Find the right accounts, research them, reach out, and learn from every campaign
Evidence status: Capability pattern grounded in current Oddlynew organizations, types and metadata, search, API and MCP server connections, the Gmail integration, Loops and Lenses. It is not presented as a customer testimonial.
Who this is for
Founders, sales and growth teams, and the people who build their outbound and account research with agents: anyone who keeps a list of companies and people to reach, researches each one before writing, and wants the next campaign to start from what the last one taught.
Outcome
Every account and every contact you care about is one document, with typed fields your agents can filter on: industry, size, the signals that make it a fit, its score, its owner and its stage. Research, outreach and replies land on that document, so anyone on the team, and any agent, picks up an account where the last one left it. Each campaign is written down with what it tried and what it got, and the next one starts from that.
The working pattern
Describe who you sell to. Ask your agent to write the ideal customer and the disqualifiers into one document, and to create the types you need: Account and Contact, with the fields you filter and score on, and Campaign.
Find the accounts. Ask in plain words, for example "companies with more than 30 sales reps and a free trial on their website" or "companies hiring for international expansion". Your agent searches the web with its own tools and any data source your organization connected, and writes one Account document per company with its sources.
Bring your own data sources. An administrator connects the data providers you already pay for once, as an API from its OpenAPI description (Connect an API to your organization) or as an MCP server (Connect an MCP server to your organization), and every member's agent then uses that connection. When one source has no answer, your agent asks the next.
Research and score. For each account, the agent reads what is public, notes the signals it found with their links and dates, and sets the score from the rules in your ideal customer document. Change the rules, and the agent rescores.
Write and send. The agent writes a message per contact from that research and leaves it as a draft in the mailbox your organization connected. A person reads the drafts and sends them, or approves sending.
Learn from each campaign. Replies, meetings and bounces go back onto the contact and the campaign. A Loop with one measure, such as meetings booked per hundred messages, compares campaigns and writes what worked into the ideal customer and messaging documents.
Keep it current. The same Loop rechecks your best accounts on a schedule for new signals, such as a new job opening or a new leader, and marks what changed.
Useful connections and plugins
Web research and browser tools in your agent: for finding and reading what is public about a company or a person.
Your data providers: connected to the organization as an API or an MCP server.
Gmail: the organization connects a mailbox once; agents read replies and write drafts there.
Your CRM: connected as an API, so accounts and stages stay in step.
Spreadsheet tools: for importing a list you already have.
Availability depends on your client and your organization's connections.
A good first task
Ask your connected agent: "Write our ideal customer into a document from what you know about us, and create Account and Contact types with the fields we should filter and score on. Then find twenty companies that fit, one Account document each with its sources and a score, and show me the five best with why. Draft one message for a contact at each of the five, and leave the drafts for me to read."
Boundaries
Personal data and outreach law. Contacts are personal data, and cold email is regulated differently by country. Keep contacts in a private organization, record where each address came from, and follow the rules where your recipients live. Oddlynew is not a compliance approval (How to Start).
People send, or approve sending. Agents write drafts. Sending in your name is an action an organization can put behind an administrator's approval (Setting up your client).
Sources age. A signal is as current as its date; the agent says when it found it.
Everyone in the organization sees every account. Use a separate organization when an account list belongs to a different audience, such as one client of an agency (One audience per organization).
Data you buy stays under its license. What a provider lets you store and for how long is in your contract with it.
Why it compounds
A spreadsheet holds the list; this holds what you learned. Every campaign leaves its result on the accounts it touched and in the documents that say who to target and what to say, so the next campaign, and the next person or agent, starts from what worked. Nothing is locked into one tool: any agent your team uses reads the same accounts, and the knowledge outlives every campaign and every tool change.