TLDRocket
Sign in

Scarlett brings recurring reports and multi-step automation to Slack

tryscarlett.ai

Scarlett is a new AI "coworker" living in Slack that can run recurring reports and multi-step tasks on its own. It's less chatbot, more automated intern with its own cloud computer and access to 3,000+ business tools.

Based on reporting by tryscarlett.ai — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Slack bots that summarize threads are old news. Scarlett wants to be something else entirely — an AI teammate that actually finishes the job. Install it from the Slack App Directory, hook it up to Stripe, Notion, Google Ads, or whatever else runs your business, and it starts pulling live data, building dashboards, and shipping the same weekly report every Monday without anyone nagging it.

The pitch leans hard on the word "colleague" rather than "tool." Scarlett gets its own cloud computer to write and execute code, and it picks different models depending on the job — Opus for conversation, Fable for planning, Claude 5.6 for coding, according to the company. That's a deliberate jab at simpler Slack integrations like Claude Tag, which mostly live inside a single chat thread. Scarlett instead keeps a persistent workspace and can chain together multi-step workflows: scanning support tickets every morning, watching a Drive folder for new contracts, or turning merged GitHub pull requests into changelog notes.

The premade workflow templates are where the ambition shows. There's a daily 7 a.m. team brief, a meeting-prep briefing that gathers attendee context before calls, and an inbox sweeper that finds unanswered Gmail threads and drafts follow-ups. One quoted customer says Scarlett pulled Q3 numbers from six different tools in under a minute, a task that used to eat half a workday. Another simply says it feels like hiring a senior IC that never sleeps — the kind of line built for a landing page, sure, but it captures the sales angle plainly enough.

Security questions get more attention than usual for this category, probably because giving an AI standing access to your CRM and email is a bigger ask than letting it summarize a channel. Credentials go through an authenticated connections page rather than chat, OAuth tokens sit behind a proxy, and custom API keys get AES-256-GCM encryption, injected only when an approved action runs. Scarlett is also built to pause and ask for a human sign-off before anything irreversible, like sending an email or deploying code. Workspaces are isolated, and the company says it doesn't train on customer data — the standard promises, but ones worth having in writing when a bot has write access to your Stripe account.

What's notable isn't any single feature — plenty of agent startups now promise integrations and automations — it's the framing. Scarlett isn't sold as a productivity trick. It's sold as headcount. Whether that holds up once teams actually depend on it for a Monday report is a different question than whether it looks impressive in a demo.

My take — AI-written commentary, not fact-checked reporting

This is the same pitch every "AI coworker" startup makes right now — replace the boring recurring task, call it a colleague, charge per seat. I'm not against automating a weekly CPC report; I'm against pretending an agent that still needs human approval for high-stakes actions is somehow not just a very fancy scheduled script with API keys. Useful, sure. A coworker, no.

Read more about this at: tryscarlett.ai

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.