AI Virtual Assistant Services: How a VA Business Can Package Them

A VA business can sell AI as fixed-scope monthly packages, such as inbox triage, meeting write-ups or report drafting, in three tiers. Price each tier at the human hours it still takes, plus a 15 to 25 percent review buffer, times the hourly rate you want to earn. Never price from the cost of a ChatGPT subscription.

The trap is hourly billing. If AI cuts a ten-hour inbox job to four hours and you still bill by the hour, your revenue drops by 60 percent while the client receives the same result. Packaging fixes that, because the client buys an outcome (an inbox cleared by 9 a.m., a report in their folder every Monday) and you keep the efficiency as margin.

Follow me on Instagram@sagnikteaches

Why the tool's cost is the wrong place to start

A paid assistant seat costs a small fraction of one VA hour. ChatGPT Business lists at $25 per user per month billed monthly or $20 billed annually, with a two-seat minimum, so a solo VA either pays for two seats or uses an individual plan instead. Whatever you choose, the software is a rounding error next to your own time.

Connect on LinkedInSagnik Bhattacharya

So pricing from tool cost leads to silly answers: "it costs me $25, so the AI part is worth $25." Clients do not pay for the tool. They pay for three things you carry: the process that makes the output usable, the review that makes it safe, and the accountability if it goes wrong. Those are the same things they paid you for before AI arrived, which is why a VA's price per finished item should not collapse just because the first draft is faster.

Subscribe on YouTube@sagnikteaches

If you want typical rate bands and a wider look at hourly versus retainer billing, the tutorial on virtual assistant pricing when AI does the drafting covers it.

One client's bill, hourly versus packaged

Every figure here is illustrative; use your own rates. Take a VA who bills $30 an hour and has a client on a monthly inbox-management arrangement of 10 hours.

The client saves $40 a month in the packaged version and gets faster triage and a written summary every morning. You earn $256 for four hours of work after the tool share. Neither side is being cheated: the client pays less than before for more, and you are paid for the judgement that makes the AI draft safe to send. The fee came from a simple rule: charge somewhat less than the old total so the client feels the benefit, but price against the value of the outcome rather than the minutes.

The same sum with hourly billing after AI shows why VAs get squeezed. Working four hours for $120 means the client captures every minute saved. Packaging is how you stay in the game.

Choosing what to package

Package work that is repeatable, has a clear "done", and where a human check takes minutes rather than hours. Work that needs deep judgement about people or money stays on an hourly or project basis. Seven candidates that fit, with where the human time goes:

• Inbox triage and draft replies. AI sorts and drafts; you approve, edit tone and send. Review is the main time cost. The tutorial on AI email management for virtual assistants sets out the rules and handover.
• Meeting notes to action lists. Transcript in, summary and owners out. You check names, dates and anything that sounds like a decision that was not actually made.
• Weekly or monthly reporting. Pull figures into a template, draft the commentary, and verify every number against the source.
• Research briefs. A one-page summary of a topic, competitor or supplier set with sources you have opened yourself.
• Content repurposing. One webinar or newsletter becomes five social posts and a short email, in the client's voice, approved before scheduling.
• CRM and data hygiene. Deduplicating, tagging and updating records from forms or emails, with a sample check each week.
• Travel and event admin. Itinerary drafts and confirmation emails, with a human booking every actual reservation.

A useful screen: if the sentence "the client would not notice if the AI got this wrong" is true, the task is too low-stakes to be worth a premium. If the sentence "a mistake here could cost the client money or a relationship" is true, you can package it, but the review step has to be explicit and priced in.

A three-tier menu you can copy

Three tiers is enough. Fewer looks like a single take-it-or-leave-it price; more makes the client do your sorting. Use the tiers below as a starting layout, with figures you can swap for your own.

Notice how the tiers differ: volume caps, turnaround and review depth, not "more AI". Clients understand limits on messages and meetings. They do not understand tiers named after models.

Naming the tiers

Skip clever names. Use the job: "Inbox Essentials", "Inbox and Meetings", "Inbox, Meetings and Reporting". A client skimming a proposal should see what lands in their week. Avoid tier names that imply guarantees, such as "Priority", without a stated turnaround, because the first missed deadline turns the word into a complaint.

Pricing logic in four steps

1. Time a real sample. Run five typical weeks of the task with your AI workflow and log human minutes only. Use the median, not the best week.
2. Add a review buffer. Add 15 to 25 percent for the week an AI draft is wrong in a way that costs you an hour to unpick. Setup time for the first month goes in a separate one-off fee.
3. Convert to your target rate. Multiply buffered hours by the hourly rate you want to earn on that work, not the rate you used to charge.
4. Test it against the client's alternative. The client's alternatives are doing it themselves, hiring a part-timer or buying software. If your price is within what their own time costs, it sells without a discount.

A short worked pass for meeting write-ups. The median is 22 human minutes for a one-hour meeting, and a 20 percent buffer takes that to about 26 minutes. Six meetings a month is 2.6 hours. At a target of $55 an hour that is $143, which you round to $150 as the meeting line inside the Standard tier. Compare that with a client writing the notes personally: six hours of their own time a month, which is worth far more than $150 to most founders, so the line sells.

The setup fee

Charge a one-off onboarding fee for the first month. It covers access, writing the voice guide, building the prompts and agreeing the review rules. A fee of one to two times the monthly price is a reasonable starting range to test; the point is that setup is real work that does not repeat. Clients who balk at it often turn out to be clients who would also balk at scope limits.

Scope limits that protect your margin

A package without limits is an hourly contract with a flat fee, and AI makes it worse because clients quickly assume a fast tool means unlimited requests. Put these in writing:

• Volume caps. Messages, meetings, reports, posts. Say what happens at the cap: overage per item or an upgrade offer.
• Revision rounds. One round included, further rounds as an add-on. AI makes it easy to generate "just one more version", and clients ask for it.
• Turnaround windows. Working hours, time zone as agreed, excluding weekends unless stated.
• What you will not do. Send anything that commits the client to a price, a legal position or a refusal without their explicit approval. List it in the agreement, not in your head.
• Source of truth. The client's systems hold the real data. If the AI summary disagrees with the source, the source wins and you flag it.
• Requests outside the list. Anything that falls outside the listed deliverables is quoted, not absorbed.

An illustrative mistake shows why. Picture a VA who offered an unlimited "quick drafts" benefit on a mid-tier package. Within a month the client was sending forty one-line requests a day, each of which looked trivial but needed a read, a prompt, an edit and a send. The VA's four-hour estimate became eleven. The fix was to cap drafts at 120 a month, keep a $2 overage rate per draft, and keep the same fee. The client's usage fell to 95 once they could see the counter.

Client disclosure: what to say and where

Disclose in three places: the proposal, the service agreement and the first onboarding call. The wording should be plain. A filled-in example you can adapt:

How we use AI in your package

- We use AI tools to draft first versions of replies, meeting
  summaries and reports. A person at [business name] reads and
  approves every item before it reaches you or anyone outside
  your business.
- The tools we use are listed in your onboarding sheet. We will
  tell you before we add or change a tool that handles your data.
- We use business-tier accounts that do not train on your
  content by default. We do not paste passwords, card numbers or
  full identity documents into any AI tool.
- You can ask us to switch AI off for any message, client or
  document. We will do that task manually and tell you whether
  the time or price changes.
- If we find an error that came from an AI draft after it was
  sent, we will tell you the same working day and correct it.

Two things in that wording matter more than the rest. "A person approves every item" is a promise, so only write it for the tiers where it is true (see the review row in the table above). And "business-tier accounts that do not train on your content by default" should be checked against the vendor's current terms for each tool you use, because defaults and plan names change. For a deeper checklist, the tutorial on VA confidentiality when using AI goes through consent, redaction and account set-up step by step.

Whose account holds the data

Decide this per client. Using your own paid account is simpler but puts client material in your workspace. Using an account the client owns keeps ownership clear but means you work inside their settings. Whichever you choose, write it down. A client with sensitive records, such as a care agency or a veterinary practice holding owner details, may need to own the account and keep AI features off for anything with personal details.

How the packages look in four different businesses

The packages flex to the client's sector; the structure stays the same.

A music school sends many parent emails about lesson changes. The Standard tier works well: one inbox, custom drafts from a library of approved phrasings, and a weekly summary of cancellations and make-up lessons. The review step matters because a draft that tells the wrong family their child's slot has moved is a real complaint.

A care agency needs far more caution. The package covers rota-confirmation emails and recruitment enquiries only, with a hard rule that no care-recipient information goes into an AI tool. The scope sheet says so in one line, and the price stays modest because the allowed task set is narrow.

An optician's practice wants reminders and review requests drafted after appointments. The AI writes message variants; the VA schedules them in the practice's own system and removes anyone flagged as opted out. The package is priced per 100 messages, because that is the unit the practice can count.

A veterinary practice wants after-hours enquiry summaries ready at 8 a.m. The package is a morning digest: enquiries grouped as urgent, booking and information, with a draft reply for each. Clinical advice is explicitly out of scope, and every urgent item is phoned or flagged to a vet rather than answered by text.

Presenting it: a filled-in package page

A package description that sells is short and specific. Here is the Standard tier written out, as illustration:

Inbox and Meetings - $420 a month

You get
- Two inboxes triaged every working day by 9 a.m.
- Up to 120 custom draft replies a month, each approved by
  a person before sending
- Notes and action lists for up to 6 meetings, delivered
  within 4 working hours
- A weekly one-page summary of open items

Limits
- 900 messages and 6 meetings a month; extra items quoted
  or billed at the add-on rate
- One round of edits per deliverable

Setup
- One-off onboarding fee of $600: access, voice guide,
  review rules and a two-week trial run

Notice what is absent: any mention of which AI model you use, how clever it is, or how many hours you save. Clients care about the Monday morning experience.

Moving existing hourly clients across

Do this client by client and say why. A before-and-after of the message that goes out:

Before: "Hi [first name], I'm moving to package pricing, and your rate will be $380 a month from next month." That sounds like a price rise with no reason.

After: "Hi [first name], over the last quarter you've averaged 11 hours a month with me, mostly on inbox and meeting notes. I've changed how I do those so you get the notes the same day and a morning inbox summary. I'd like to move you to a fixed monthly package at $380 that includes both, with a cap on volume that you're well inside. You'd no longer be billed for hours, and the first month runs alongside your current terms so you can compare. Shall we talk through it on a ten-minute call?"

The second version names the client's past usage, a benefit, and a trial period. It makes the move a service upgrade rather than a quiet rate change.

Five numbers that show a package is paying

After the first four weeks, look at five numbers per package:

1. Average human minutes per deliverable against the minutes you priced. Over by more than 20 percent means tighten the limits or raise the price at renewal.
2. Edit rate. What share of AI drafts you changed materially. A high rate means your prompts or voice guide need work, not that the client needs to pay more.
3. Error escapes. How many mistakes reached the client or a third party. Target zero and treat every one as a process fix.
4. Cap usage. Clients routinely at 95 percent of a cap are ready for the next tier; clients below 40 percent may be on the wrong one.
5. Effective hourly rate. Fee divided by logged human hours, after the AI tool share. Below your old hourly rate means the package is underpriced or over-delivered.

A quick illustration. A client on a $420 Standard package where you logged 6.2 hours has an effective rate of about $68 an hour, which is healthy. If the same client needs 11 hours, the rate falls to $38, close to your old hourly rate, and either scope or price has to move.

When a package is the wrong answer

Use hourly or project pricing when the work is unpredictable (launching a new product, clearing a three-year backlog), when the client's volume swings widely week to week, or when the client demands bespoke AI on their confidential material and will not agree to limits. Packages also fit badly where the client wants you on call all day. Use quoted service levels for that kind of role instead; the tutorial on chat support VA software covers the tool side.

Packaging is also not a way to hide AI from clients. If the pitch only works when the client does not know a tool is involved, rewrite the pitch. For the wider question of how VAs stack tools across several clients, the tool-stack tutorial pairs well with this one, and how virtual assistants use AI to manage more clients covers capacity planning once the packages sell.

Launching in two weeks

A realistic order of work for a solo VA:

1. Days 1 to 3. Pick the three tasks you do most often for clients. Time them under your AI workflow for a week.
2. Days 4 to 6. Write the tier table and limits. Price using the four steps above.
3. Days 7 to 8. Write the disclosure wording and add it to your service agreement. Have a qualified adviser look over the agreement wording, since limits of liability and data clauses are legal matters.
4. Days 9 to 10. Offer the packages to your two friendliest existing clients on a trial month.
5. Days 11 to 14. Read the time logs, fix the limits, and publish the menu on your site.

If you would rather work out the packages with someone who has seen how small service businesses wire AI into their day, that is what my AI implementation consultation covers.

Questions VA owners ask before launching AI packages

Should I hide that AI drafts the work?

No. Tell clients which steps use AI, which person reviews the output and where their data goes. A client who finds out later is likely to assume the worst, while one told up front can ask about turnaround and price instead. Put the disclosure in your service agreement and repeat it in each package description so nobody has to hunt for it.

Do I charge less because AI does the first draft?

Not automatically. Charge for the finished, reviewed result and the responsibility you carry for it. Your price can fall per item if volume rises or if you want to win clients, but cutting it because a tool is involved just moves your saved time into the client's pocket. Work the sums with your real review minutes first.

Which tier should I push first?

The middle one. Build the entry tier so it is clearly limited and the top tier so it is clearly roomy, then price the middle tier at the level you actually want most clients on. Watch what people pick in the first ten sales conversations and adjust the limits, not the prices.

What if a client wants to supply their own AI tool?

That can work if you write it into scope. Agree who pays for the tool, whose account holds the data, and that you will work inside it under their settings. Price the package for your labour only, and note that you cannot vouch for settings you do not control.

How long before I know a package is profitable?

Four weeks of time logs is usually enough for monthly retainers. Record minutes per deliverable for every client on the package, compare the average against the minutes you priced in, and adjust limits if the average runs more than 20 percent over.

Further reads

• How to Choose a Virtual Assistant Agency: Questions and Prices — See what buyers ask when comparing VA providers.
• AI Assistant vs Virtual Assistant: Which Should You Pay For? — The buyer's side of the same decision.
• ChatGPT Prompts for Virtual Assistants: Inbox, Scheduling, Reports — Prompts you can run inside the packages.
• Virtual Assistant for QuickBooks: Intuit Intelligence or a Human VA? — A bookkeeping-adjacent package and where AI stops.
• Virtual Assistant or AI for a Real Estate Agent: What Goes Where — A niche example of splitting tasks between human and AI.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT Business pricing and plan pages; the rest is working guidance with illustrative figures.