Auxot
Tutorials Docs Pricing Login

Triage and follow up on inbound leads

Build an agent that reads new sales leads the moment they come in, scores them by fit and timing, drafts a personalized first response, and surfaces who needs a follow-up — so leads stop sitting cold.

Plus: three prompts that turn 'we'll get back to you' into a same-hour first touch and a follow-up rhythm that doesn't depend on your memory.

Audience Everyone · Admins
Time ~15 min
Prerequisites An Auxot account on any tier. [Create an agent from scratch](/tutorials/create-an-agent-from-scratch) finished — you've built at least one agent. Helpful: [Add your first context file](/tutorials/add-your-first-context-file) if you want to give the agent your ideal customer profile and product positioning, [Connect Slack to your agents](/tutorials/connect-slack-to-your-agents) if leads come through Slack, [Add an MCP server](/tutorials/add-an-mcp-server) if you want to plug into a CRM.
You'll end up with A Lead Triage agent that reads inbound leads and scores them, drafts personalized first responses, and a clear path for who to follow up with — tested on a real recent lead.

When a tutorial shows italic text in quotation marks, it usually mirrors a label or helper string inside Auxot. Product copy changes between releases — if something reads differently in your workspace, trust what you see on screen.

Callouts with a Worth knowing gold accent are meant as must-read context before you move on. Blockquotes that open with Tip are lighter, optional depth.

Why this matters

Inbound leads cool fast. A form fill answered in the first hour converts much better than the same lead answered 24 hours later, and a lead that sits 48 hours is mostly cold. Most small teams know this and still let leads sit, because the speed is gated on a human reading every form fill, judging fit, and writing a personal first touch. The labor is real.

You don’t need a generic auto-responder. You need an agent that reads the lead like a salesperson would (does this person fit our ideal customer profile, are they likely to buy soon, and what should the first message say given who they are), and produces a draft you can send in under 30 seconds. Today, you build that. Leads stop sitting. The drafts get sharper as the agent learns what’s converting.

Quick start

  1. Sign in: open Auxot in your browser and log in.

  2. Open chat with the Admin Agent: click Chat in the left menu, make sure the agent picker reads “Admin Agent.”

  3. Build the Lead Triage agent: paste this:

    I want a "Lead Triage" agent. When I paste in a new inbound lead — form fill, email, LinkedIn message — it should: (1) score the lead's fit against my ideal customer profile from 1–5, (2) score timing urgency 1–5 (how soon should I respond — same hour, same day, or this week), (3) draft a personalized first-touch response in our voice, and (4) recommend the next step (book a call, send a resource, or deprioritize). Give it a clear job description and make it ready to use.

  4. Answer the Admin Agent’s questions: what your ideal customer profile looks like, what your voice is, and how you currently handle leads. Two or three exchanges.

  5. Test on a real recent lead: paste a form fill or email from the last week into chat with the agent. Watch what it does. Compare to what you actually did.

Done? The agent is in Settings → Agents and selectable from the agent picker. Optional next step: wire it to your real lead source: Slack channel, CRM, or email inbox.

The agent can do that?

Lead triage gets dramatically better with three small additions most people skip. Each one takes 30 seconds to set up.

1. Have the agent score timing AND fit, not just fit

Open chat with Lead Triage and ask:

When you score a lead, separate fit (1–5: how well do they match our ideal customer profile) from timing (1–5: how urgent is the response — 5 means respond within an hour, 1 means same week is fine). Don't collapse them. A high-fit lead with low timing urgency gets a different response than a low-fit lead with high urgency. Show both scores at the top of every triage.

Why it’s non-obvious: Most lead-scoring collapses fit and timing into one number, which loses the most useful signal. “They asked specifically about pricing for 50 seats” is timing. “They’re a 200-person company in our ideal customer profile” is fit. Different responses, different urgency. Splitting the score lets you triage your day: respond to high-timing first, work the high-fit-low-timing into your weekly nurture flow.

2. Have the agent draft three first-touch options by tone

After the first week, ask the agent:

For every triage from now on, draft three versions of the first-touch response: a warm version (lead with empathy, ask one diagnostic question), a direct version (assume they want to buy, propose a specific next step), and a curious version (ask one sharp question that surfaces what they're really trying to solve). Label each. I'll pick one and edit.

Why it’s non-obvious: Picking-and-editing a draft is faster than writing-from-scratch and faster than approving-or-rejecting one draft. The three-version pattern means you almost always have the right starting point in front of you. The agent learns over time which one you keep picking and starts ordering them by likelihood.

3. Build a “Follow-Up Tracking” agent for non-responders

Inverted-usage move. Open chat with the Admin Agent:

Build me a "Follow-Up Tracking" agent. Its job is to track leads I've responded to and flag when I haven't heard back in [3 days for high-timing, 7 days for medium, 14 days for low]. For each, recommend whether to send a nudge, deprioritize, or book a different angle (e.g., "they asked about price; maybe send the case study instead"). Run weekly. Attach it to the same context files as Lead Triage.

Why it’s non-obvious: First-touch is the part everyone obsesses over; follow-up is where most deals are actually lost. The follow-up-tracking agent does the routine work you’d do if you had time. Pairs perfectly with Set up your Monday morning briefing: the follow-up-tracking report becomes one section of your weekly briefing.

Go deeper

The two scores: fit and timing

A lot of sales-tooling lore treats lead scoring as a single number (“MQL score: 73”). For triage, split it:

  • Fit = how well this person matches your ideal customer profile (industry, size, role, or problem they’re describing). Stable. Doesn’t change much between when they fill the form and when you respond.
  • Timing = how urgent your response should be. “We’re evaluating vendors this week” is a 5. “Just exploring” is a 2. Volatile: answers in their form fill or email tell you almost everything.

Fit decides whether to respond at all. Timing decides how fast.

Where leads come from determines the wiring

Pick the source that matches where your leads actually arrive:

  • Form fills (most common): wire your form to drop submissions into a Slack channel (Connect Slack to your agents); Lead Triage watches it. Or, if you’re on a CRM with an MCP server (Add an MCP server), pull from the lead object.
  • Email inbox: an MCP server that reads new emails matching certain criteria (sender domain, subject keywords).
  • LinkedIn messages: mostly manual today. Paste them in. The agent is still useful for the drafting and follow-up tracking.
  • Multiple sources: pick one to start. Don’t try to wire all three at once. Add the second after the first is producing usable triage.

Email reply wiring (Gmail and Outlook)

For Gmail, wire Google’s official Gmail MCP. It is free and draft-only by design: the agent’s first-touch reply lands in your Gmail Drafts folder; you skim and click send. The Drafts folder is the structural review gate. See Wire Gmail drafts through Google’s official MCP for the full wiring.

For Outlook, route through Composio Outlook with the agent’s tool policy restricted to draft tools only (Create email draft, Create draft reply). Exclude the send tools from the policy explicitly; otherwise the agent can send autonomously. See Lock Outlook drafts via a draft-only tool policy for the full wiring.

The same draft-only discipline applies to follow-ups (power move 2 and the follow-up-tracking agent). Drafts go to your Drafts folder; you click send. Never wire the agent to send autonomously, even for the highest-timing leads. The gain from “in your inbox in 30 seconds” comes from the draft already being right, not from skipping the human send. Keep send_email out of your agent tool policies walks through why this stays true even when the lead is high-timing.

The “first response in under 30 seconds” rule

The whole point of the agent is collapsing the time between lead arrives and first response sent. If your workflow is “agent triages → I switch tabs → read the draft → edit → switch tabs again → send”, you’ve added latency, not removed it. Two ways to keep it fast:

  1. Same-tool sending. If the agent’s draft lands in Slack DM, reply to the lead from Slack (with the right integration). One tab.
  2. Send the draft as-is. For the highest-timing leads, the agent’s draft is the response. Skim and click send. The 5-second loss in personalization is dwarfed by the gain from responding in 4 minutes instead of 4 hours.

Troubleshooting

  • The agent’s drafts sound like every cold-outbound email you’ve ever ignored. Its context file doesn’t have enough of your voice. Same idea as Worth knowing above: paste 5–10 examples and tell it to match.
  • Every lead gets scored 4 or 5 on fit. The agent’s definition of an ideal customer is too loose. Edit its description to include explicit examples of low-fit leads (“a freelancer asking about enterprise pricing is a 1; a 5-person agency with no AI experience is a 2; a 200-person company with budget approved is a 5”).
  • You’re getting good triage but still not responding fast. The agent isn’t your bottleneck: the loop from triage → draft → send is. See “first response in under 30 seconds.”
  • The follow-up-tracking agent recommends nudges that feel pushy. Its description needs your follow-up philosophy explicitly. “Never more than two follow-ups. Second one should add a new angle, not repeat the first.”

Variations & edge cases

  • Free tier: the agent works at all tiers. CRM integrations via MCP servers (Add an MCP server) work too.
  • Business-to-business vs business-to-consumer: Business-to-business benefits more from this pattern: fewer leads, higher value per lead, personalization matters. For high-volume consumer sales, sampled review (see Set up a content review pipeline) on the agent’s auto-sends might be a better fit than human-edited every lead.
  • Account-based selling: fit is mostly pre-decided (you’re working a list). The agent’s value shifts to drafting personalized first-touch and tracking follow-ups, not scoring.
  • When NOT to use this: if your leads are a few per month and high-touch, the time saved doesn’t beat the time invested in setup. Manual is fine. Build this when you’re seeing more than 5–10 inbound a week.

Walkthrough

Step 1: Open chat with the Admin Agent

Click Chat in the left menu. Make sure the agent picker at the top reads “Admin Agent.”

Step 2: Build the Lead Triage agent

Paste this:

Build a "Lead Triage" agent. When I paste in an inbound lead — form fill, cold email, or LinkedIn message — it does four things in order:

1. Scores the lead's fit against my ideal customer profile from 1–5, with one-line reasoning.
2. Scores timing urgency 1–5 (5 = respond within an hour; 1 = same week is fine), with one-line reasoning.
3. Drafts a personalized first-touch response in our voice.
4. Recommends the next step: book a call, send a specific resource, deprioritize, or add to nurture.

Give it a clear job description, ask me what context files to attach (ideal-customer-profile description, product positioning, and sample replies), and make it ready to use.

The Admin Agent will ask follow-up questions: who fits your ideal customer profile, what’s your tone, what counts as urgent timing, and what does deprioritized look like. Be concrete: “Our ideal customer profile is 50–500 person professional services firms in the US,” not “mid-market companies.”

Step 3: Attach the right context files

The agent needs three things to do its job well:

  • Ideal-customer-profile description: explicit examples of high-fit and low-fit leads.
  • Product positioning: one page on what you do, who it’s for, and what your differentiation is.
  • Sample replies: 5–10 of your best recent first-touches, the ones you’d want the agent to learn from.

Add your first context file covers the mechanics. The third one matters most: that’s what teaches the agent your voice.

Step 4: Test on a recent real lead

Pick a lead from the last 2–4 weeks where you remember what happened. Paste it into chat with Lead Triage. Verify:

  • Does the fit score match what you’d score it?
  • Does the timing score match the urgency you actually felt?
  • Is the drafted first-touch usable, or would you rewrite it from scratch?
  • Is the next-step recommendation what you actually did?

Adjust the agent’s description based on what’s off. “Your fit scores are too generous, especially for sub-50-person companies — tighten the ideal-customer-profile filter,” etc.

Step 5: Wire it to a real source

Once the agent is producing usable triage, pick one source and wire it up:

Start with Slack if you have it. Add the second source only after the first is producing reliable triage.

Step 6: (Optional) Build the follow-up-tracking agent

Once Lead Triage has been running for a week and producing usable output, head back to the Admin Agent and use power move 3 to build the follow-up-tracking agent. Wire its weekly report into your Monday briefing (Set up your Monday morning briefing) so you don’t have to remember to check it.

What’s next

Reference

Browse all tutorials