The problem this solves
At the very top of an account list, the economics of outreach invert. For the ten accounts most likely to close, the constraint is no longer volume; it is distinctiveness. These champions receive dozens of competent, personalized-ish emails weekly, and text alone - however well researched - arrives in a format their inbox has taught them to skim. Standing out for the accounts that matter most calls for an artifact, and artifacts are expensive to make by hand.
A genuinely personalized visual - one that speaks to the company's industry and its current initiative - normally requires a designer, a brief, and days of turnaround, which is why nobody does it for outbound. The research burden compounds the problem: knowing what each account's 'current moment' actually is means reading their announcements, launches, and hiring before any creative work can start. Ten accounts means ten research memos and ten briefs, and the play dies in the queue.
And because these are your best accounts, nothing should send without a human pass. A generated image and opener that miss the tone at a top-ten account cost more than they save. The workflow has to end in drafts a person approves, with the whole package inspectable before a single send. The review pass is what makes the ambitious version of this play safe enough to run at all.
How the mission runs
- Rank the accounts with Intent Data. The agent takes your target list and orders it by live intent, selecting the ten accounts with the strongest current buying signals. Anchoring on intent means the expensive personalization lands on accounts demonstrably in motion, where a distinctive touch has the best odds of converting attention into a reply.
- Research each company's moment with Web Research. For each of the ten, Web Research builds a picture of the current initiative: recent announcements, product launches, expansion moves, public statements. The output is a per-account brief naming the moment worth speaking to, which grounds both the visual and the opener in something real.
- Generate the personalized visual. Image Generation produces a landscape visual for each account, composed around their industry and their researched moment. Each image is account-specific by construction, made from that company's brief, so the recipient sees something that could only have been made for them.
- Write the three-line opener. The agent drafts a three-line opener per champion that connects their researched moment to the reason you are reaching out, with the visual as the centerpiece. Three lines is a discipline: long enough to prove attention, short enough to be read entirely.
- Stage everything as Gmail drafts. Each package - recipient, opener, linked visual - is created as a draft in your Gmail. Nothing sends. You open each draft, judge the image and the message against your knowledge of the account, edit freely, and send only the ones that earn it.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
Ten Gmail drafts, one per highest-intent account, each containing a three-line opener grounded in that company's researched current initiative and a personalized landscape visual linked within it. Alongside the drafts, the per-account research briefs show what the agent found and why each creative choice was made. Every send remains yours: review, edit, and release account by account.
Make it yours
- Refresh monthly as intent reshuffles the top ten, so the accounts entering the window get the distinctive treatment while it matters.
- Swap the visual style per audience - a technical diagram register for engineering buyers, an editorial illustration for brand-side champions.
- Extend the package to a second touch: a follow-up draft referencing the visual, staged alongside the opener for a two-step play.
Frequently asked questions
Will the visuals feel generic despite the personalization?
Each image is generated from that account's specific research brief - their industry and their named initiative - so the inputs differ meaningfully per account. The drafts stage lets you judge the output before anyone sees it; an image that misses simply gets regenerated with a sharper brief or dropped.
Does anything send automatically?
No. The mission's terminal state is drafts in your Gmail, deliberately. For top-of-list accounts the human pass is the point: you control tone, timing, and whether each package sends at all. The agent does the research and production; the send decision stays with you.
How does the agent know what an account's current initiative is?
Web Research reads what the company has made public recently - announcements, launches, leadership commentary, expansion signals - and the brief cites what it found. If the brief for an account looks thin or misreads the moment, you see that before the draft goes anywhere, and can redirect it.