What AI usage is
Your AI agent runs on Xconversion AI. What it does for your customers is counted in AI credits, and your workspace holds a balance of them. The AI usage screen shows both: what the agent used, and the credits you have left.
Open it from AI usage in the menu, or go straight to it.
What uses a credit, and what does not
| What happens | Credits | On the screen as |
|---|---|---|
| Your agent sends a reply to a customer. | One credit | Replies |
| A customer sends a voice note or a picture, and your agent replies to it. | One credit for the reply: listening and looking are part of it | Replies |
| You try your agent in its preview: each answer. | One credit | Studio previews |
| A page or a document is read into your catalogue: each part of it. | One credit | Catalogue imports |
| Your agent writes a reply, then hands the conversation to a person instead of sending it. | None | Not shown |
| Your agent looks something up in its knowledge, or reads a document or a website you add to it. | None | Not shown |
| An answer fails, or is called off. | None | Not shown |
| You ask Xconversion AI a question in the Help panel. | None | Not shown |
| Your team writes a message, or a template goes out from the inbox, a campaign or a workflow. | None: no AI is used | Not shown |
A reply is one credit however long it is, and however many tries it took to write. The tokens on the screen are what the agent read and wrote to do it, for you to follow; they are not what you are charged.
Your credits
The first card is AI credits: the credits your workspace has now, in whole credits. It is shown to the people who handle the account. Everybody else sees the screen in tokens: see If you do not see credits.
| On the screen | What it does |
|---|---|
| AI credits | The credits left. A balance below zero has a minus sign: it happens when the agent goes on answering for a little while after the credits are used up. The badge beside it says where you stand: see When credits run low. |
| Buy credits | Takes you down to Buy AI credits on the same screen. See Buy AI credits. |
Under the balance, a line says how many of the credits your plan includes this month are left, and the day they end.
Choose the period
Everything under the balance is for one period: the last 30 days, until you choose another.
Press 7 days, 30 days or This month.
The figures, the charts and the history change at once. The first and last day of the period are written under This period.
For other dates, fill the two date boxes and press Show these dates.
The first box is the first day, the second is the last. A range ends today at the latest and covers a year at most. If the dates cannot be used, for example the first is after the last, the screen shows the last 30 days.
Days are counted on your workspace’s time zone, set in Workspace settings. The period is part of the page’s address, so a view can be bookmarked or sent to a colleague.
The figures
| On the screen | What it does |
|---|---|
| Credits used | The credits used in the period: the replies, the previews and the catalogue parts. |
| Tokens in / out | What your agent read, and what it wrote, in tokens. Large numbers are shortened: M is millions, K is thousands. |
| AI conversations | How many conversations your agent answered, or listened to a voice note or looked at a picture in. |
| Per conversation | The credits a conversation used, on average, in the period. |
A figure with nothing behind it says No data, never zero.
Day by day
| On the screen | What it does |
|---|---|
| Credits by day | A bar for each day: the credits used that day. Over the bars are the period’s total, the average a day and the busiest day. The dotted line is the average. |
| Tokens by day | A bar for each day, in two parts: In is what the agent read, Out is what it wrote. |
Point at a bar, or tap it on a phone, to see its date and its exact figure.
What the credits went on
| On the screen | What it does |
|---|---|
| By what the agent did | The credits split between Replies, Voice notes and pictures, Studio previews and Catalogue imports. |
| By WhatsApp number | The credits used on each of your numbers. No number is what belongs to no number, such as the previews. |
| By agent | The credits each of your AI agents used. No agent is what an agent that has since been deleted used. |
Each line gives its share of the period’s credits and how many, the largest first.
The conversations that used the most
Conversations that used the most credits lists the ten conversations that used the most credits in the period: who the conversation is with, the tokens read and written, and the credits. Press a line to open the conversation in the Inbox.
The list only holds conversations you are allowed to open.
The rest of the screen
- AI credits history, every credit in and out for the period: see Top-ups, receipts and history.
- Buy AI credits, to buy more: see Buy AI credits.
- Low-credit warning, the line under which your admins are warned: see When credits run low.
- Top-ups, every purchase of credits and its receipt: see Top-ups, receipts and history.
The credits your plan includes
A plan includes a number of AI credits every month. How many depends on your plan: each plan’s card says it, on the plans’ own screen. See Plans and what you pay for.
- As each calendar month begins, on your workspace’s time zone, the month’s credits are added. In AI credits history it is the line Credits included in your plan.
- They are used first, before the credits you bought.
- The AI credits card says how many are left, and until which day.
- When the month ends, what was not used is taken back: the line Included credits not used last month. They do not carry over. Credits you buy never expire.
When credits run low, and when they run out
The badge beside AI credits says where you stand.
| On the screen | What it does |
|---|---|
| In good shape | Enough for now. |
| Running low | Your credits are under your warning line, or at your recent use they run out within a few days. |
| Used up | The credits have reached zero. Your agent keeps answering a little longer, and the balance goes below zero, until you buy more. |
| AI paused | That little extra is used too, and your agent has stopped. |
When your credits fall under the warning line, the workspace’s admins are emailed, the bell says so, and a banner shows on every screen to the people who handle the account, with Top up. It warns once each time: again only after your credits have gone back above the line.
To set your own line, type a number of credits in Warn below (credits) under Low-credit warning, then press Save.
Leave it empty to use your plan’s line, which the hint under the box gives. The card shows once your workspace has had its first credits.
While the AI is paused:
- Your agent answers nobody, listens to no voice note and looks at no picture.
- Customers’ messages still arrive. Each conversation the agent would have answered goes to your team in the Inbox, with the note
AI paused: no AI credits left. - Everybody in the workspace sees a banner saying that AI replies are paused. The people who handle the account also get Top up on it.
- The agent’s preview is paused as well, and says why.
As soon as credits are added, your agent answers new messages again. A conversation that went to your team while the agent was paused stays with your team until somebody hands it back: see Assign and hand over.
A daily limit
Each plan also has a limit on how much your agent may do in one day, to keep a mistake or an unusual day from using all your credits at once. You do not see it on the screen. When it is reached, your agent hands each new conversation to your team for the rest of the day, and the conversation says why: the day’s AI limit was reached. The next day, your agent answers again by itself.
If this happens often, write to support: your plan may be too small for your volume.
If you do not see credits
Credits are shown only to people whose access includes The workspace and billing. Everybody else who can open AI usage sees the same screen in tokens and conversations, with a line saying so.
- No balance, no credits used, no way to buy credits, no history.
- Tokens in / out and AI conversations for the period.
- Tokens by day.
- Conversations that used the most: the ten conversations that used the most tokens, among those you may open.
Access is given on the Team screen: see Roles and access.
Good to know
- Who can open the screen: anybody whose access includes AI usage, which every role has unless it was taken away. Credits need The workspace and billing, which an org admin has.
- The screen keeps itself up to date: the figures move as your agent works, without reloading.
- A new workspace gets its free credits the first time its AI agent is used. Until then the screen says so, and shows tokens.
- If the screen says AI billing not active yet, nothing is taken for your agent for now. The screen still shows what it used.
- A workspace that is not billed for AI is told so on the screen, and has nothing to buy.
- Your plan payments and credit purchases are listed with their receipts on Plan & billing: see Payments, receipts and billing details.
Questions
What is a token?
The unit AI text is counted in: a short word, or a piece of a longer one. What your agent reads counts as tokens in, and what it writes as tokens out. You are charged in credits, not in tokens.
Does a long reply use more credits?
No. A reply sent is one credit, whatever its length and whatever the customer sent.
Does my agent stop the moment my credits reach zero?
No. It keeps answering a little longer while the badge says Used up. It stops when that is used too, and the badge then says AI paused.
Do the credits I buy expire?
No. Only the credits your plan includes each month end with the month.
Is what Meta charges for WhatsApp messages in these figures?
No. This screen is only your agent’s AI usage. Meta bills its own charges to you directly: see What WhatsApp charges.