How to Find Customers on X and LinkedIn: The 5-Step B2B Playbook

By The Odeist AI team · 2026-10-08 · 8 min read

Short answer: to find customers on X and LinkedIn, write down the questions your buyers ask before they buy, turn each one into a saved search on both networks, and reply to the fresh posts from real buyers with a useful answer, not a pitch. Do it on a fixed daily routine, have one person approve what goes out, and prune your searches every week. That is how B2B teams like DataGait reach their buyers: by answering the questions those buyers already ask in public.

How do B2B teams like DataGait find customers on X and LinkedIn?

DataGait sells to data engineers. Here is how the team describes its approach, in its own words:

"Odeist AI became our outbound team on X and LinkedIn. It finds data engineers asking the questions DataGait answers and replies in our voice. The Odeist AI team tuned that voice with us on day one."

That one sentence holds the whole method. A clear buyer (data engineers). A trigger (a question the product answers). An action (a reply that sounds like the team). No bought list, no cold messages to strangers.

The Applier uses the same pattern for a different buyer: "Odeist AI puts The Applier in the LinkedIn and X threads where job seekers ask for help. More new users now find us through those replies."

The rest of this guide is that loop, step by step. The searches and replies below are our own examples for a made-up company that sells a data pipeline monitoring tool to data engineers. They are not DataGait's.

Step 1: Which questions do your buyers ask before they buy?

Most teams start with keywords. Start with questions instead, because that is what people post.

Mine them from places you already have:

• Sales and demo calls. Ask every new customer: "What were you trying to fix when you started looking?"
• Support tickets and onboarding emails. The first question a new user asks is often the one they posted publicly a week earlier.
• Competitor names. People post when they are leaving a tool.
• Your own FAQ page. Rewrite each question the way a tired engineer would type it.

Aim for 15 to 20 questions, then sort them by how close the person is to buying:

Shortlist and switching posts are the warmest. Pain and method posts are the most common, and they are where people first see your name.

Step 2: How do you turn those questions into searches on X and LinkedIn?

Each search is a phrase from the question, a word from your field, and a few words to exclude. The two networks read searches differently.

On X, wrap phrases in quotes, use OR for alternatives and a minus sign to exclude a word. -filter:links drops posts with links, which are often promotion, and lang:en keeps one language. After searching, switch to the Latest tab.

• "schema change" (broke OR broken) dbt -filter:links lang:en
• "how do you" testing "dbt models" -hiring
• "alternative to" ("data quality" OR observability) -filter:links

On LinkedIn, the search bar takes Boolean strings: quotes for exact phrases, AND, OR and NOT typed in capitals, and parentheses for grouping. LinkedIn's help page says minus signs and wildcards aren't officially supported, so use NOT. Then pick Posts, sort by Latest and narrow Date posted to the past week.

• "data quality" AND ("any recommendations" OR "what do you use")
• ("schema change" OR "schema drift") AND (broke OR broken) NOT hiring

Then test each search. Run it and count how many of the first 20 results are a real buyer asking a real question. Keep the searches that find at least three. Rewrite or drop the rest. You should end up with eight to twelve searches worth running every day.

Step 3: Which posts deserve a reply today?

A good search still returns more posts than you should answer. Score each one from 0 to 2 on five things before you write a word:

Reply to anything at 7 or more. A LinkedIn post can still score on freshness after a day or two, since conversations there move slower. For more on reading a post's author and fit, see how to find leads in LinkedIn and X conversations.

Step 4: What should a reply to a buyer actually say?

The basics of a good reply (show you read the post, give one useful thing, end with a question) are in our playbook for getting customers from X replies. What changes in B2B is the stage. A pain post wants a diagnosis. A shortlist post wants honest options. Pitching to someone who only asked how others do it is how a reply gets ignored.

Here is an example for the method question above:

If they answer and ask what you use, that is the moment to mention your product, and to say you made it: "Full disclosure, I work on [product], which runs that check for you. Happy to share how to set it up by hand too."

On X, keep it to two or three lines. On LinkedIn, a few short sentences and one question read naturally. On both, write like an engineer talking to an engineer: plain words, one specific detail, no adjectives doing the selling.

Step 5: How often should you search and reply?

Threads go cold, so this works as a routine, not a campaign. Here is a sample week for one person:

Judge the channel on conversations and sign-ups, not likes, and give it a month before you decide.

Who should approve replies before they go out?

Every reply speaks for your company in public, so one person should own the yes. A simple rule set:

• Start by approving every reply. You will learn fast which drafts are safe.
• Always keep these for a human: complaints about your own product, threads about pricing or security, and angry posts about a competitor.
• Hand over the rest gradually, once the drafts stop needing edits. More on drawing that line in what an AI marketing agent should and shouldn't do without you.

This is the step where most teams stall, and where Odeist AI fits. Its X and LinkedIn agents run your keywords on a schedule, draft each reply in your voice, and send each draft to the app, WhatsApp, Telegram or Slack as a card with Approve, Edit and Skip. See how approving from WhatsApp works.

What makes this stop working?

• Searching for your category instead of the question. Nobody posts "seeking data observability platform".
• A search list that never changes. Buyers' words change. Your list should too.
• Replying to everything. Ten good replies beat forty thin ones.
• Counting likes. A reply with no likes that starts a conversation with a buyer is the win.
• Hiding who you are. If you work on the product, say so when you mention it.

How does Odeist AI run this playbook for you?

Odeist AI is a team of AI agents that finds the people looking for what you sell, answers them in your voice and reports back. For this playbook, it takes over the searching, the drafting and the daily routine.

• It learns your buyers and voice. Paste your website and it writes six brand documents, including Buyers and Voice, in about a minute. It can also learn how you write from your own posts.
• It finds the questions. The X and LinkedIn agents look for people asking for what you sell or about your competitors. You set the keywords, your rules on who to answer, and how often they look.
• It drafts in your voice. Each reply starts from your brand documents.
• You choose the switch. Each agent has "Ask me first" or "On its own". Every skip and edit teaches the agent a rule it follows next time.
• It costs little per reply. A reply is 2 credits, plus a little for each search. Every plan gets every agent, from $29 a month. See pricing.

Chivvy put it this way: "When Chivvy was two people doing everything, Odeist AI was our social media team. It found people asking for what we’d built." And the wrkr: "We’re pre-launch and Odeist AI is already growing our waitlist, joining conversations about flaky end-to-end tests."

Try it with 7 days of Growth and 1,000 credits, no card needed. Start free, or enter your website and Get a free growth report to see what your marketing is missing.

Whether you use a tool or not, the loop is the same: know the questions, search for them every day, and be the most useful reply in the thread.