To recover a self sourced pipeline number that stands at 20% of target after month 1, an account executive needs one LinkedIn campaign that runs every working day and a fixed 30 minutes to answer the people who reply: 5 minutes on results, 20 on replies, 5 on one small fix. You do it from your own AI chat, on your own LinkedIn account, without waiting for an SDR or for marketing.
Picture Tomas, an AE who sells procurement software to hospital groups. His quarter has a self-sourced target of 300,000 in new pipeline. It is the last Friday of month 1 and he has opened 60,000, all from one former colleague. He has eight weeks, a full calendar of demos, and no hour in the day that he controls except the one before lunch.
This routine uses the GTM Bud connector in Claude or ChatGPT. The minute counts are our estimate for a typical day, not measured data.
Why 20% after month 1 is a worse position than it looks
It is worse than it looks because outbound has a delay, and the delay eats month 2. A person you reach today may accept next week, reply the week after, and take a call the week after that.
Count backward from the end of the quarter. An opportunity opened in week 12 counts. A conversation that starts in week 11 will not be an opportunity by then. So the real window to start conversations closes around week 9, and you are at week 5.
The usual AE answer is a burst: one Saturday of hand-written messages, then nothing for three weeks. A burst gives one wave of replies and no follow-up. A campaign that sends a steady number every day, with follow-ups on their steps, is the only thing that fits inside four working weeks.
The one-time setup: about two hours this weekend
Build one campaign, not three. One audience you already know how to sell to, one offer, one sequence.
The audience sentence. Take it from your last two wins. Tomas wrote:
“Heads of procurement and supply chain directors at hospital groups with 3 to 30 hospitals in the UK and Ireland.”
Type the sentence into your AI and ask for a campaign draft. Your AI writes the targeting criteria, and GTM Bud checks each one and saves the draft. Tomas had added “with a tender coming in the next six months”. The check flagged that line and suggested a rewrite, because a future tender cannot be read from a LinkedIn profile, a company page or public web content.
The preview. Ask for a prospect preview and wait one to three minutes. Mark each person Good fit or Not a fit. Tomas marked a procurement lead at a single private clinic as Not a fit, with the reason “one site, too small”, and told the AI that he never wants people like that one. GTM Bud keeps the mark with the campaign. A mark does not rewrite the criteria, so he also had the AI add a criterion for the number of hospitals.
The seller profile. For this draft, set your company website and your own personal LinkedIn profile URL. GTM Bud reads both so the messages are about you.
Offers and messages. Your AI gets the value offer brief, researches, and proposes offers. Pick one you can deliver alone, in under an hour, with no approval from anyone. For Tomas that was a one page comparison of what five product categories cost across the lead’s own hospitals, made from a price file the lead sends. Then your AI writes the message sequence, and GTM Bud checks and saves it.
Launch. Run the launch readiness check. It lists what is not met, for example a LinkedIn account with no sending schedule. The schedule and the account connection are set on gtmbud.com. Then ask for the launch, read the preview, and confirm.
The playbook. Before the first reply, save a short reply playbook: the goal is a 25 minute call with you, no pricing in a message, your free slots, and your booking link if you use one. GTM Bud instructs your AI to state only facts that the playbooks, the campaign and its offer, or the lead’s profile contain, and to use a booking link only if a playbook holds one.
The 30 minutes, minute by minute
| Minutes | Job | What you type |
|---|---|---|
| 0-5 | Read the numbers | “Results for my hospital campaign, next to the numbers I noted yesterday.” |
| 5-25 | Answer every waiting reply | “List replies that need an answer and write one for each.” |
| 25-30 | One fix or one note | “Remove this lead.” or a line in your own tracker |
Put it at the same time each day. Tomas chose 11:30, after the morning demos.
Minutes 0 to 5: what do the numbers say?
They say one of three things, and each has one response.
- Requests go out, few are accepted. The connection request carries no note, so the audience judges your profile and nothing else. Do nothing for a week. Then check the preview marks and your own profile.
- Accepted, but quiet. The first message is the problem, not the list.
- Replies arrive. Go straight to the next block.
The results show leads reached by channel (connection request, InMail, email), connection requests sent and accepted, replies, and replies by label. They refresh when you ask and can be up to about 15 minutes old. The result carries the exact time.
Do not read the numbers twice a day. Nothing you can act on changes in four hours.
Minutes 5 to 25: how do you answer replies in 20 minutes?
You let the AI write, and you spend your minutes on judgment. Ask for the replies that need an answer. For each one, the AI reads the whole conversation and the playbook, writes an answer, and shows it. You read, edit and approve. GTM Bud instructs the AI to send only the text you approved. Treat that as a rule the AI follows, and still read every word, because the message goes out under your name.
An example. A supply chain director wrote: “We are locked into a framework agreement until next April.” The AI wrote:
“April gives you time to see the numbers first. The comparison uses your own price file and takes me about an hour. Would it help to have it before your renewal talks start?”
Tomas approved it without a change. He then labeled the reply Warm, and wrote “April renewal” next to the name in his own tracker.
Some messages need no answer. An out-of-office note, or a bare “thanks”. The AI is instructed to say so and to write nothing, which saves you from sending a message that only exists to fill silence.
If a lead asks something the playbook does not cover, such as a security certificate, the AI should answer the part it can and ask one question. Do not let it guess. Add the fact to the playbook at the end of the session, and the next reply can use it.
Minutes 25 to 30: one fix
One, not five. Good fixes for this slot:
- Remove a lead you know personally, or a customer who slipped in. The connector has no function that excludes your existing customers, so you remove a name when you see it. To keep a whole company out of every campaign, add it to the Excluded companies list on your account page on gtmbud.com.
- Add one sentence to the playbook that today’s replies showed was missing. The AI shows you the new text and saves it after you approve.
- Label any reply that is still unlabeled.
Bigger changes do not belong here. The audience of a live campaign cannot be changed from the chat. To test a second audience, duplicate the campaign into a new draft on a Saturday and change the audience of the copy.
Does a LinkedIn conversation count as self-sourced?
It counts when your company’s rule says it does, so read the rule this week. Most plans count an opportunity as self-sourced when the AE created it from their own outreach and no SDR or marketing touch came first. Some require the first activity to be logged in the CRM under your name.
That last point matters here. The connector does not write to a CRM. When a lead agrees to a call, you create the contact and log the first touch yourself, on the same day. If your AI app also has a connector to your CRM, it can do both steps in one chat, but that is your AI and your CRM connector at work, not a GTM Bud feature.
A two-line habit protects your credit: log the contact when you label a reply Interested, and paste the date of your first message into the record.
Three limits an AE should know first
Your LinkedIn account does the sending. Messages go out from your profile, inside the sending schedule you set. By default, each LinkedIn account can send up to 40 replies a day, and replies from the GTM Bud Inbox count toward the same limit. A single campaign will not come near that.
Company rules come first. If your employer has a policy on outreach tools or on putting prospect data into an AI app, follow it. This article does not know your policy.
No promised number. We do not give a reply rate or a meeting count, because we cannot know your audience or your offer. The routine makes sure that a campaign runs every day and that no reply waits overnight. What that produces is for you to measure.
Trade-off: One campaign is easy to keep alive in 30 minutes and slow to teach you anything about a second segment. Two campaigns teach faster and break the routine. With eight weeks left, keep one.
For the math that connects pipeline to quota, read our guide to the pipeline coverage ratio. For more on working an inbox from a chat, see managing LinkedIn replies from Claude or ChatGPT. If your audience sentence feels vague, how to build an ICP for outbound will sharpen it.
Frequently asked questions about self sourced pipeline for an AE
What counts as self sourced pipeline for an account executive?
Self sourced pipeline is the value of the opportunities that the account executive opened from their own outreach, and not from marketing leads, SDR handoffs or partner referrals. Each company writes its own rule for the edge cases, so read your compensation plan before you count a LinkedIn conversation as self sourced.
Can an account executive use GTM Bud without the company buying it?
Yes, as a personal account with your own LinkedIn account connected. You add the connector to your own Claude or ChatGPT by its address and sign in. Check your employer policy on outreach tools and on customer data in AI apps before you start, because that decision is not yours alone.
How long before a new LinkedIn campaign produces a first conversation?
Plan in weeks. On the connection path, a request has to be accepted before the first message goes out, and follow-ups come days apart. Nobody can give you a fixed number, since it depends on your audience and your offer. That delay is the reason to launch in the first days of month 2 and not in month 3.
What if my company already has SDRs who work my accounts?
Then choose an audience the SDRs do not cover, and tell their manager which one. Two people from one company writing to the same buyer in the same week costs both of you the conversation. The connector has no function that excludes accounts another person works, so the split is an agreement between people.
Is 30 minutes a day really enough?
It is enough to keep one campaign healthy and every reply answered on the day it arrives. It is not enough to build the campaign, which takes about two hours once. It also does not cover the discovery calls that result, and those are the reason you are doing this.
Launch this weekend, and keep the 11:30 slot
Self sourced pipeline is won or lost in month 2, because conversations that start later do not become opportunities in time. Build one campaign this weekend, then hold 30 minutes a day: 5 on results, 20 on replies, 5 on one fix. Add the connector with the steps on GTM Bud for Claude, sign in to GTM Bud, and put the daily slot in your calendar before Monday’s first demo.