Lead enrichment, scoring, routing and CRM hygiene

0:00 Lead pipeline

Here's a pattern you'll see in almost every business. A lead fills in a form on Friday evening, nobody sees it until Monday, and by then they've booked a call with a competitor. Two things decide whether inbound leads convert: how fast you respond, and how well you understand them. Automation can fix both, if it also keeps your CRM clean instead of filling it with duplicates. In this lesson you'll build the full lead pipeline: capture, normalize, dedupe, enrich, score, route and notify, plus hygiene automations that keep your CRM trustworthy.

0:40 Lesson roadmap

Here's the roadmap. First, why speed and understanding decide conversions. Then a triage analogy for the whole pipeline. Then enrichment, including compliance, then scoring and routing. Then two examples: CRM hygiene automations, and a full lead pipeline for a B2B firm in Dubai. We'll finish with common mistakes and a scoring exercise you can run on your own leads.

1:06 Why it matters

Why does this matter? Speed to lead is one of the most consistent factors in inbound conversion, and it's exactly what automation does best. But speed without context wastes the call. A rep who knows what the company sells, its size, and that the lead mentioned a budget and a deadline has a much better first conversation. And a CRM full of duplicates and stale data makes every report and every campaign worse.

1:38 Analogy: hospital triage

An analogy. Think of a hospital emergency department. Patients arrive through different doors. At triage, a nurse records details in a standard format, checks whether they're already a patient, looks up history, scores urgency with a consistent protocol, and sends them to the right specialist, fast. Nobody guesses. That's your lead pipeline: standard intake, dedupe, enrichment, explainable scoring and routing.

2:04 Enrichment

Enrichment adds context. Data providers offer company size, industry, location and technology, but check their coverage in your markets, terms and data protection posture. An AI website summary is a cheap alternative: fetch the company homepage, respecting robots rules and terms, extract the text, and ask a model for a structured summary of what they sell, who they sell to, and likely needs. It works well for small businesses in Pakistan and the Gulf, where data vendors often have thin coverage. Add first-party signals like pages visited and campaign source. And remember compliance: B2B enrichment still processes personal data, needs a lawful basis and transparency, and scraping against platform terms is off-limits.

2:53 Scoring

Scoring combines fit and intent. Fit asks: does this lead match your ideal customer profile, by industry and size? Intent asks: are they ready, did they mention a budget, request a call, or visit the pricing page? Here's the key principle. Let AI extract signals like budget and urgency from free text, but keep the score calculation deterministic and explainable. The lesson text has a scoring function that adds points for target industry, size, budget and a call request, subtracts for free email domains on a B2B offer, assigns a band, and records the reasons, so reps can see why a lead is an A.

3:39 Routing

Routing sends each lead to the right person. Use rules by territory, like country, emirate or city, by language, by service line, by existing account ownership, round-robin with capacity limits, and working hours. And always have a default owner, so nothing falls through the cracks. Then notify and acknowledge: a Slack alert with the summary for the rep, and a prompt acknowledgment to the lead in their language, promising a realistic response time.

4:11 Example 1: CRM hygiene

Example one, simple: CRM hygiene. Upsert instead of create, matching on email, phone and company domain. Standardize country codes, phone formats and lifecycle stages whenever you write. Run a weekly workflow that flags likely duplicates for merge review, without auto-merging. Alert owners when deals have no activity for fourteen days. Turn missing required fields into tasks. And have every automation write a note saying which workflow and version updated the record.

4:42 Example 2: Dubai B2B firm

Example two, realistic. A Dubai B2B services firm gets leads from web forms, LinkedIn lead forms and events. The pipeline normalizes and deduplicates on email and domain, enriches with a data provider where coverage exists, otherwise with an AI website summary, extracts budget and urgency with structured output, and computes a rule-based score. Saudi leads go to the Riyadh team, UAE leads by service line, and existing accounts to their owners. Reps get a Slack summary, leads get an acknowledgment in English or Arabic, and high scorers get a task due within four working hours. A weekly hygiene job flags duplicates and stale deals. High-score leads now hear back the same hour instead of the next day.

5:33 Common mistakes

Three mistakes to avoid. Letting AI score leads with no explainable rules, so nobody trusts or can audit the scores. Creating duplicates because you match only on email, when the same company comes in through a different person or a phone number. And enriching with data you don't have the right to use, or scraping platforms against their terms. Each one erodes trust in the CRM, or creates legal risk.

6:03 Watch me do it: enrich, score, clean

Watch me do it. I add enrichment and scoring to the Make build. After the AI extraction, I check whether the lead has a company website. If yes, an HTTP module fetches the homepage, a text parser strips the HTML, and I keep only the first few thousand characters. An AI module summarizes it into JSON: what they sell, likely customer type and industry from our list. I parse it. Then scoring: I don't let AI score. I use a set variable module with an if formula that adds points for target industries, budget above our threshold, a call request and urgency, subtracts for free email domains on B2B services, and builds a reasons text. The score and band go into the CRM along with the reasons. Next, hygiene. I change the CRM search to match on email or on company domain, so a second person from the same company attaches to the existing company rather than creating a duplicate. I add a separate weekly scenario that lists contacts with the same domain and similar names and posts them to a sheet for merge review, never auto-merging. And a stale-deal check: deals with no activity in fourteen days notify the owner. Finally, I score last month's fifty leads with the same formula and compare conversion by band. Band A converted best, so the weights hold.

7:41 Recap + try this now

Recap. Build the pipeline: capture, normalize, dedupe, enrich, score, route, notify, and log. Use AI to extract signals and summarize websites, but keep scoring rule-based and explainable. Route with clear rules and a default owner. Automate CRM hygiene with upserts, standardization, duplicate review and stale-deal alerts. Try this now: take the scoring function in the lesson text, adapt the target industries and ranges to your business, run it on your last fifty leads, and check whether A-band leads actually converted better.

8:16 Try this now

Try this now. Export your last fifty leads with their outcomes. Adapt the scoring function in the lesson text: set your target industries, size range and budget threshold. Run it over the export and add the score, band and reasons. Compare the conversion rate by band. If B-band leads convert as well as A-band, your weights are wrong, so adjust and re-run. Then put the final function into your lead workflow.