How many seats does a small support team actually need?

Capacity planning in support used to be one sum: conversations per day, divided by what one person can answer, rounded up, plus cover for holidays. An AI agent breaks the sum in a useful way. The agent’s capacity isn’t hours, it’s conversations. The humans’ capacity is still hours, but the conversations that reach them are the slow ones. So you plan the two separately, and Anchor’s pricing makes the first half unusually easy to count.

The unit is a conversation, not a message

One thing to fix in your head first. In Anchor a conversation is a 24-hour thread with one customer and it costs 10 credits, however many messages it takes. A customer who asks four questions in one sitting is one conversation. So the number you need isn’t messages — it’s threads. Most teams can get it from a week of WhatsApp and website chats: count the customers who wrote in, not the lines.

Step 1: Size the pool

Every seat brings a monthly bundle of credits, and every seat’s bundle pours into one shared pool for the workspace. Two plans, same product, different bundle:

So the seat count is decided by two things at once: how many people need a login, and how many conversations the pool has to cover. A two-person team with 100 conversations a month is two Starter seats — GHS 500 for a 1,000-credit pool. A two-person team with 300 conversations is two Growth seats — GHS 1,200 for 3,000 credits. A one-person shop doing 150 a month is a single Growth seat, GHS 600, and no second login it didn’t need.

The comparison to make is against the pool, not the seat. Two Starter seats and one Growth seat cost roughly the same; the Growth seat buys more conversations and the Starter pair buys a second person. Which one you need is a question about your team, and it’s fine to answer “the person” this month and “the conversations” next.

Step 2: Expect the human share to get slower, not smaller

The mistake teams make after the agent goes live is measuring the humans on volume. The agent takes the where’s-my-order layer, which was fast. What’s left is the duplicate charge, the customer on day 32 of a 30-day return window, the parcel that says delivered and isn’t. Each of those took twenty minutes before the agent and takes twenty minutes after it. So “conversations per person per day” falls, and it should. Plan on the humans handling fewer conversations, each one heavier, and don’t read that as a productivity problem.

The inbox helps you see it honestly: every conversation marked Waiting on human shows a live wait timer, so the number that matters — how long a real problem waits for a person — is visible without a report.

Step 3: Protect the upkeep hour

Part of the humans’ capacity now goes on the agent: reading the Worklists, rewriting the document that keeps escalating, pruning the handoff list. It’s an hour a week for a small team, and it’s the hour that moves the resolution rate, so it’s the last thing to cut when the inbox is busy — and the first thing that gets cut if it isn’t on the calendar. Budget it like a seat.

What happens at zero

Plan for the edge, because Anchor’s is a hard one on purpose. When the pool runs out, the agent pauses. New conversations go straight to your team; the customer sees nothing about billing — the widget stays open and their message still arrives — but nobody is answering automatically until credits are back. There is no serve-now-pay-later. Nothing runs past an empty balance.

Three things keep you off the edge. Monthly credits renew at the start of every billing period (unused ones don’t carry over). Top-ups can be bought at any time at the rate shown at purchase, and top-up credits never expire; they’re spent after the monthly bundle. And auto top-up buys a set number of credits whenever the balance falls below a threshold you choose — the setting to turn on before a promotion, not after. All three are for the workspace owner.

Treat the plan as a set of assumptions

You’ll guess wrong the first month, and the guess should be written down so you can see by how much. The Segments tab in Analytics gives you Volume by channel and Volume by intent, which is exactly the two things the guess was made of: how many conversations, and what kind. Revisit monthly. Seat and plan changes prorate at checkout, so moving from two Starter seats to one Growth seat and back isn’t a decision you have to get right forever.

The checklist

• Count threads for a week, not messages, and multiply up to a month.
• Size the pool to that number, then decide whether the seats are for people or for credits.
• Plan on humans handling fewer, heavier conversations. Watch the wait timer, not the count.
• Put the upkeep hour on the calendar and treat it as capacity.
• Turn on auto top-up before the busy month, not during it.
• Re-read Volume by intent monthly and adjust. Changes prorate.

Where Anchor fits

The figures above are Anchor’s published prices — Starter at GHS 250 and Growth at GHS 600 per seat per month, 10 credits per conversation — and the pricing section has a configurator that does the arithmetic for any seat count. The free trial is 300 credits, about 30 conversations, which is enough to count your first week properly before you size anything.

Common questions