AI sales prospecting should help you find customers who fit for a reason—not produce another giant lead list for someone to clean up. The useful result is a smaller set of companies you can understand, challenge, and confidently act on.

Strawberry is a browser with built-in AI agents. It can learn from the context you already have, work across LinkedIn, company websites, connected apps, and public sources, and investigate the signals that matter for your business. That makes it possible to search beyond the standard fields in a shared lead database.

You do not need a perfect ICP or a fully automated outbound system to begin. Start with what you know about a good customer, review a real set of prospects with your companion, and use your feedback to sharpen the wider search.

Start with what makes a customer worth finding

A useful search starts with the problem you solve and the evidence that a company might need it. Your website, ICP, strong customers, poor-fit examples, won accounts, CRM, and previous research can all help Strawberry understand that judgment without making you explain everything again.

The criteria can begin rough. Tell your companion what you know, what you are unsure about, and which exclusions matter. It can help turn that context into a working definition of fit, then improve it from real examples.

Find me 30 new leads

Use this to define the search, find a first set of prospects, and refine the result before scaling.

I want to find 30 new leads.

Choose between a wide search and a deep one

A broad pass can screen more companies quickly and at a lower credit cost per prospect. A deeper pass can investigate harder-to-find signals, connect evidence across sources, and make a stronger case for why an account fits.

You rarely need to choose only one. For a large market, Strawberry can screen widely using lighter signals, narrow the pool, and spend the deeper research on the companies most likely to matter. Agree on the likely depth, scale, and destination before starting a large run.

Look beyond the usual lead databases

Useful sales signals often live in the messy edges of the web: company pages, niche directories, public databases, association sites, job listings, product changes, and other sources specific to the market. They may reveal a problem, technology choice, expansion, or moment of change that a standard company profile misses.

Because Strawberry is an agentic browser, it can open the real sources, follow useful links, search within sites, and keep digging when the first result is incomplete. It can combine that public evidence with relevant context from your CRM, files, tabs, and connected apps, while keeping the source behind the important findings.

Calibrate the judgment before scaling the list

When the criteria are new or subjective, ask Strawberry for a small, varied first set. Review why each company was included, look at a few borderline examples, and tell your companion what feels right or wrong. If the definition is already trusted and the search is straightforward, you can move faster without forcing an extra review step.

This loop matters because prospecting depends on judgment. Your feedback can sharpen the criteria, sources, fields, and level of research before time and credits are spent on a larger run. Once the approach works, save those decisions as a custom skill so the next search starts further ahead.

Make every prospect easy to trust or reject

A useful prospect list shows more than a name and a score. Keep a concise reason for fit, the important source links, requested company or contact fields, and visible uncertainty beside each prospect. Check the CRM and existing lists for duplicates, and never pad the result with weak matches to reach a number.

If you need people or contact details, research them only to the depth the next step requires. Confirm that someone still holds the relevant role and keep verification status visible for contact information. The result can stay in chat or land in a table, spreadsheet, CRM, document, or dashboard—wherever it is easiest to review and use.

Enrich a lead or account list

Use this when the companies are already known and the list mainly needs contacts or missing fields.

I have a lead or account list that needs better data. Help me decide which fields would make it useful, enrich a small sample for review, and then update the rest after I confirm the approach.

Turn an approved list into the next useful step

Once the prospect set feels right, your companion can research the strongest accounts more deeply, enrich the records you need, or build a customer outreach sequence in your voice. Finding prospects, adding them to the CRM, drafting messages, and enrolling or sending are separate decisions—you stay in control of each step.

Share the approved list when teammates need the result, or save and share the accepted sourcing method as a team skill when they should reuse the process. If the team needs the same customer context across prospecting, research, outreach, and other sales work, share the full companion. After the process has worked several times, it can become a Routine that looks for new companies or useful signals at the right cadence. The goal is not simply more leads. It is a prospecting system that gets better at recognizing the customers your business can genuinely help.

Send personalized outreach that sounds like me

Use this after the prospect set is approved to draft and send reviewable outreach.

Help me send personalized outreach to this list. Use what you know about my voice and context, research only where it will help, and start with a small set I can review. Before sending, show me the exact recipients, messages, channels, and timing—and explain when thoughtful follow-ups or a sequence would be useful.

Official Strawberry skill

name: find-new-customers
description: Find and qualify prospective customers from the user's offer, best customers, target market, geography, buying signals, or other criteria. Use when someone wants to discover a targeted set of new prospects or improve how they source them. If they already have a list and mainly need contacts or missing fields, use enrich-lead-list instead.

Find New Customers

Find prospects who fit for a reason, not just a long list. If the user already has a sourcing skill shaped around their business, follow that instead.

Workflow

1. Understand what a good prospect looks like

Start with the result the user needs and what makes a prospect worth their time. Use relevant tabs, connected apps, files, memory, and prior work already available before asking them to repeat themselves.

Good prospecting depends on context about how the user's business works: what they sell, who gets value, strong and poor-fit customers, exclusions, CRM history, and feedback from earlier searches. Use that context to judge fit the way the user does, rather than relying only on generic filters. If a useful source is unavailable, briefly explain what connecting it would improve without making setup a requirement.

Find the kind of prospect the user actually sells to. This might be an organization, school, location, property, team, individual, or something else. Research specific people or contact details only when the requested result needs them.

2. Propose the search and first sample

For someone new to Strawberry or this workflow, propose:

a concise definition of who you will look for, with any important uncertainty made clear; and
a short plan covering the first sample, feedback, wider search, and delivery.

Include sources, breadth versus depth, approximate scale, useful fields, destination, and review point only when they matter. Explain tradeoffs comparatively; give specific time or credit estimates only when they can be estimated reliably. For a trusted repeat workflow, keep the plan brief and reconfirm only material changes or a substantially larger, deeper, or more expensive run.

3. Find and calibrate an initial set

Choose a search approach that fits the situation. It may use explicit criteria, lookalikes, a market or account universe, signs of a relevant problem or change, or a combination that becomes clearer through research.

Search where the useful evidence is likely to live. Because Strawberry is a browser, it can follow promising paths across LinkedIn, company sites, connected apps, public databases, niche directories, and other web sources rather than relying on one shared lead database.

When the criteria are new, subjective, or expensive to apply at scale, begin with a small and varied set. Explain why each prospect appears to fit and preserve the important sources. Include borderline or rejected examples when they would help the user correct the approach. Do not force calibration when the user already has a trusted rubric and straightforward scope.

Require evidence of fit for every prospect. Treat a credible timing or "why now" signal as helpful when it exists, not as a universal requirement. Use lighter, clearly labeled evidence for an initial screen; go deeper and cross-check material claims for the final shortlist.

4. Expand the accepted approach

Use the user's feedback to improve the criteria, sources, fields, and research depth before expanding. Do not pad the result or silently loosen the criteria to reach a requested count. If too few prospects meet the bar, explain why and offer useful choices: broaden a criterion, search other sources, accept more uncertainty, or keep the smaller list.

When a CRM, existing list, prior work, or useful memory is available, check it for duplicates and prior contact before deep research. Flag existing prospects rather than silently excluding them unless the user asked to see only new ones.

When working in LinkedIn, narrow the search before opening many profiles, work sequentially at a human pace, and stop on any warning, challenge, rate limit, or unexpected state. Parallelize only independent off-platform research.

5. Deliver a result the user can trust

For every prospect, include:

the prospect's name or other clear identifier;
a concise reason it fits;
the material evidence and source links.

Make important uncertainty or confidence caveats visible when they could affect how the user interprets or acts on the result.

Add timing signals, requested fields, contact details, prior-relationship status, or a suggested outreach angle only when relevant. If people or contact details are included, confirm that each person still holds the relevant role and keep verification status visible. Never invent missing companies, people, profile URLs, contact details, or buying signals.

Prefer a compact table when delivering more than a few prospects. Include how many were reviewed or filtered when that context is meaningful. Before delivery, check duplicates, missing values, stale roles, inconsistent formats, broken links, and low-confidence fields. Keep observed facts separate from inference.

Deliver to the agreed destination. Default to a reviewable result in chat when no other destination matters.

6. Improve and reuse the workflow

Finding prospects, drafting outreach, adding CRM records, and enrolling or sending messages are separate steps. Do not change external records or contact anyone without the user's approval. If the user wants to contact approved prospects, use strawberry/sales/send-personalized-outreach.

After the user has reviewed and improved a useful result, offer to save the accepted criteria, sources, fields, destination, and review points as a custom skill. When useful, also offer to share the approved prospect list or accepted sourcing method with the relevant teammates in their agreed destination. Keep sharing the list, sharing the method, and changing external systems as separate choices. When teammates need the same context and several Sales workflows—not only this sourcing method—offer to share the full companion. Once the workflow works reliably and has a useful cadence or trigger, offer to turn it into a Routine.