An AI agent that sends photos and videos: when a file closes what text cannot
Contents
The common mistake is to think that photos and videos belong to marketing, not to selling. The most persuasive material a business has, a before-and-after album, a video tour of an apartment, a shot of the room with the view, sits on Instagram or in the owner's phone gallery. And inside the conversation itself, the place where the customer decides, they get text and a link: "You can see our work on Instagram." It looks reasonable, because the material is already there and there is no reason to duplicate it. But the link takes the customer out of the conversation exactly when they became interested, and leaves them to find, on their own, the picture that answers their question. An AI agent that sends photos and videos does not improve the look of the answer. It brings that material back to the place where decisions are made. This article is about when and why: at which moments in a conversation a file does the work of a whole paragraph, what goes into the library, what not to send, and what happens the other way round, when the customer is the one sending.
Why "check out our Instagram" is an expensive answer
The customer asked something specific: what lips look like a week after a filler, whether the balcony faces the sea, whether the double room is really as big as in the main photo. A link to a profile answers a general question. To find their answer, the customer has to scroll through two hundred posts, half of them holiday greetings and offers that have ended. Some do. Many write "thanks, I'll have a look", and the conversation ends there. No report shows it, because nobody counts conversations that ended with "I'll have a look".
The second problem is timing. When the owner replies in the evening "I'll send photos tomorrow", the material arrives once the interest has cooled, and sometimes after the customer has seen photos from another business that answered faster. This is the difference between material that exists and material that is available: the best photo in the world does not help if it is sent the next day.
And the third: a file sent inside the conversation stays there. The customer can show it to a partner, come back to it in the evening, forward it to a friend. A link to a profile stays in nobody's memory. And once the picture is there, the customer's next message is usually "nice, when can I come?", which is the next step in the sale rather than another question.
Four moments when a file does the work of a sentence
When the question is "what does it look like"
This is the obvious moment. No verbal description of a treatment, an apartment or a room matches one good photo. The agent recognizes the question, even when it is worded differently every time, and sends the photo that fits it, not the whole album. The article on an AI agent for aesthetic clinics shows how many of the questions in such a clinic are really a request to see.
When the price sounds high
A customer who gets a price without context compares it with the lowest price they have seen. A before-and-after photo, a short video of the work process or a shot of the materials used gives the difference a reason. The agent does not haggle and does not drop the price. It shows what the money pays for, then asks the question that moves things forward: check a slot, or hand over to a person. This is how an agent that qualifies leads and moves them forward works: media at the right moment turns a qualifying chat into a convincing one.
When the question is about logistics
"Where exactly is the entrance?", "Is there parking?", "What should I bring?". A ten-second video from the street to the door saves a phone call on the day of the appointment and a fifteen-minute delay. A preparation document sent right after the booking saves the question that would otherwise arrive the day before.
When the customer needs to compare
A menu, a table of room types, a model comparison, a course syllabus. Here one tidy PDF beats five text messages, because the customer sees all the options together and chooses. The agent sends the file and then helps choose, based on what the customer has already said in the conversation.
What to send and when, by type of business
| Type of business | The question in the chat | What the agent sends | What it saves |
|---|---|---|---|
| Aesthetic clinic | "What does it look like after?" | Before-and-after photos chosen by the clinic | A long explanatory phone call |
| Real estate | "Which way does the balcony face?" | Photos of the property and a short tour video | A viewing of a property that does not fit |
| Guesthouses and boutique hotels | "What does the room with the jacuzzi look like?" | Photos of that specific room and of the view | A separate exchange about every room |
| Renovations and contractors | "Have you done a kitchen like this?" | Similar jobs from the portfolio | An assessment visit to a customer who does not fit |
| Shop | "Do you have it in blue?" | A photo of the product in that color and a catalog page | Repeated questions about every item |
| Courses and workshops | "What do you actually learn?" | A PDF syllabus and a short clip from a lesson | A consultation call about the basics |
What every row has in common: the file answers a question that was asked, and is not sent "just in case". In real estate, for example, a tour video sent before a viewing is arranged filters out the person who would have come and found the kitchen too small, which saves time on both sides. We describe how such an agent differs from an ordinary listing page on the AI agents for real estate page.
The other direction: what the customer sends
Sending is half the picture. Customers send businesses more and more files, not only text. A voice note instead of typing. A photo of what broke. A screenshot of an error message. Another business's quote with the question "can you do it cheaper?". A short video of a strange noise in the engine.
In a business without an agent, files like these wait for someone to open them, and when there are dozens of conversations a day, some get lost. An agent that understands them answers on the merits: it hears what was said in the voice note and carries on from there, recognizes what the photo shows, reads the document and answers from what it says. The file itself is saved on the customer's record in the CRM, so whoever picks up the case sees it in the right place and does not have to scroll back through the chat.
And there is a clear limit. When a patient sends a photo of redness the day after a treatment and asks "is this normal?", the agent does not assess and does not reassure. It writes that a member of the team will check, and hands the conversation to a person together with the photo. The same goes for anywhere a professional opinion is needed: a diagnosis, an insurance damage assessment, legal advice. The agent understands what is in the file in order to route it correctly, not to replace the person qualified to decide.
Want to consult with us?
We can help you choose, build and deploy the perfect AI solution for your business. Leave your details and we'll get back to you.
How to build a media library the agent can use
Most of the work here is yours, and that is good: you know what convinces your customers. Seven steps:
- Start from the questions, not the files. Read last month's WhatsApp conversations and write down the ten questions where you would have wanted to send a photo, a video or a document.
- One file per moment. Not forty before-and-after photos, but the best three or four for each treatment or product. More than that drowns the customer.
- One sentence per file. What it shows and when to send it. The agent chooses by the context of the conversation, and a file called IMG_4471 tells it nothing.
- A short, vertical video. Under a minute, shot for a phone. WhatsApp limits file size, and nobody watches a five-minute video inside a chat anyway.
- Consent and rules. Photos of real customers only with the written consent of the person in them. Clinics should also check what the professional rules of their field allow in advertising. The agent sends only what you uploaded, and does not create or edit images itself.
- Sending rules. No file without a question that justifies it, and after every file a question that moves forward: "Shall I check a slot on Thursday for you?".
- A review date. A PDF price list from last year is wrong information in a customer's hands. Set a date to go through the library and replace whatever is out of date.
One more thing worth knowing about WhatsApp: within 24 hours of the customer's last message, a business can send them any message, including photos, videos and documents. Outside that window it may only send a template message that Meta approved in advance. So the right time to send the file is inside the conversation, when the customer asks, not "tomorrow, when we get back to them".
What WhaleBiz does today
On WhatsApp the agent works with full capability: it sends photos, videos, documents and catalogs from a library you define, and understands the voice notes, photos, videos and documents a customer sends. On the other channels these capabilities are being added step by step, and the project manager tells you in advance what is available on yours. If you are still deciding where to start, the comparison between a WhatsApp bot and a website chatbot helps.
Sending media starts on the Standard plan, together with WhatsApp and calendar booking. The Business plan adds a larger media library, as well as Instagram and Facebook. Every agent reply counts as one message, including a reply to a voice note, a photo or a file, with no double counting for media. You will find each plan's quotas and prices in the plan comparison, and the full detail of the capability on the media and documents in chat page.
What the agent does not do: it does not take, design or edit your photos, and it does not give a professional opinion on what it is sent. You choose the material, and a person makes the professional decisions.
Frequently Asked Questions
Does the agent send photos in the website chat too, or only on WhatsApp?
On WhatsApp the agent both sends and understands everything: photos, videos, documents and voice notes. On the other channels the capabilities are being added step by step, and the project manager tells you in advance what is available on the channel you chose.
Does a reply to a voice note or a photo count twice?
No. Every agent reply counts as one message, including a reply to a voice note, a photo or a file. Messages the customer sends are not counted at all. Each plan's quota is on the pricing page.
Which plan lets the agent send media?
Sending media starts on the Standard plan, together with WhatsApp and booking. The Business plan adds a larger media library, plus Instagram and Facebook. The full detail is on the pricing page.
Is it allowed to send customers before-and-after photos of patients?
Only with the written consent of the person in the photo, and clinics should also check what the professional rules of their field allow. The business decides which photos go into the library, and the agent sends only what was uploaded to it.
What happens when a customer sends a photo that needs a professional opinion?
The agent does not diagnose and does not assess. It tells the customer that a member of the team will check, hands the conversation to a person, and the photo is saved on the customer's record in the CRM so whoever answers sees it at once.
Michael Romm
Michael is the founder and CEO of WhaleBiz, leading business and marketing strategy. An expert in data (SQL, Python) and developing automation and AI solutions for businesses.