When a customer asks for your price list in a text conversation, they want it right now. Every minute between that question and a real answer is a minute they're scrolling to a competitor. An AI agent wired with resource delivery sends the document in the same thread, in seconds, without anyone on your team doing anything.
This post covers exactly how that works: what triggers a delivery, what kinds of assets make sense, how to set it up, and where it fits inside a broader agentic system. It is a specific capability, not magic. Understanding the mechanics helps you use it well.
What does in-conversation resource delivery actually do?
In-conversation resource delivery is a configured behavior where an AI agent detects a trigger phrase in a conversation and responds with a direct link to a document, PDF, image set, or webpage. The customer asks about pricing; the agent sends a URL to the PDF service menu within the same exchange. That is the whole mechanism.
It sounds simple because the concept is. The value is in the timing. Most service businesses handle this the old way: someone asks, the owner or front desk says "I'll email that over," the customer gets the email four hours later, forgets the context, and never opens it. The friction between the question and the answer is where leads go quiet.
In-conversation delivery collapses that gap. The customer stays in one thread. They get what they asked for before their attention moves elsewhere. And because the resource is a URL, it works on any text-based channel: web chat, SMS, and direct messages on social platforms. The customer taps the link and the document opens on their phone.
What types of documents should you set up for delivery?
The best assets for in-conversation delivery are the ones your team sends manually the most. Any file accessible via a URL works: PDF service or treatment menus, care instruction guides, before-and-after photo galleries, quote templates, and FAQ sheets.
Across the systems we've built, the most commonly triggered documents fall into three categories. First, pricing and service menus: a customer asks what something costs or what services are offered, and the agent sends the menu PDF. Second, care or prep instructions: after a booking is confirmed, the agent sends the pre-appointment guide or post-treatment care sheet without waiting for staff to remember to do it. Third, proof assets: before-and-after galleries or portfolio pages that a customer asks to see before committing.
What does not work well is anything that requires customization before sending, like a personalized quote with specific line items calculated on the fly. That is a different capability entirely, closer to what an AI quote and estimate agent handles. Resource delivery is for fixed documents that are the same for every recipient asking that question.
How do trigger phrases work, and how do you set them up?
A trigger phrase is a word or short phrase that, when detected in the conversation, causes the agent to fire a specific response. The response includes the resource link alongside a natural-sounding message. "Here is our treatment menu" followed by the URL, for example. The agent does not just drop a raw link; it wraps the delivery in context so the exchange reads like a real conversation.
Setup starts with an inbox audit. When we wire up resource delivery for a new client, the first thing we do is pull the last 90 days of conversations and tag every request that came in for a document or piece of information. That list almost always has five or fewer unique items accounting for the vast majority of requests. Those become the triggers.
Each trigger is assigned to one document. "Price list," "how much does it cost," "do you have a menu" could all map to the same PDF. The agent handles variations in phrasing because it reads intent, not exact matches. You upload the document once (or provide the URL if it already lives on your website), assign the triggers, and the configuration is done.
Average time businesses take to respond to an inbound lead inquiry, according to a Harvard Business Review study from 2011. The document gap is part of that delay.
How does this affect the sales cycle for a service business?
It removes the gap between interest and information. That gap is where most service business leads die. Not because the customer lost interest in the service, but because the friction of waiting made them think the business was hard to work with.
A pest control company we onboarded was having their sales rep manually email a six-page treatment menu PDF to every single inquiry. He was doing this from his phone between jobs. Fifteen to twenty times a day. The email would go out hours after the original question. Some customers had already called someone else by the time it arrived. Once that document was set as a trigger in the agent, it went out in seconds. The sales rep stopped doing it entirely. The customers who received it quickly tended to ask a follow-up question in the same thread, which kept the conversation moving toward a booking rather than going cold.
That pattern shows up across different verticals. The sooner the customer has the information they asked for, the more likely the next message is a booking question. Delivery in the same thread keeps the momentum alive in a way that a deferred email does not.
Does this work over SMS, and what does the customer actually receive?
Yes. Because the delivery mechanism is a URL, it is channel-agnostic. The agent sends the link through whatever channel the conversation is happening on: your website's chat widget, an SMS thread, a Facebook or Instagram DM. The customer taps the link and the file opens in their browser. For PDFs, most phones open them natively. For hosted web pages (a photo gallery or service page), it opens in Safari or Chrome.
The format of the message matters. The agent does not send a naked link. It sends a short, natural line of text with the link embedded. "Here is our care guide for the procedure we discussed:" followed by the URL. That framing is part of how the AI customer support agent maintains a conversational tone across the whole exchange rather than feeling like a bot firing off automated responses.
One thing to get right during setup: the documents themselves should be mobile-friendly. A 20-page PDF with small print is not useful on a phone screen. The best-performing assets in this context are short (one to four pages), visually clean, and designed for someone reading quickly on a small screen.
How does the agent avoid sending the wrong document or repeating itself?
A well-configured agent uses context, not just keywords. It tracks what has already been sent in the current conversation and avoids resending a resource that was already delivered earlier in the same thread. If a customer received the service menu in message three and then asks a different question in message eight that happens to touch on pricing, the agent responds to the pricing question directly rather than resending the PDF they already have.
The agent also reads intent. "My sister mentioned your price list" is not a request for the document. "Can you send me your price list?" is. The difference is handled by how the trigger is configured, whether as an exact phrase or as a semantic match. For most service businesses, the simpler approach works fine: a small set of well-chosen triggers mapped to specific documents, with a human-readable fallback when nothing matches.
Knowing when not to deliver a resource is just as important as the delivery itself. That logic is part of a broader set of guardrails built into an agent that runs across web, text, and phone. The goal is a conversation that feels competent and attentive, not one that fires automated responses on any keyword match.
What do you actually need to get this running?
Three things: the documents themselves, a place to host them, and an agent configured to match triggers to deliveries.
Your documents need to be accessible via a public URL. If they already live on your website (a PDF linked from your services page, for example), you can use that URL directly. If they are sitting in a Google Drive folder or on your desktop, they need to be moved somewhere stable: a hosted page, a PDF storage service, or a dedicated folder on your site. The URL must work on mobile and must not require a login to view.
The trigger configuration is part of building out the agent's knowledge base. We pull the frequent-request list from the inbox audit, write the triggers, and test each one in a sandbox conversation before going live. The resource delivery trigger is one of the simplest wins we build into any agent setup. Within the first week it is usually the most-used feature, and the one clients notice immediately because customers stop asking the same question twice.
The rest of the agent handles everything else: qualifying the lead, answering general questions, booking appointments, routing complex situations to a human. Resource delivery is one capability inside a larger system. If you are thinking about AI text follow-up for leads, this is a natural complement: the initial agent delivers the resource on request, and the follow-up sequence re-engages anyone who received the document but did not take the next step.