How to Announce New Features to Users Inside the Conversation
Firstflow Team
Most teams announce new features the wrong way
Most teams announce new features the same way: an email blast, a changelog entry, maybe a banner at the top of the UI. For traditional SaaS products, this works well enough. Users are logged in, they see the banner, they investigate.
For agent products, it doesn't work at all.
Your users aren't browsing a UI. They're in a conversation. They don't see banners. They don't read changelogs. And an email about a new feature lands in an inbox they may check days later — if ever. By the time they see it, the context is gone and the motivation to try something new has faded.
The right place to announce a new feature to an agent product's user is inside the conversation itself, at the moment they're already engaged.
Why in-conversation announcements work differently
An announcement delivered mid-conversation has something no email or banner has: the user's full attention. They're already talking to the agent. They're already in the mindset of using the product. When the agent says "we just added something you might find useful" at the right moment, the user is ready to hear it — and ready to try it immediately.
This is the advantage that's unique to conversational products. The distribution channel for new features is the same channel as the product itself. You don't need to reach users through a separate medium and hope they come back. You reach them where they already are.
The result: in-conversation feature announcements consistently drive higher trial rates than email or in-app banners — not because the message is better, but because the timing and context are.
What a good in-conversation announcement looks like
The instinct is to announce everything at once — a changelog-style message listing every new capability. Resist it. In a conversation, this reads as an interruption, not an announcement. The user didn't ask for a list. They're in the middle of something.
A good in-conversation announcement has three parts:
1. A hook tied to what the user just did.
The announcement should connect to the current context. If the user just asked the agent to summarize a document, and you've just released a new feature that extracts structured data from documents, that's the moment to mention it. "By the way — we just added the ability to pull structured tables out of documents like this one. Want to try it?"
Context-tied announcements feel like helpful suggestions, not marketing pushes. The user understands immediately why this is relevant to them, right now.
2. One feature, one sentence.
Don't announce multiple features in a single message. Pick the one most relevant to this user, at this moment, and describe it in a single sentence. Users who are mid-conversation don't want to context-switch into evaluating a feature list. They want one thing they can act on immediately.
3. An immediate next step.
The announcement should close with an action: "Want to try it?" or "Should I show you how it works?" This converts awareness into trial in the same message. The user doesn't have to go find the feature, navigate to a settings page, or remember to try it later. They just say yes, and the agent demonstrates it.
Targeting: not everyone needs to hear every announcement
The power of in-conversation announcements is that they can be targeted in ways that email blasts can't. You know who each user is, what they've used, and what they haven't. Use that context to make every announcement feel relevant.
Segment by capability usage. If a user has never used capability A, and you've improved capability A significantly, they're a high-priority target for that announcement. A user who already uses A daily probably doesn't need the announcement — they'll discover the improvement on their own.
Segment by onboarding stage. New users should hear about foundational capabilities before advanced ones. An announcement about a power feature delivered to a user who hasn't completed basic onboarding will land flat. Sequence announcements based on where the user is in their journey.
Segment by role or profile. If you've collected any structured data about your users through onboarding flows or surveys, use it. A feature designed for sales teams should be announced to users who've identified as being in sales — not to everyone.
Time it to high-engagement moments. An announcement delivered during a session where the user is actively engaged will perform better than one delivered at the start of a session where the user hasn't yet found their rhythm. Track engagement signals within the conversation — response rate, follow-up questions, time between messages — and trigger announcements when they're high.
Measuring whether it worked
Unlike an email announcement with an open rate, an in-conversation announcement gives you richer measurement:
- Impression rate: What percentage of targeted users received the announcement?
- Response rate: What percentage responded positively ("yes, show me")?
- Trial rate: What percentage actually used the new feature within the session or within 24 hours?
- Feedback after trial: What did users rate the feature after trying it for the first time?
This funnel tells you not just whether users saw the announcement, but whether they tried the feature and whether they found it valuable. No email announcement gives you that last data point automatically.
If response rate is high but trial rate is low, the announcement generated interest but the feature is hard to get started with. If trial rate is high but post-trial feedback is low, the feature doesn't live up to what the announcement promised. Both are actionable in a way that an email open rate never is.
The release cycle shift
Once you have in-conversation announcements as a distribution channel, the way you think about releasing features can change.
Instead of batching features into a big release so there's enough to warrant an email, you can ship smaller improvements more frequently and announce each one to the users most likely to benefit — immediately, in context, at the right moment. The feedback loop tightens. The feature gets real-world signal faster. And users feel like the product is alive and improving, because it keeps introducing them to new things through the conversation they're already having.
Get started with Firstflow today and start building in-chat experiences that help AI agents activate users within minutes.