Agents Lookup

Recipes / Sales & outreach

Inbound Lead Qualification and Reply Drafts

An agent scores each inbound lead against your written ICP, checks the company's public website, and drafts a reply with a suggested next step that a rep reviews and sends; nothing goes out automatically.

CategorySales & outreach
Steps9
Human sign-offs4

The problem

Inbound leads arrive through a website form, a shared inbox, or a chat tool, and in a small team they are handled unevenly. Good-fit leads wait a day for a reply because nobody owns the inbox, while time goes to students, vendors pitching services, and spam. Reps who do reply often start from scratch, look up the company by hand, and guess whether it's worth a call. This recipe adds a consistent first pass. When a lead comes in, the agent scores it against ideal customer profile criteria you have written down, looks only at the company's own public website for basic context, and drafts a reply with a suggested next step: book a call, send a resource, or politely decline. The draft and score land in your CRM or a Google Sheet. A rep reads it, edits, and sends. Nothing is sent automatically, and consent and anti-spam rules apply to every lead.

Who it's for

Not a fit when

Tools

Web form or GmailWhere leads arrive; a form submission, a webhook, or a Zapier or Make trigger on new emails to a sales address starts the workflow
HubSpot, Pipedrive, or Google SheetsStores the lead record, score, reasons, draft reply, owner, and status; a Sheet is a free starting point
Zapier or MakeMoves form or email data to the agent, fetches the public website, and writes results to the CRM or Sheet
Claude or a similar LLMScores against your ICP, summarizes the public website, and drafts a reply with a suggested next step
Website fetch stepRetrieves the homepage and one or two public pages of the lead's company domain for context; no paid data brokers or personal profile scraping
Slack or email alertsNotifies the assigned rep when a high-fit lead arrives
Scheduling linkOptional: included in replies the rep approves for qualified leads

Inputs you need first

Step-by-step build

  1. Write the ICP and scoring rubric

    Turn your ideal customer description into a scoring rubric of four to six criteria, each worth points. Example: industry in target list (0 to 3), company size in range (0 to 3), region you serve (0 or 2), stated use case matches your offer (0 to 3), timeline or budget mentioned (0 to 2). Add hard disqualifiers that set the score to zero: competitor, job seeker, vendor pitch, student project, outside your service area. Set bands such as 9 or more is high fit, 5 to 8 is medium, below 5 is low. The rubric is yours to adjust.

  2. Set consent and anti-spam rules

    Write down what each lead agreed to. A person who filled in a contact form asking about your product generally expects a reply to that request; that isn't the same as consent to marketing emails. Replies should answer their inquiry only. Don't add leads to newsletters or sequences unless they opted in. Honor any request to stop contact right away. Include your business name and a way to reach you in replies. Check the email and privacy rules that apply in the regions you serve, and don't contact other people at the company the lead didn't mention.

  3. Set up the lead record

    In your CRM or a Google Sheet, create fields for: lead_name, email, company_name, company_domain, source, message, consent_marketing (yes or no), fit_score, fit_band, score_reasons, website_summary, suggested_next_step, draft_reply, assigned_rep, status (New, Draft ready, Sent, Disqualified, Do not contact), sent_by, sent_at. In HubSpot or Pipedrive, use custom properties or notes for the agent output. In a Sheet, add data validation for status and fit_band so values stay consistent.

  4. Connect the trigger

    Start the workflow when a new lead arrives, using a form webhook, a native form trigger, or a Zapier or Make trigger on new emails in a sales inbox. Pull the sender's name, email, company, and message. Get the company domain from the email address or a form field, and skip enrichment for free email providers. Remove duplicates by email before creating a new record. Leads that are clearly spam, such as empty messages or link-only text, can go straight to a Spam status for a weekly human check instead of the agent.

  5. Enrich from the public website only

    Fetch the lead's company homepage and, if easy, one public page such as About or Services. Pass the page text to the agent to summarize what the company does, who it serves, and rough size signals stated on the site. Don't use paid data brokers, personal social profiles, or scraped personal contact details in this recipe. If the site can't be fetched or the domain is a free email provider, the agent works only from the form message and says so in website_summary.

  6. Score and draft the reply

    Send the agent the rubric, the lead's message, the website summary, and your reply templates. It returns a score per criterion with a short reason, the total and band, a suggested next step (offer call, send resource, ask a clarifying question, refer out, or decline), and a draft reply in your voice that responds directly to what the person asked. For high fit, the draft can include your scheduling link. For low fit, it drafts a short, polite reply. It never invents pricing, features, customer names, or results.

  7. Route to a rep for approval

    Write the output to the lead record with status Draft ready and assign a rep using your routing rules. Alert the rep in Slack or by email for high-fit leads, and include low and medium leads in a twice-daily digest. The rep reads the lead, checks the score reasons, edits the draft, and sends it from their own email or CRM inbox. They then set status to Sent with their name and time, or mark it Disqualified or Do not contact. No step in the automation sends email.

  8. Test with past leads

    Collect 20 to 40 past inbound leads, including ones that became good customers, ones that went nowhere, and obvious junk. Run them through the workflow with alerts off and have a rep score the same leads blind. Compare: where does the agent score high on leads the rep would skip, or low on leads that turned into real deals? Adjust rubric weights and disqualifiers, and tighten the reply templates where drafts sound generic. Repeat until rep and agent mostly agree on the band.

  9. Go live and maintain

    Turn on the live trigger and set a reply target, such as high-fit leads answered within business hours on the same day. Every month, review a sample of scored leads with a rep, check disqualified leads for false negatives, and update the ICP rubric when your offer or target market changes. Keep the Do not contact list respected across all tools. Check that the website fetch step still works and that the agent isn't using anything beyond the public company site.

Example agent instructions

Adapt this to your business. Placeholders in braces are filled by your automation tool.

You help qualify inbound leads for {company_name} and draft replies for a sales rep to review. You never send messages. A rep edits and sends every reply.

Ideal customer profile and scoring rubric:
{icp_rubric}

Disqualifiers (score becomes 0):
{disqualifiers}

Lead form data:
{lead_data}

Public company website text (may be empty):
{website_text}

Reply templates:
{reply_templates}

Instructions:
1. Summarize in two sentences what the lead's company does, using only {website_text} and {lead_data}. If website text is empty, say so.
2. Score each rubric criterion with a one-line reason. Do not guess facts that are not in the inputs; score 0 and write 'not stated'.
3. Apply disqualifiers first. Total the score and assign the band: high, medium, or low.
4. Suggest one next step: offer_call, send_resource, ask_clarifying_question, refer_out, or decline.
5. Draft a reply that answers the lead's actual message, in a friendly, direct tone, signed by {rep_name}. Include {scheduling_link} only for offer_call.
6. Do not invent pricing, features, timelines, customer names, or results. Do not add the lead to any marketing list.

Return JSON with: website_summary, criteria_scores, total_score, band, suggested_next_step, draft_reply, flags.

Human sign-off points

Guardrails

Expected time saved

Estimate: if your team gets about 60 inbound leads a week and a rep spends about 7 minutes on each to look up the company, decide on fit, and write a reply, while reviewing a scored lead with a draft takes about 2 to 3 minutes, that could save roughly 4 to 5 hours a week. This assumes a clear written ICP and good reply templates. Your results depend on volume and data quality.

How to tell it's working

FAQ

Why limit enrichment to the company's public website?

It keeps the workflow simple, low cost, and respectful of privacy. The company site usually tells you enough to judge basic fit: what they do, who they serve, and where they operate. Pulling personal profiles or broker data brings in consent and accuracy questions that a small team may not want to manage. If you add other data sources later, check their terms and your privacy obligations first.

Can we let the agent reply instantly to high-fit leads?

This recipe keeps a rep in the loop for every send, because a wrong claim, a misread request, or an off tone in the first reply is hard to undo. To respond faster, you can send an instant form confirmation that only says you received the request, and then have the rep send the personal reply. Alerts for high-fit leads help keep that review quick.

Do we need a paid CRM for this?

No. A Google Sheet with the fields listed here works for low to moderate volume and costs nothing beyond your automation and AI usage. HubSpot and Pipedrive both offer entry-level plans if you want pipeline views and email logging. Start with what your team already uses, and move to a CRM when tracking follow-ups in a sheet gets messy.

Want this set up for you?

Build it yourself with the steps above, or tell us what you need. We scope the work with you and send a written quote before anything starts.

Request a quote

Sources

Written by Agents Lookup. Official pages for the tools this recipe uses: