Transitioning away from brittle, rule-based automation nodes toward a centralized, proactive memory layer built on official platform APIs (Meta, X, TikTok) is a massive unlock for social CRM scaling. Setting up complex keyword-matching trees or writing fragile manual webhooks for every distinct permutation of follower intent is a major maintenance bottleneck. FanBase shifts this paradigm by treating user interactions as a continuous context-retrieval problem. The core engineering win is the auto-calibrating memory system that continuously ingests historic comments, mentions, and team activity to map a deterministic brand voice profile. Automating low-latency comment-to-DM loops based on autonomous buying intent categorization—without forcing users away from their native app environments—drastically compresses top-of-funnel conversion friction.
Curious how the AI actually learns my voice over time. Does it need a batch of past posts to get started, or can it pick up my style as it goes from the first few replies?
@selma1777037 Hi Selma!
When going through our onboarding, the AI Copilot will go through your recent posts to get a first understanding of your voice, your audience, and build an initial knowledge which gets refined over time.
So while the AI Copilot gets smarter over time, we spent a lot of efforts making the onboarding an actual useful step to let you get value out of the product instantly.
the buying-intent-triggers-a-DM part is the one I'd want guardrails on. unsolicited brand DMs already have a spammy reputation, and if it's triggered off a public comment, there's a real risk it reads as "why is this brand watching me that closely" instead of helpful. is there a delay, rate limit, or human review step before the DM actually fires, or does it go out the moment the signal is detected?
@galdayan Hi Gal! Yup, at the moment DMs aren't triggered automatically for buying intent.
In the case of a "keyword-to-DM" automation, those are sent automatically.
But if the AI Copilot detects a comment worth answering to, it goes to the Review section where the user can edit/approve the DM before sending it out.
@goathier that's a sensible split - fully automated for the keyword stuff where the trigger is unambiguous, human-in-the-loop for the fuzzier AI-judgment calls where a wrong read costs you the relationship. makes sense you'd want a person to catch it before a DM goes out over something the AI misread as buying intent.
how does the AI actually pick up on my voice and tone without me having to feed it a bunch of old posts first
@ksulunhat56036 It learns your voice based on older posts, but you don't have to "feed it" or do anything manually. When you connect your socials (part of the onboarding or later on via the settings), the AI Copilot will do an initial scan of recent posts to improve the knowledge/memory it has about your, your style of writing, the emojis you use etc... and get smarter over time.