简体中文
English
Website review
This review may include AI-generated content and is not individually verified. Domain, certificate and other facts come from the observed data on this page.
What is Lobby?
Lobby is a tool built around a predictive email workflow agent. Its stated goal is to create "extensions of human cognition": a behavioral model that learns how you think and act, then acts on your behalf without requiring prompts. In practice, that means you connect your email and calendar tools, the model predicts your email workflows end to end, and you approve its output rather than writing instructions each time.
What it actually does
The product's page describes one concrete application: coordinating email workflows across tools you already use, with no prompting. The pitch is a shift from reactive agents (you ask, they respond) to proactive ones (they anticipate and you approve). The page claims 92% prediction accuracy, a prediction depth of five tasks and seven days, and 80+ sub-agents, and it says the behavioral model is user-owned and portable across systems.
Who it suits
- People drowning in coordination work—founders, sales and fundraising staff, product leads—who repeatedly triage, draft and schedule similar emails and would rather review predictions than compose from scratch.
- Teams already using many connected tools, since the value depends on the model seeing enough context across email, calendar and related apps.
- People comfortable delegating judgment, because the core trade-off is trust: approving predicted actions is faster only if the predictions are reliably right for your style and context.
It is a weaker fit if you handle sensitive correspondence you would not want modeled, or if your email habits are too irregular for a behavioral model to learn useful patterns.
Practical next step
Pick one recurring email workflow—say, scheduling follow-ups after meetings—and judge the product on how often its predictions match what you would have written. That approval rate, not the headline accuracy figure, is the number that determines whether Lobby saves you time.
How does Lobby's predictive email agent work without prompts?
Lobby's approach is to learn a model of your judgment rather than wait for instructions. According to the site, you connect your existing tools, and the system builds a "behavioral model" that predicts your email workflows end to end; you then approve its proposed actions instead of writing prompts. The page frames this as a shift from reactive agents (which need prompting and upkeep) to a proactive one that acts "with full fidelity to who you are." It also states that the model is user-owned and portable across systems, and that the email application is its first product.
What "no prompt" means in practice
- Learning from your history: the model infers how you handle email—who gets replies, how you route requests, what you delegate—from connected tools, not from typed instructions.
- Predicting whole workflows: the site claims end-to-end prediction of email workflows, with approval as the human step.
- Stated metrics: 92% prediction accuracy, a prediction depth of 5 tasks over 7 days, and 80+ sub-agents. Treat these as vendor claims to test against your own inbox.
A concrete way to judge it
Pick a recurring, low-risk pattern—say, scheduling intro calls or triaging vendor emails. Connect only the tools that pattern touches, run it for a week, and count how often you accept the prediction unchanged versus edit or reject it. High acceptance on routine threads with clear edits on sensitive ones is the signal that matters; a single overall accuracy figure hides exactly that split.
For context on how promptless email tools compare, see Superhuman and Shortwave.
How accurate is Lobby's behavioral model, and what does prediction depth mean in practice?
Lobby reports 92% prediction accuracy for its behavioral model, with a prediction depth of five tasks and a seven-day horizon to reach 80% convergence. Those figures come from Lobby's own page, so treat them as vendor-reported rather than independently verified. The practical question is what they mean for your inbox.
What the numbers describe
Accuracy (92%) is the share of predictions the model reportedly gets right when it anticipates your email workflows. In use, that means most of the time the agent's proposed action—routing, drafting, scheduling, archiving—matches what you would have done, and you approve rather than correct.
Prediction depth (5 tasks) describes how far ahead the model plans within a workflow: it can chain roughly five steps (for example, read a thread, identify the ask, draft a reply, propose a meeting time, and file the thread) before it needs your input. Deeper depth means fewer interruptions per workflow.
Seven days to 80% convergence is a learning curve, not a per-email metric. It suggests the model needs about a week of observing your behavior before it reliably reflects your judgment in most cases. Early on, expect more corrections; after that, fewer.
What this means in practice
A concrete scenario: you run sales or fundraising and handle dozens of threads a day. During the first week, review the agent's drafts closely—it is still calibrating. By the second week, if your patterns are consistent (how you qualify leads, how you phrase follow-ups, when you defer), approvals should become quick taps. The 92% figure matters most for high-volume, repetitive workflows; unusual one-off emails will still need your judgment regardless of the score.
The trade-off is transparency versus control. A model that predicts end-to-end and only asks for approval saves time, but you need to trust its calibration and spot the 8% it gets wrong—especially on sensitive threads. If your work is highly varied or relationship-critical, a shorter learning period with more manual oversight may suit you better than leaning on the depth figure.
Decision criterion
Ask whether your email patterns are repetitive enough for a behavioral model to learn within a week, and whether you are comfortable approving rather than composing. If both hold, the reported accuracy and depth are a reasonable basis for a trial; verify the numbers against your own correction rate in the first two weeks.
Which email and calendar tools can I connect to Lobby, and how does the $42/month plan differ from the 50+ tools addon?
The page does not list which specific email or calendar tools Lobby connects to. It only says the product coordinates email workflows "across the tools you already use," and the plan description mentions "Unlimited emails and calendars" plus a "50+ tools addon." So the exact integrations and what the addon unlocks are not documented in the available page content.
What the page does establish
- Core behavior: Lobby connects to your tools and predicts email workflows end to end, with you approving rather than prompting.
- Base plan signals: From $42/month, described as including unlimited emails and calendars.
- Addon: A separate "50+ tools addon" is mentioned, implying broader tool coverage beyond the core email and calendar connection, but the page does not name those tools or state what changes in practice.
How to decide between the two
The meaningful question is whether your work lives mainly in email and calendar, or spans other systems where context matters.
| Situation | Likely fit |
|---|---|
| Your coordination happens mostly through email and scheduling | Base plan is the relevant starting point |
| Your email decisions depend on CRM, docs, project trackers or similar systems | The 50+ tools addon is the one to evaluate |
| You need a named integration confirmed before committing | Ask Lobby directly; the page does not answer this |
For a concrete case: a founder juggling investor threads and calendar holds may find the base email-and-calendar scope sufficient. A sales lead whose follow-ups depend on pipeline records would need the addon—but should confirm the specific CRM is supported before relying on it.
Next step: Request the current integration list and a precise definition of what the addon adds (tool count, which categories, and whether it changes prediction scope) before choosing a plan.
Who benefits most from Lobby's email coordination—sales, fundraising, or product teams?
Sales teams are the clearest fit, with fundraising a close second; product teams benefit least from the email-coordination feature specifically.
Why sales and fundraising lead
Lobby's page describes an agent that predicts your email workflows end to end and lets you approve rather than prompt, aimed at people juggling many relationships and contexts at once. That maps directly onto roles where email is the work:
- Sales: high volume of threads, follow-ups and scheduling across many contacts. The promised "predicts your workflows... no prompt, you just approve" model suits repetitive but judgment-heavy outreach. The page also lists "Sales" among its signal categories.
- Fundraising: fewer but higher-stakes threads with donors and stakeholders, where tone and timing matter and context is scattered across tools. The "50+ tools addon" and cross-tool coordination are the relevant hooks here.
- Product teams: their core work happens in docs, tickets and design tools, not email. They'd gain less from an email-first agent.
A practical way to decide
Ask whether your week is dominated by email coordination with external people. If yes (sales, fundraising, partnerships), the approval-based model is a plausible time-saver. If your bottleneck is internal collaboration or building, look elsewhere first.
For context on the crowded AI-email space, compare official pages such as Superhuman and Shortwave, and check Lobby's own materials at Lobby before committing.
How does Lobby keep my behavioral model private and portable across systems?
Lobby's stated approach is to make the behavioral model the user's own artifact: the page describes it as "user-owned and portable across systems," so the model is meant to travel with you between tools rather than stay locked inside one app. Privacy is framed the same way—the model is described as an extension of your cognition that you own, rather than a profile the vendor controls.
What the page actually commits to
- User ownership: the behavioral model is explicitly called user-owned.
- Portability: it is described as portable across systems, so it isn't tied to a single email client or tool.
- Scope of learning: the first application coordinates email workflows across "the tools you already use," which implies the model reads signals from connected calendars and tools to predict your workflows.
- Human approval: Lobby says it predicts workflows end to end with "no prompt, you just approve," so you remain the checkpoint before actions go out.
The page does not detail encryption, data residency, export formats, deletion rights or how portability is technically implemented. Treat "private and portable" as the product's stated intent, not a verified guarantee—check the Privacy and Terms pages and ask directly about export and deletion before connecting work accounts.
Practical test for a privacy-conscious user
If you're considering it for a role handling sensitive mail (fundraising, sales pipelines, legal), run a small trial: connect one calendar and one mailbox, confirm exactly which signals are ingested, then ask what happens to the model if you cancel. Portability only matters if you can actually take the model out—so request a concrete export example rather than accepting the phrase.
For comparison on how other agent tools frame data handling, OpenAI and Anthropic publish their own enterprise data policies; read those alongside Lobby's terms to judge which posture fits your organization.
Cloudflare
Vercel
Google Workspace
User reviews (0)