Ship an AI lead qualification workflow with Perplexity, Claude, and Make in a weekend

AI lead qualification workflow Make setups can be shipped in a weekend by chaining Perplexity for research, Claude or GPT‑4/GPT‑5 for scoring, and Make for routing into your CRM with clear next actions.¹⁶¹⁰

What is an AI lead qualification workflow in Make and why does a 3‑stage pipeline work?

An AI lead qualification workflow in Make is a 3‑stage pipeline that enriches, scores, and routes inbound leads automatically into your CRM.¹¹¹

AI lead qualification uses software and models to decide which prospects are worth a salesperson’s time and how fast to engage them, combining enrichment, predictive scoring, behavioural intent, and conversational logic.⁹ A practical workflow for 2025–2026 teams is a three‑step pipeline: capture and enrich lead data, score and prioritise against your ICP and behaviour, and route qualified leads into your CRM with clear next actions.¹¹⁰

In Make, that usually translates to:

• A form or CRM trigger (Typeform, Tally, HubSpot, monday sales CRM).
• A research stage using Perplexity or similar web‑search AI.
• A scoring stage using Claude or OpenAI’s GPT models.
• A routing stage writing into your CRM, Google Sheets, or both.⁴⁶

If you keep each stage explicit and testable, you can ship this without rebuilding your stack or confusing your sales team.

How do you design the 3‑stage AI lead qualification pipeline before touching Make?

You design the pipeline by explicitly defining data intake, research fields, scoring rubric, thresholds, and routing actions before you open Make.⁶¹¹

Start on paper or in a doc rather than in Make:

1. Stage 1 – Capture & enrich
   Decide what your form or intake already captures (name, email, company, website, role, budget) and what you want AI to enrich: industry, headcount, tech stack, recent funding, and relevant news.¹⁵
   Perplexity (or a similar web‑search AI) can research the company website and public sources to append firmographics and context to each lead record automatically.⁵⁶

2. Stage 2 – Score & prioritise
   Build a simple ICP‑based scoring matrix. Assign points for firmographic fit (industry, size, region), role seniority, and behaviour (pages visited, form completeness, email engagement).⁶¹⁰
   Modern teams weight behavioural signals higher because behavioural data is roughly 3× more predictive than firmographics for qualification quality.²

3. Stage 3 – Route & action
   Define score bands and outcomes before any automation: e.g. ≥80 = sales‑ready with immediate routing to a rep, 50–79 = marketing‑qualified with nurture, <50 = long‑term nurture or disqualified.¹⁰¹
   Document exactly what happens in CRM for each band: owner, SLA, sequence, and whether a meeting link is automatically sent.

You want one page that any sales leader can read and challenge. Avoid opaque “clever scores” no one can explain; clarity is more important than sophistication.⁵

Sample 0–100 scoring matrix for service businesses

Here’s a realistic starting matrix for a B2B service firm.

Dimension	Points range	Example rule
Industry fit	0–25	Core verticals 25, adjacent 10, out‑of‑scope 0
Company size	0–15	Target 10–200 staff 15, 5–10 or 200–500 staff 8
Role / seniority	0–20	Founder/VP 20, manager 10, non‑decision maker 0
Behavioural intent	0–25	Pricing page + form 25, blog only 5
Budget / timeline	0–10	Budget declared + <60‑day timeline 10, unknown 4
Disqualifiers	−20	Student, competitor, clear misfit

Leads scoring 80+ are “sales‑ready” and should go straight to a rep; 50–79 enter a structured nurture motion.¹⁰

How should you write and test the AI scoring prompt for Claude or GPT?

You write a 400–600‑word prompt encoding your ICP, disqualifiers, scoring rubric, and output format, then test it against at least ten real past enquiries.⁶

Most of the heavy lifting happens in your system prompt for Claude or GPT:

• Describe your product/service and ideal customer profile in concrete terms.
• Specify firmographic ranges, geographies, and explicit disqualifiers.
• Embed your scoring matrix (dimensions, point ranges, bands).
• Ask for a structured JSON response with: numeric score, band label, reasoning, and recommended next action for sales.⁶

According to specialist guidance, “a well‑written qualification prompt is usually 400–600 words” and should be tested on “ten real past enquiries before going live” to catch false positives and negatives.⁶ Use historical leads you know were good, mediocre, and bad, and see whether the AI’s scores align.

During testing:

• Compare scores against closed‑won vs closed‑lost outcomes, not just gut feel.⁹
• Adjust behavioural weighting if high‑intent leads score too low.
• Keep the rubric comprehensible; sales should be able to explain why a lead scored 82.

Treat this scoring logic as a living asset. Without a feedback loop from actual deal outcomes, AI scoring will decay over time.²⁹

How do you use Perplexity in Stage 1 to enrich leads before scoring?

You use Perplexity as a research step that turns raw form fields into a richer profile: industry, size, tech stack, and recent signals appended to each lead record.⁵⁶

In the Stage 1 Make scenario:

• Trigger: New form submission or new lead in CRM (e.g. monday sales CRM, HubSpot).⁴¹⁰
• Lookup: Take the company name and website from the lead and send them to Perplexity via an HTTP or custom app module.
• Research: Ask Perplexity to summarise key facts: industry, approximate headcount, relevant products, whether they sell online, and any recent events (funding, launch, hiring).⁵⁶
• Write‑back: Append the researched fields into your lead record (e.g. Google Sheets enrichment columns or CRM custom fields).⁶

A 2026 Make + Perplexity workflow example uses Google Sheets as a lead database with two tabs: Raw Leads and Qualified Leads. The scenario searches for unprocessed rows, sends company data to Perplexity, passes the research to GPT‑4 with your criteria, then moves only “Qualified” and “Needs Review” companies into the clean sheet.⁶

The same pattern works with live inbound form data. The important point is that Stage 1 gives your scoring model a richer, more recent context than the form alone.¹¹¹

How do you build the full AI lead qualification workflow in Make in a weekend?

You build three Make scenarios (or one with three segments): intake + research, scoring, and routing, each testable end‑to‑end.⁴⁷

A realistic weekend plan:

Day 1 morning – Intake & data hygiene

• Connect your form (Typeform, Tally, Webflow, etc.) or CRM “new lead” trigger to Make.⁴
• Standardise fields: company name, email, website, role, source. Data hygiene and freshness are critical; stale or messy data will degrade AI scoring.²¹⁰
• Add basic validation (email deliverability, phone format) using existing Make modules or third‑party APIs.³⁵

Day 1 afternoon – Perplexity research scenario

• Add a Perplexity module (HTTP or app) to call web‑search AI with company details.⁶
• Map the response into enrichment fields: industry, size band, online/offline, notable signals.⁵⁶
• Save the enriched lead into a staging sheet or CRM record.

Day 2 morning – Claude / GPT scoring scenario

• Connect Claude or OpenAI via Make’s AI modules.⁴⁶
• Paste your tested 400–600‑word scoring prompt and pass in:
  • Original form fields
  • Perplexity research summary
  • Any behavioural data (pages viewed, UTM, email opens) if available.
• Ask for: numeric score (0–100), band label, reasoning, and suggested next action.
• Log everything to a Google Sheet “Qualified Leads” tab so you can eyeball results.⁶⁷

Day 2 afternoon – Routing into CRM and next actions

• Add a router step in Make based on the score and band.¹¹⁰
• For sales‑ready (≥80):
  • Create/Update lead in CRM with status = “Sales Ready”.
  • Assign owner, add tasks, and optionally trigger a Slack alert.⁴⁵
• For marketing‑qualified (50–79):
  • Tag “MQL – AI” in CRM and enrol in a nurture sequence.¹⁰
• For nurture / disqualified (<50):
  • Tag appropriately and keep them out of high‑touch flows.

Make is built for exactly this orchestration: from a form or CRM trigger, call AI for research and scoring, then push structured outcomes into Sheets or CRM with tags or status fields such as Hot/Warm/Cold.⁴⁸

What CRM fields and actions should you configure for this workflow?

You configure explicit fields for score, band, and AI notes, then attach automation rules to each band for follow‑up.¹⁰

In monday sales CRM or similar systems, your AI lead qualification workflow should write to:

• AI Score (0–100) – numeric field.
• AI Band – dropdown (Sales‑Ready, MQL, Nurture, Disqualified).
• AI Notes – short reasoning (visible to reps).¹⁰
• Source – to track which funnel entry produced the lead.

Routing rules:

• Sales‑Ready: Auto‑assign to an SDR/AE queue, create a task, and trigger a meeting link or personalised intro email.¹⁰¹¹
• MQL: Route into marketing automation with content tuned to their segment.
• Nurture: Add to low‑frequency campaigns, avoid immediate sales contact.

Common practice is to define thresholds like ≥80 = sales‑ready, 50–79 = marketing‑qualified, <50 = nurture and treat those as explicit CRM states.¹⁰¹

How do you test and ship this AI lead qualification workflow safely?

You run structured dry‑runs with 10–15 hypothetical leads, validate routing for each band, and check mobile form completion rates before declaring it live.⁸

Before production:

• Form UX: Test your intake form on desktop and mobile. Typical best practice is to track completion rate and aim for above 70%; if people abandon, you probably ask too many questions.⁸
• Scenario sanity checks: Run 10–15 hypothetical leads through the full flow, including edge cases and obvious misfits.⁸
• Routing validation: Confirm tags, owners, emails, and booking links for at least five profiles across all paths (Sales‑Ready, MQL, Nurture, Disqualified).⁸
• Sync cadence: Run CRM sync every 15 minutes rather than daily to keep data fresh for AI scoring.²

Once live, add a monthly feedback loop: pull closed‑won, stalled, and rescued deals, compare their AI scores, and adjust the rubric and thresholds based on real revenue outcomes.⁹² That’s the difference between a one‑off experiment and a dependable system.

How does this Make‑based stack compare to “rebuild your funnel” AI platforms?

A focused Perplexity + Claude/GPT + Make setup gives you most of the value of dedicated AI lead platforms without replacing your CRM.

Approach	Pros	Cons
Make + Perplexity + Claude/GPT	Uses existing forms & CRM; flexible; fast to ship in a weekend	Requires thoughtful prompt & scenario design
All‑in‑one AI lead platforms	Prebuilt UX, analytics, chat widgets	Often force new forms/CRM; heavier rollout
Manual scoring in CRM	Simple, no new tools	Slow, subjective, hard to scale

For teams aiming at fast, credible sales wins this quarter without a stack rebuild, the Make‑based workflow is the pragmatic choice: it adds intelligence where you already work and leaves your CRM as the system of record.¹¹⁰