Start with seats for people who will use ChatGPT every week, not every person with an email address. That is the cleanest answer. For most companies, the right first order is 10% to 30% of staff, then expand after 30 days of real usage data.

The simple seat formula

Here is the quick version.

Needed seats = weekly AI users + managers who review AI work + builders who create shared workflows.

That is it. No magic spreadsheet goblin required.

A seat should go to someone who gets real value from ChatGPT Enterprise. Not someone who might try it once because the office Slack got excited. Not someone who says, “Cool, AI,” then disappears into email for six weeks.

Think of seats like gym memberships. Buying 1,000 does not create 1,000 healthy people. It creates 1,000 invoices. Usage is the thing that matters.

Who should get the first seats?

Give seats to teams with repeated text, research, data, or planning work. These people can save time fast.

• Sales: Writing follow-up emails, account notes, call summaries, and proposal drafts.
• Customer support: Drafting replies, summarizing tickets, and spotting common complaints.
• Marketing: Creating briefs, outlines, campaign ideas, SEO drafts, and social posts.
• HR: Writing job posts, policy drafts, interview questions, and onboarding material.
• Legal and compliance: Reviewing long documents, summarizing clauses, and preparing checklists.
• Finance: Explaining reports, building commentary, and checking formulas or assumptions.
• Product and engineering: Writing specs, summarizing bugs, creating test cases, and reading logs.
• Executives: Preparing board notes, strategy drafts, and quick summaries.

Start where the pain is obvious. Boring work is your gold mine. If a team copies, pastes, rewrites, summarizes, or searches all day, give them seats first.

The 10-30-60 rule

This is a simple rollout model. It works well because it stops people from buying too much too soon.

• 10% of employees: Good for a careful pilot.
• 30% of employees: Good when several teams already have clear use cases.
• 60% of employees: Good when AI is part of daily work across the company.

For a 200-person company, that means:

• 20 seats for a tight test.
• 60 seats for a serious rollout.
• 120 seats when adoption is proven.

Honestly, it feels like some companies buy seats the way people buy snacks before a road trip. Too much. Too random. Then half of it sits untouched.

A useful starting chart

Company size	Starter seats	Best fit
50 employees	10 to 15 seats	Founders, sales, support, marketing
100 employees	15 to 30 seats	Core teams with heavy writing and analysis
500 employees	75 to 150 seats	Department rollout with training
1,000 employees	150 to 300 seats	Phased rollout by business unit

This chart is not a law. It is a sanity check. Your numbers may shift if your company is very technical, very document-heavy, or very regulated.

Do not forget the builders

Some users do more than chat. They build prompts, workflows, internal assistants, knowledge tools, and team templates. These people are important.

Give them seats early.

They help everyone else move faster. One strong builder in customer support can create useful reply drafts for 50 agents. One operations person can turn messy process notes into clean checklists. That is not fancy. That is just less chaos.

Pick one builder per major team. If the team is large, pick two or three. Make them the local AI helper. Give them time to test things. Do not make them do it secretly at 9:47 p.m. after their normal job. That is how good projects become sad little folders.

Use activity data after 30 days

Your first seat count is a guess. A smart guess, sure. Still a guess.

After 30 days, check usage. You want simple answers.

• How many users logged in each week?
• How many sent useful prompts?
• Which teams used it most?
• Which teams barely touched it?
• Which tasks saved real time?

If 80% or more of users are active weekly, you probably need more seats. If only 35% are active, stop. Fix training. Improve use cases. Remove seats from ghost users.

Expect to waste time on vague prompts at first. People will type “make this better” and hope for wizardry. It will produce bland soup. Teach better prompting. Show examples. Give teams prompt libraries. This can turn weak adoption into strong adoption fast.

Good signs you need more seats

Add seats when demand is real. Not when someone says AI is “the future” in a meeting and everyone nods like a dashboard dog.

Good signs include:

• Teams are sharing prompt templates without being asked.
• Managers ask for more access for their teams.
• Users report saving 2 to 5 hours per week.
• Support replies get faster.
• Sales notes become cleaner.
• Marketing briefs stop taking half a day.
• Employees complain because they do not have access.

That last one matters. If people are annoyed they cannot use it, you have organic demand. That is the good kind of problem.

Bad signs you bought too many seats

Too many seats create budget fog. Nobody likes that.

Watch for these signs:

• Many users log in once and never return.
• Teams cannot name three useful use cases.
• Managers do not know who uses it.
• Training attendance is low.
• People use free personal tools instead.
• No one tracks time saved or work improved.

If this happens, do not panic. Reduce the next purchase. Reassign seats. Run a better enablement session. Make examples specific to each team. Generic AI training is where enthusiasm goes to nap.

Should every employee get a seat?

Sometimes, yes. Usually, not at first.

Company-wide access makes sense when ChatGPT is part of daily work. It also makes sense when security rules are strict and you want people using approved tools only. If employees are already pasting work into random AI tools, a wider Enterprise rollout may reduce risk.

But if half the company rarely writes, researches, analyzes, or creates documents, start smaller. A warehouse worker, field technician, or retail associate may still benefit. But they may need shared workflows, mobile access, or manager-led tools more than a full seat right away.

A practical rollout plan

Use this four-step plan.

1. Pick high-value teams. Choose teams with clear tasks and frequent pain.
2. Start with 10% to 30% of staff. Keep the first group focused.
3. Train with real work. Use actual emails, reports, tickets, and briefs.
4. Review after 30 days. Expand, reassign, or pause based on usage.

Do not make rollout a giant ceremony. Make it useful. Give people five tasks they can do this week. Then ask what saved time.

The final seat answer

You need enough ChatGPT Enterprise seats for your active weekly users, plus a small group of managers and workflow builders. A good first purchase is usually 10% to 30% of your headcount. Expand when usage proves the need.

Buy for behavior, not hope. Start focused. Track activity. Move seats away from silent users. Add seats when teams can show saved time, faster output, or better work.

That is the boring answer. It is also the one that saves money.