AI Prompt

Research an enterprise account and write the first email in one pass

Tools: Claude ▪ Lusha

Personal details in live examples on this page are masked: names to initials, the account described by industry and size, email domain hidden, no phone numbers shown.

Enterprise account research usually means four tabs and an afternoon: news, the org chart, contact details, then a blank email. This prompt runs the whole sequence through the Lusha connector for Claude in one pass. It pulls the account’s dated signals, maps the buying group, confirms the people are still in the seat, reveals only the work emails you need, and writes the first email around the strongest signal.

It also sets aside the signals that would put a rep on the phone with the wrong sentence: readings that contradict each other, and news about events that happened years ago.

The prompt

Open in Claude Open in ChatGPT

<context>
I sell: [your product, in one line].
The team that buys it: [department], signed off by [title].
Target account: [company domain].
</context>

<task>
Work this account end to end with Lusha, in this order.

1. Credits. Run account_usage, which is free, and estimate the run before spending anything.
2. Account brief. Pull the account's signals for the last 90 days. List the ones I can open with, each with its date. Set aside any reading that contradicts a later reading of the same signal, and any news item whose event date is much older than the article date, and say why.
3. Buying group. Map the buying group at the account. If the decision maker sits outside [department], search the account for the highest-ranking [department] leaders and pick the one closest to what I sell.
4. Confirm the people. For the decision maker and one champion, check they are still in the seat and when they started, and reveal work emails only. If only a personal email comes back, don't use it, and give me another route to that person.
5. First email. Write one email to a person with a verified work email, under 90 words, opening with the strongest dated signal and connecting it to what I sell. No flattery and no "hope this finds you well".
</task>

<output>
- Account brief: signals to use and signals set aside, with dates
- Buying group table: name, title, role, in the seat since
- The first email, with a subject line
- A next step for anyone without a verified work email
- The credits the run used
</output>

<constraints>
- Do not send anything or write to my CRM.
- Do not reveal phone numbers unless I ask.
- If the same person appears twice, merge the records.
</constraints>

What you'll get back

What you’ll get back

The situation

We ran the prompt on October 25, 2026 against a public identity security company with 7,938 employees, as a seller of a testing and monitoring platform for AI agents. The team that buys it is engineering.

Account brief: five signals to use

Signal to open with	Date	What it says
Product build	Oct 14, 2026	Building single sign-on for AI agents
Hiring surge	Oct 12, 2026	106 new jobs posted in a week, 14% above the account’s average
Partnership	Sep 24, 2026	An AI agent security alliance with a major cloud provider
Headcount growth	Oct 1, 2026	7,938 employees, up 10% over 12 months
Acquisition	Jul 30, 2026	Bought an AI security startup

Three signals set aside

Set aside	Why
IT spend increase of 245%, Aug 8, 2026	Two later readings went the other way: down 20% on Sep 3 and down 23% on Oct 3. Three readings that disagree are not a budget story to open with.
A legal settlement in the news, article dated Sep 22, 2026	The event itself dates to 2022. The article is new, the event is not.
A second acquisition story, article dated Jul 30, 2026	Its event date is December 2023, so it describes an old deal.

Buying group: who to contact

The buying group came back with eight people: one decision maker, five potential champions and two end users. The champions were AI software and automation engineers, a direct match. The decision maker was a director of technology alliances, outside engineering, so the prompt ran a title search for engineering leaders at the account. It found 13, including one senior leader listed twice under two records, which it merged.

Contact	Title	Role	In the seat since	Work email
K.K.	VP of AI Strategy and Forward Deployed Engineering	Decision maker, from the engineering title search	Aug 1, 2026, an internal move from a director-level role	Not returned. Lusha returned a personal address only, which the prompt does not use
S.K.	Senior AI Software Engineer	Potential champion, from the buying group	Aug 1, 2026, joined from an AI startup	s.k@[company].com, confidence A+, updated Oct 25, 2026

The first email

Subject: Testing agents before they get their own sign-on

Hi S.,

Saw that [company] is building single sign-on for AI agents, and that you joined the AI engineering team in August. Once agents carry their own identity, every new tool call is a new access path to test before it reaches a customer.

We help AI engineering teams replay agent runs and catch failing tool calls before release. Worth 20 minutes to compare how you test agent access today?

[Your name]

Next step for the decision maker

Lusha returned a personal address for K.K. and no work email, so the prompt left it out of the outreach. The options it gave: reveal K.K.’s phone number for 5 credits, or ask S.K. for an introduction once the first conversation happens. K.K. moved into the role on August 1, 2026, which is itself a reason to reach out in the first quarter of a new remit.

Credits the run used

47 credits: 36 for the signal events in the account brief, 8 for the buying group, 1 for the engineering title search, and 2 for email reveals. The credit check at the start is free.

Contacts confirmed live via Lusha connector, October 25, 2026. The seller and product in this example are illustrative. The account, signals, people and credit counts are from the live run.

Why it works

Each step in Lusha feeds the next one inside the same Claude conversation, so nothing gets copied between tabs and nothing goes stale between steps. Signals arrive dated, which lets the prompt keep the ones a rep can open with and set aside the ones that contradict each other or describe old events. Buying group roles show who decides and who champions, and a title search fills the gap when the decision maker sits in the wrong department. Seat checks and reveals come after the shortlist, so credits go only to the people you will contact. Lusha collects business contact information only and does not scrape social platforms: data sources and the Trust Center.

FAQ

What if the buying group doesn’t match the team I sell to?

It happened in our run. The buying group ranks people against your account’s Target Profile, so the decision maker can sit in a different department from the one you sell to. The prompt checks this and runs a title search for your department’s leaders when it does.

Why did the decision maker have no work email?

Lusha returned a personal address only. The prompt does not use personal addresses for outreach, so it offers a phone reveal or an introduction through the champion instead.

Why set aside a 245% IT spend increase?

Because the next two readings of the same signal showed decreases. One reading that the following two contradict is not something to put in a first line. More on reading signals: five checks before you trust a buying signal.

What does a run cost?

Our run used 47 credits. The account brief is the largest part, at 1 credit per signal event returned. A buying group costs 1 credit per contact returned, a search 1 credit per 25 results, an email reveal 1 and a phone reveal 5. Current plans: lusha.com/pricing.

Can I run it on a list of accounts?

It is built for one account per pass, so each brief gets read before anything is revealed. To pick which accounts to run it on, start with the stacked signals play.

Ready to run this?

One data connection. Works in Claude, ChatGPT, your CRM, or any agent you build.

Start for free Read the Docs