Skip to navigation
Other

Images

Overview

Your AI Agent can read images that end users send and use them to resolve inquiries. A vision model produces a text description of each image, and the Agent reasons over that description. The Agent never receives the image itself.

Supported formats and channels

The AI Agent reads images on these channels:

  • Messaging
  • Email
  • SMS and MMS, through Twilio
  • WhatsApp, through Twilio or through Sunshine Conversations
  • Instagram, connected directly to Ada or through Sunshine Conversations
  • Facebook Messenger and Twitter DM, through Sunshine Conversations

The Agent also reads images on the Sunshine Conversations Web Messenger. This is available on request. Contact your Ada team to enable it.

See Messaging conversation guidance for image and handoff-upload behavior.

  • Formats on the Messaging SDK and Email: PNG, JPEG, GIF, HEIC, and HEIF. WebP is read on the Messaging SDK only. AVIF, SVG, PDFs, documents, and video are not read. Format support on other channels differs. This page does not list it.

  • HEIC and HEIF: on the Messaging SDK and Email, the Agent converts each HEIC or HEIF image it reads to JPEG. The JPEG is stored and is passed to human agents on handoff. Images wider or taller than 4096 pixels are scaled down during conversion. Images over 50 megapixels are not read. An image the Agent does not read keeps its original format, for example the fourth image in an email.

  • Size: on the Messaging SDK, each image can be up to 50 MB. On Email, the whole message, including every attachment, can be up to 30 MB. On SMS and MMS and on WhatsApp through Twilio, each image can be up to 10 MB. On Instagram connected directly to Ada, each image can be up to 25 MB. Channels through Sunshine Conversations have no Ada-side size cap.

  • Messaging SDK: the AI Agent reads each image an end user uploads. In the hosted chat window, an end user can attach up to 3 images to one message. The Agent replies once for each image. Image uploads are off by default. The Enable image uploads to your AI Agent toggle controls them. The Agent does not read images from legacy Chat. To let end users upload images:

    1. Go to Config > CHANNELS > Messaging.
    2. On the Appearance tab, open the General drawer.
    3. Enable Enable image uploads to your AI Agent.
    4. Click Save changes.

    On the legacy Ada Web Chat, the toggle is locked because image uploads are not available there. Contact your Ada team to move your AI Agent to the Messaging SDK. This toggle does not change file uploads to human agents.

  • Email: the Agent reads up to 3 images per email. The inbox’s attachment handling setting decides whether it reads them:

    • Hand off to a human (default): every email with an attachment is handed off. No image is read.
    • AI Agent replies without the attachment: attachments are discarded. No image is read.
    • AI Agent replies using the attachment: images are read, and every accepted attachment is kept for a later handoff.

    AVIF and SVG attachments are never read. They count as non-image attachments.

What the Agent reads

The AI Agent receives a description of each image. It includes all visible text: order numbers, error messages, codes, dates, and amounts. It also includes the visible content: products, damage, packaging, and the screen shown in a screenshot. Text inside an image is treated as end-user input, never as instructions.

Sensitive images

An image is withheld when the AI Agent detects any of the following anywhere in it. The Agent learns only the document type, not its contents, and the filename is dropped.

  • A payment card with any part of the card number visible
  • A government ID document: passport, driver’s license, national or state ID card, or residence permit
  • A credential document that is not a government ID, such as a social security card or a bank statement that shows full account numbers

A receipt or order confirmation that shows only the last 4 digits of a card is read normally. A full card number anywhere in an image is withheld.

Detection is automatic and best-effort. A card number that detection misses is still masked by the built-in payment card redaction before the Agent sees the description. Any other text that gets through is handled like typed text, including your Redaction settings.

Payment cards are always withheld. You choose whether the Agent withholds the other two types under Image redactions. Both are enabled by default.

By default, the Agent tells the end user that it did not read the image, and names the document type, for example “your passport or ID”. It does not ask them to resend the image. The Agent then gives one next step. For a payment card, the Agent may ask for the last 4 digits as text. For an identity document or other credential, the Agent does not ask for the number. It offers a human agent when a handoff is available. When no handoff is available, it says that human support is not available and keeps helping without the number. While a Playbook defers handoffs, or after the Agent has handed off the same issue, it does not offer a human agent and keeps helping without the number.

In the dashboard, a withheld image appears as [Image withheld: sensitive content]. The Agent does not read a withheld image. On Email, the withheld image still goes to the human agent on handoff, like any other accepted attachment.

Redactions

Redaction settings apply to image descriptions the same way they apply to typed messages.

Retention

Images that Ada stores are deleted 7 days after upload, consistent with other attachments. Ada does not store an image that the AI Agent reads on a channel connected through Sunshine Conversations. The text description remains part of the conversation transcript.

Unreadable images

If an image cannot be read, the AI Agent tells the end user and asks them to describe it or send it again.

Have any questions? Contact your Ada team, or email us at .