Conduit Intro Assist: automatically write personal email intros
Launching: Generate contextual email intros with work and relationship history, as a ready-to-send draft
Today, Conduit is launching Intro Assist, a first-of-its-kind AI that instantly writes email intros between your contacts for you — with details about your relationship, ready to edit and send.
Intro Assist combines the powerful context of your Personal Graph with natural language technology and best industry practices. Generating an intro this good has been impossible until now.
Intro Assist-generated emails are really good — people say over 92% of them can be sent without any changes. And if you write more than 3 introductions a day, Intro Assist saves over a full days’ worth of work every single month.
Generated intros open as a draft for you to review, personalize, and send — in Gmail or any local email client including Airmail and Outlook. We recommend adding personal touches and asking for double opt-in first (i.e. both people know about and want the introduction). There is never any branding on a generated intro.
Intro Assist includes context, relationship, and work details
When appropriate, Intro Assist looks for ways to incorporate—
- Contact details, as recipients and in the subject line
- Work details, like what each person does and where they work
- Your long-term history and recent interactions with each contact (such as how long you’ve known them, or that you met over a recent dinner)
- A call to action — for follow up, a review, a call, or more
Intro Assist alerts you when we guess or are unsure
Technology like this won’t be perfect for some time. If we guess, are unsure, or work with imperfect data, Intro Assist will flag a prominent warning for you to review and confirm. This way, you can use the feature with confidence.
Even if you decide not to use the written intro, Conduit can still save you time by pre-filling the email recipients and subject line.
Intro Assist is ready to use now
Starting now, Conduit and Conduit VIP users can open Intro Assist from any contact’s page. We’re focused on the double opt-in use case now, but are looking to handle more situations soon. If something can be improved, just let us know — this is a beta :)
So how does it all work?
Intro Assist works well because we have deep knowledge of your relationship with both parties. What sets Conduit apart is our Personal Graph, the data- and AI-driven model we’ve built to understand relationships and interactions.
When generating intros, we’re helping you put your own data to use. Intro Assist also tackles some interesting engineering challenges, a few of which we’ll share below.
Natural language contextualization
People have all sorts of job titles, and some of them are more intuitive than others. Intro Assist features a custom-built engine that contextualizes work positions and adapts them to fit a conversational, professional tone.
Handling real-world job titles is tricky for a number of reasons —
- There is only one CEO, but there can be multiple partners and many analysts. Sentence patterns differ for each: “the CEO”, “one of the partners”, “an analyst”.
- Positions can include a division, which can be implicit (“Director of Growth” becomes “Jack leads growth at”) or explicit (“Director, NYC Region” becomes “Jill directs the NYC region for”).
- Job titles aren’t clean. They can include adjectives and add-on terms. People like to include all sorts of irrelevant information in their LinkedIn. Where we can’t figure it out, we flag the potential mistake and try our best.
Relationship cadence and action point detection
One of the ways we leverage your Personal Graph in the introduction is to look at calendar, email, and LinkedIn history to detect “relationship cadence” and your most recent “action point”.
We define relationship cadence as the start behavior and aggregate behavior over time between you and the other person. This influences commentary in the introduction, which helps establish the level of rapport you have. We look at when you connected on LinkedIn, your first in-person coffee chat or meal, and more.
The action point is the meeting, call, or email that actually triggers the introduction. It’s usually either a recommendation from you or a request from the party that wants the intro.
While the action point is often the most recent interaction, this isn’t always the case. For example, if you grab coffee with John and send a couple of emails afterwards, we skip the emails and suggest that the intro was a result of the coffee chat.
Including cadence and the action point helps establish authenticity and purpose. The Personal Graph helps us rapidly detect, digest, and structure this information.
Predicted relationship dynamic (early testing)
One thing we’re exploring is to detect and then surface a predicted relationship dynamic, or what we think is the introduction’s purpose. This is most useful for generating intro opt-ins, which we’re testing internally.
To predict relationship dynamic, we pattern-match the role and company of both parties. As it turns out, the vast majority of introductions are made within a narrow band of reasons —for example, a founder to a potential investor; a potential hire to a recruiter; or a deal to a potential partner.
During a matched scenario like this, we can suggest or add an option to include this context.
Conduit is the #1 tool for connectors and leaders to better understand, grow, and engage their network. We’ve taken a new, data- and AI-driven approach to managing relationships, all powered by our Personal Graph platform.
