Scroll

Most AI knowledge tools ask you to trust the answer. Scroll asks you to check it, and that single design choice tells you who this is for. We'd reach for it when the answer will be challenged — a rep quoting product specs to a customer, a compliance lead answering a regulator's question, a coach whose advice needs a paper trail back to the interview it came from. The ingestion model suits teams that already have the corpus. If your product docs, policy PDFs, and recorded calls live in a shared drive and go stale weekly, the hourly/daily refresh schedules are the feature that keeps the agent honest without someone babysitting it. Where it bites is volume. At 50 credits per query, an agent used by fifty people a day burns through allowance quickly, so model the query load before you commit to a tier. Audio ingestion at 3 credits per hour and web pages at 5 credits each are cheap by comparison — the query meter is the one to watch. Pass if your agents need live, real-time data feeds rather than scheduled refreshes, or if you need offline and mobile access to the knowledge base. Pass too if you want to pick your own underlying model — Scroll is a fixed, citation-first pipeline, and the citation behavior is the point. The Free tier's 300 pages is enough to test ingestion quality on one narrow corpus, not enough to run a real department. Teams at $200/month ($160 annually) is where branding controls and shared agents come in; enterprise security controls sit above that. Compared with a generic chatbot builder, the tradeoff is flexibility for verifiability. You give up model choice and get an answer you can defend in a meeting. For teams where an unsourced answer is a compliance incident, that's the right trade.