Lead Desk
Every new lead scored, understood and answered in draft — by the time you've made coffee. You send. Nothing else does.
What it is
A Claude Project you paste new leads into. For each one it scores fit against signals you wrote down, says what the person actually wants, drafts a short reply in your voice, and gives you one line for your spreadsheet. Then it stops. You read, change, and send.
It scores and drafts
1 to 5 against your signals — and "unknown" when a signal isn't in the message, never a guess. Then a four-line reply, not a page.
You send
From your own email or account, in your own words. The desk never sends, never follows up on its own, never invents a fact about the lead. That's the rule, written into it.
Your day with it
- 9:30
- Paste what came in overnight — the form, the email, the DM. Two minutes later: a score, what they want, a draft, and a ledger line for each.
- 9:35
- Read the drafts. Fix anything that doesn't sound like you. Send the good ones. Bin anything under 3 — that's the desk working, not failing.
- Friday
- "Who's gone quiet." Paste the ledger, get ten lines: which good leads went silent, and what the next line to them is.
Good for
The reply you meant to send
Tuesday's lead answered on Tuesday, not Friday. That alone changes who you close.
Knowing which one matters
Three leads, one afternoon. The desk says which two are real and why, so the afternoon goes to them.
Following up without nagging
Three follow-ups, spaced, in your voice, then stop. Never "just checking in".
A CRM that's one sheet
One row per lead, eight columns. Enough until you have fifty rows, and most people never do.
The honest part: the automated version of this exists, and it sends thousands of emails a day. This one doesn't, on purpose. In the markets we work, finding the buyer is a relationship and the reply is your reputation — neither survives volume. The desk gives you back the ten minutes; it doesn't take the decision.
Where do leads arrive?
Where do you keep them?
What happens to a lead that isn't a fit?
Your desk. Do these in order.
What happens.
- Week 1
- The scores feel harsh. They usually aren't. Fix the signals if the desk keeps scoring your best kind of lead a 3 — the instructions are yours to edit.
- Week 3
- You notice you're replying the same day. The ledger has twenty rows and you can see which source sends the real ones.
- Month 2
- "Who's gone quiet" catches a good lead you'd forgotten. That one pays for the whole thing.
What it won't do
- Find leads. That's still you — your network, your form, your introductions. It only works on what arrives.
- Send anything. Ever. If you want that, the automated version exists and it isn't this.
- Know things it wasn't told. If the message doesn't say who signs, the score says "unknown", not "probably".
Prefer to read it? The whole kit in plain text, and what we changed in translation.
The kit, in plain sight
This is the general version. The tool above builds yours, with your signals in it.
A lead desk is a Claude Project you paste new leads into. For each one it scores fit against signals you wrote down, says what the person actually wants, drafts a short reply in your voice, and gives you one line for a spreadsheet. You read, change, and send — it never does. Twenty minutes to set up, ten minutes a day to run. The automated version of this exists; this one keeps finding, sending and replying human on purpose.
# Lead Desk — how to run it I sell: [what you sell, to whom] Leads arrive by [website form / email / DMs / introductions]. I keep them in [a spreadsheet / my CRM]. A good lead shows these signals (in order of weight): 1. [e.g. they name a specific problem, not "interested in your services"] 2. [e.g. they're the person who'd sign, or they say who is] 3. [e.g. a timeline, a budget, or a trigger event] When I paste a lead (or several), do this for EACH one: 1. Score it 1–5 for fit. 5 = every signal present and specific. 1 = none, or clearly the wrong buyer. Say which signals you saw and which you didn't. Never guess a signal that isn't in the message — write "unknown". 2. Say what they actually want, in one line. If it's unclear, say so; that's a finding. 3. Draft the reply, in my voice, short. Score 4–5: answer their question, ask one qualifying question, propose the next step. Score 3: answer, ask two questions, no next step yet. Score 1–2: a two-line polite no. 4. Give me one ledger line, tab-separated: date · name · company · source · score · what they want · status (new) · next step. Rules: - You never send anything. Every draft is for me to read, change, and send myself. - Never invent facts about the lead, their company, or what we can do for them. - Short beats thorough. A four-line reply that goes out today beats a page that doesn't. - If two leads look like the same person or company, say so.
Lead desk. Here's what came in. Score, what they want, draft, ledger line — for each. --- [paste the lead(s) here]
date name company source score what they want status next step
Follow-up ladder for [name]. They haven't replied since [date]. Draft three: day 3 (one line, adds one useful thing), day 7 (two lines, a different angle), day 14 (one line, closes the loop with no pressure). After the third, we stop. My voice, no "just checking in".
Who's gone quiet. Here's my ledger. Tell me: every lead scored 4–5 with no reply in 7+ days, and what the next line to them should be. Then every lead sitting at "new" for more than 2 days. Ten lines, not a report. --- [paste the ledger]
What changed in translation
| In the source | Here | Why |
|---|---|---|
| Analyses what you sell by studying competitors | You write it in two lines | You already know. Two lines beat a model's guess. |
| Decides who should buy with fit scoring | Kept — your signals, scored 1–5, with "unknown" allowed | The scoring is the useful part. Guessing signals that aren't there is the dangerous part. |
| Finds decision-makers automatically | You find them | In the markets we work, finding is a relationship, not a query. |
| Writes and sends personalised emails at volume | Drafts one at a time. You send. | A reply you read is a reply you'd stand behind. Volume is the thing that breaks this. |
| Handles replies and books meetings | Drafts the follow-up ladder. Stops after three. | Persistence is a setting for a machine and a reputation for a person. |
| Optimises on performance data | A weekly "who's gone quiet" | Ten lines, from your own ledger. Enough to see what's working. |
Adapted from the six-step pipeline of Explee's AutoGTM (explee.com) — the automated version. This one is run by hand on purpose: finding, sending and replying stay yours.