Read AI, best known for AI meeting notes and summaries, has launched an email-first assistant called Ada that the company describes as a “digital twin.” The pitch is simple: instead of asking users to adopt yet another chat interface, Ada lives inside the workflow most teams already use, email, where it can coordinate meetings, draft replies, and answer questions using company context.
According to TechCrunch, users can start by emailing [email protected] with the message “Get me started,” after which Ada can be added to email threads to perform tasks while the user stays in the loop.
The timing reflects a broader shift in “meeting notetaker” tools: they’re moving beyond passive summaries into action-taking copilots that can schedule follow-ups, surface knowledge, and keep work moving after the meeting ends. Read AI is leaning into that shift by making Ada feel less like a chatbot and more like an always-available executive assistant, one that can operate when you’re out of office, in back-to-back meetings, or simply offline.
Read AI’s CEO David Shim framed it as building on existing behavior instead of adding friction: if you use email, you can use a “digital twin.”
How Ada works in real email threads
Ada is designed to be activated directly inside a conversation. When someone asks to schedule time, you add Ada on CC, and it replies in-thread with availability based on your calendar access through Read AI. If the other participant can’t make those times, Ada continues the back-and-forth and proposes new slots until the meeting is booked. Importantly, Read AI says Ada does not disclose the subject or details of your existing meetings—only availability.
This email-native approach is a deliberate choice. Instead of forcing users into a separate UI for “agent actions,” Ada tries to turn the inbox itself into the control plane for coordination—where scheduling, follow-ups, and confirmations already happen.
Answers that use your company context, not just the open web
Scheduling is only half the product. The other headline feature is Q&A grounded in workplace context.
TechCrunch reports that Ada can answer questions using a company knowledge base, topics discussed in prior meetings, and even public internet searches. A practical example given was asking for a status update like, “How are we tracking for Q1 goals?” and having Ada assemble a response from what it has access to.
When someone else asks a question in a thread, Ada can draft a response for you, then help you refine it before it’s sent—an important detail because it positions Ada as “assistive + reviewable,” not “fully autonomous.”
The privacy model: “sidebars” and approval before sensitive replies
Read AI is putting a lot of product messaging behind trust controls—because an assistant that can see calendar signals, meeting history, and knowledge sources can easily become risky if it blurts out the wrong thing.
Their approach is a concept they call a “sidebar.” For non-scheduling questions (especially those that could include sensitive details), Ada pulls the conversation into a private confirmation loop: it shows you the draft first, and asks you to approve or edit before anything goes out. Read AI explicitly calls this friction “by design,” arguing that speed without review breaks trust the first time something goes wrong.
On security and spoofing, Read AI also says Ada verifies email identity using SPF, DKIM, and DMARC, and only accesses data when you’re an authenticated participant on the email thread.
Under the hood: knowledge graph over meeting data, not MCP-dependent
A notable technical point from TechCrunch: Read AI’s VP of Product said Ada doesn’t rely on MCPs (Model Context Protocols) to connect tools. Instead, the company builds a knowledge graph from meeting data and connected services to generate more contextual answers, and it plans to make the assistant more proactive over time.
That “proactive” direction matters. Read AI’s own product pages describe actions like post-meeting scheduling and conflict resolution, plus suggestions and updates that can connect into business systems (for example CRM updates).
Rollout, scale, and pricing signals
Read AI says Ada is launching broadly, and it’s not being positioned as a niche add-on. The company states the digital twin will roll out to 5M+ monthly active users on day one and will be available globally, with Ada described as a free service for existing and new users.
TechCrunch adds scale context: Shim said Read AI has over 5 million monthly active users, sees around 50,000 sign-ups per day, and has a wider base of 100,000 users who consume Read AI summaries without creating an account. The report also notes Read AI has raised over $81 million in funding and that while 60% of users are outside the U.S., revenue is split roughly equally between the U.S. and international markets.
On the roadmap, TechCrunch reports Ada is email-first today, with plans to expand to Slack and Microsoft Teams.
How it stacks up against other AI meeting tools moving into “actions”
Ada arrives as competitors in the meeting intelligence category push toward repeatable prompts, automated follow-ups, and integrations with tools like Notion, CRMs, and project trackers. TechCrunch pointed to examples like Granola adding repeatable prompt “recipes,” and newer entrants like Quill connecting meeting output to other systems to automate tasks.
Read AI’s differentiation is the distribution bet: shipping a “do work for me” assistant through the inbox, with review gates, and a promise that it learns from what you already do across meetings and connected services.
What this launch could change for teams
If Ada performs reliably, it nudges workplace AI toward something more operational than “summarize what happened.” Email-based scheduling is a real pain point, and answering internal status questions is a daily tax on managers and operators. The bigger swing is whether the “digital twin” framing becomes a durable category, or whether users simply evaluate Ada as another assistant that must prove it won’t embarrass them.
Read AI’s design choices suggest it understands that tradeoff: make the assistant useful inside high-frequency workflows, but add explicit approvals for anything that could leak context. If that balance holds, Ada could become one of the more practical, low-switching-cost AI rollouts in the productivity space this year.