager isn't whether people use AI. It's whether the team's best ways of using it are shared, or locked in each person's own chat history. A pilot with shared skills answers that with numbers, in two weeks, and showing you can run one is a career skill in its own right.

Before you write: pick the three tasks

1. List ten tasks the team repeats at least weekly: reports, replies, summaries, first drafts.
2. Score each on how often it happens, how long it takes, how clear the rules are, and how bad a mistake would be.
3. Pick three that score high on time and clarity and low on risk.
4. Time each once as it's done today, so you have a before number.
5. Find the person who does each best, and ask them to describe their steps; that becomes the skill.

The template

Subject: Proposal: a two-week pilot of shared AI skills for [team]

Problem
Our team spends about [X] hours a week on [task 1], [task 2] and [task 3].
Each of us does them differently in our own AI, so quality varies and
good approaches don't spread. [One example: a client report redone, a
reply that went out wrong.]

Proposal
For two weeks, [names, 3-6 people] use the same three AI skills for these
tasks. Each skill is our best person's method written as steps our AI
follows: [task 1 owner], [task 2 owner], [task 3 owner]. Everyone keeps
their own Claude or ChatGPT; the skills reach them through one link.

Cost
[Free plan / N seats x price per month]. My time: about [4] hours to set
up the skills, [1] hour a week to collect results.

How we'll measure it
- Time per task, before and after (we timed each once this week).
- [Errors or rework]: how many outputs needed redoing.
- A test set of [20] real examples per skill, run before and after
  any change.

Risks and how we'll handle them
- Confidential data: we follow [our AI policy]; no [customer data / X]
  goes into the skills.
- Wrong output: every output is checked by the person sending it.
- Lock-in: the skills are plain text in an open format we keep.

Decision
On [date], I'll share the numbers and recommend continue, change or
stop.

Filling in the cost line

For a pilot of up to 3 people and 10 skills, AI Skill Pod's Free plan costs nothing. A new team also gets Business free for 14 days. After that, Team is $5 per seat a month from 3 seats, and Business is $15 per seat from 5. For comparison, Claude Team Standard is $25 per member a month billed monthly (claude.com/pricing, read 4 November 2026), and it shares skills only inside Claude.

People keep the AI they already pay for: Claude on plans with custom connectors, ChatGPT on Pro, Business or Enterprise, and Claude Code, Codex, Cursor and OpenCode. OpenAI doesn't list ChatGPT Plus for custom MCP connectors. A Plus user would paste the skill text instead.

Measuring it honestly

Time saved is the number managers ask for, and the easiest to inflate. Time the task once before, and the same task once after, by the same person. Add a quality check: Anthropic's eval guide suggests 20 to 50 simple test cases drawn from real failures as a starting set. Twenty past examples per skill, checked before and after, show whether the output got better, not just faster.

Common objections, and short answers

• "We already have AI." Yes; this shares the team's best use of it so everyone gets the same result.
• "Is our data safe?" Skills hold instructions, not customer data. The pilot follows the company's AI policy, and nothing confidential goes into a skill.
• "Who maintains it?" Each skill has one owner. If the pilot stops, the files stay ours.

Questions

How do I convince my manager to adopt AI tools?

Propose a small, measured pilot instead of a tool: three tasks the team repeats every week, two weeks, a number for time saved or errors avoided, and a decision date. Managers approve experiments with a clear stop more easily than open-ended purchases.

What should an AI pilot proposal include?

The problem in your team's own numbers, the tasks you'll test, who takes part, the cost, how you'll measure success, the risks and how you'll handle them, and the date you'll report back.

Do we need to buy new AI subscriptions for a pilot?

Usually not. Shared skills work in the Claude and ChatGPT accounts people already have (ChatGPT Pro, Business or Enterprise; Claude on plans with custom connectors), and a free plan covers a small pilot.