This is an example of an agent your team can build in Sylvie: it writes the SMS and MMS copy for promos, drops, reminders and lifecycle flows, inside the character budget, with the offer terms and opt-out language your brand and market require. It runs on your brand brain, so it knows what the brand sounds like when it only gets one line, which offers are live, what the fine print says and which past sends actually converted. Sylvie does not ship it pre-built: your team assembles this agent once on the brand brain, and it drafts every send after that.
Sylvie does not ship this agent pre-built. It is an example of an agent your team can build with Sylvie, powered by your brand brain, so it runs your workflow the way you actually run it.
The SMS Message agent is an example of what a marketing team can build on top of its brand brain in Sylvie. It does one job end to end: the text message itself, from the first four words that show on a lock screen through to the link, the offer terms and the opt-out, written in a voice a subscriber will recognize.
SMS is the least forgiving channel a marketer owns. There is no subject line to rescue a weak opening and no design to carry a flat sentence, and almost everyone opens it, including the people who will reply STOP the moment it feels like spam. It is also the channel where a wrong offer or a missing disclosure becomes a real problem fastest, which is why the copy should come from the same approved facts as everything else the brand publishes.
When you assemble the agent, you connect it to the brand brain: the compressed version of your tone (what the brand sounds like in one line, whether it uses first names, which emoji are allowed), the live offers with their real terms and end dates, product names and drop dates, the disclosure and opt-out language approved for each market, quiet-hour and frequency rules, and the sends that performed next to the ones that drove unsubscribes.
Inside Sylvie you say what the send is (a drop, a restock, a cart recovery, a booking reminder) and which segment receives it, and the agent returns the message with its character count and segment split, the link placement, the required footer and two or three variants worth testing. It checks every claim and offer against the brain, flags anything unapproved or expired, and hands the approved copy to your SMS platform. Results come back into the brain, so what converted shapes the next send.
Brand voice usually dies in compression. The agent works from a tone profile built for one line, so the text still reads like your brand rather than a template with a discount code stapled on.
It writes from the live offer list in the brain, with the real terms, dates and exclusions, so nothing goes out promising a discount that expired on Tuesday.
Opt-out wording, sender identification and the rules your team set per market are part of the draft itself, not a checklist someone remembers at 6pm before a send.
Each draft comes with alternates that change the hook rather than the punctuation, so the test tells you something real and the winner is logged back in the brain.
A peak calendar can run to forty messages across segments. The agent keeps them coherent, so the sequence reads like one conversation instead of forty separate shouts.
The sends that drove revenue and the ones that drove unsubscribes both feed back in, so the agent's instincts about length, timing and hook sharpen with every campaign.
SMS work arrives in bursts: a launch, a peak calendar, a client's first program. Here is where teams point this agent once it runs on the brand brain.
A retail client needs around thirty sends across five segments in ten days, each with a different offer and cut-off. The agent drafts the whole calendar in a morning from the approved promo list, so the team reviews instead of writes and peak week ships without a single wrong end date.
The message has to create urgency without promising stock the warehouse cannot back. The agent writes from the real drop date and inventory rules in the brain, so the alert converts on genuine scarcity and the brand skips the apology email that follows an overpromise.
The abandoned cart email went out an hour ago, and a text that repeats it burns the channel. The agent sees the flow context in the brain and writes a message that adds something new (a shipping cut-off, a size back in stock), so the pair recovers more carts while the unsubscribe rate stays flat.
A clinic or studio sends thousands of reminders that read like they came out of a database. The agent rewrites the whole reminder set in the brand's voice with clean confirm and reschedule instructions, so no-shows drop and the message clients receive most often finally sounds like the brand.
Subscribers who have not bought in ninety days need a reason that is not another 10 percent off. The agent builds variants around approved offers and product news from the brain, so the team tests three genuinely different angles in one week instead of guessing with a single send.
The brand is starting at zero: consent capture, welcome message, cadence rules and a first quarter of templates. The agent drafts the whole program off the brand brain and the market rules you set, so a channel that usually takes a quarter to stand up is live and on brand within weeks.
Those assistants live in the sending platform and know your send history, not your brand. This is an agent you build on your own brand brain, so it starts from your approved offers, your product facts, your tone and the fine print your team has cleared. You keep the platform you send from. What changes is that the copy arrives already grounded.
It enforces the rules you put in the brain: the opt-out and identification wording your team approved, quiet hours, frequency caps and per-market requirements, and it flags any draft that breaks them. That removes the most common failure, a send that goes out missing something everyone knew should be there. Final legal responsibility and sign-off stay with your team, exactly as they do today.
That is what the brand brain is for. A tone of voice document written for long-form is close to useless at 160 characters, so the agent works from a compressed profile (sentence rhythm, whether you use the subscriber's name, emoji policy, how you write a call to action) built from the texts your brand has actually sent and the ones that performed.
That is your call when you build it. Most teams have it draft and route copy for approval, then push the approved message into their SMS platform, because a text cannot be unsent. The agent can be assembled to run either way, so the approval step is a decision you make, not a limitation you inherit.
Yes. You build the agent once and run it on each client's brand brain, so every brand keeps its own tone, offers, disclosures and send history, with strict isolation between accounts. Nothing from one client's promo calendar can turn up in another client's text.
The setup work is the brand brain, and if a client already runs on Sylvie most of it is in place. Adding the SMS-specific pieces (the compressed tone profile, the approved footer, the live offer list) takes a session, and from there you get usable drafts on the first campaign and refine the agent from what converts.
Book a demo and we will map the workflows worth turning into agents, and show how each one runs on your brand brain.
Request a demo