Skip to content

Use case · Sales

Qualify every lead in the first reply, then route it by your rules.

Leads from forms, chat and WhatsApp get an answer straight away with the two or three questions your best rep would ask. A score built from rules your sales team writes decides where each one goes, and the CRM gets a clean record with a summary, not a pasted email.

At a glance

Connects to
  • Lead sources
  • Your CRM, via its API
  • Owner calendars
  • Team alerts
  • Workflow runner
  • Model provider on your account
Typical rollout
Week 1 → Weeks 5–6See the plan

The problem

Where the work gets stuck today

Leads arrive at all hours through the website, chat, WhatsApp and the shared inbox. The first reply waits until someone reads the message, looks up the company and decides who owns it. By then the lead has often heard back from someone else.

When qualification does happen it is uneven. Each rep asks different questions, required CRM fields stay empty, and the routing rules live in one person’s head, so leads bounce between owners or sit unassigned over a weekend.

  • Enquiries sent on a Friday evening get a first reply on Monday.
  • The CRM holds a pasted email instead of company, need, timeframe and role.
  • Two reps call the same lead because nobody saw the duplicate.
  • Small or off-profile enquiries take as much rep time as the best ones.

Task catalog

What the system takes on

  • Task 01

    Reply at once, on the same channel

    Every new lead gets a reply with two or three questions taken from your discovery script, adapted to what they already said.

  • Task 02

    Turn free text into fields

    Company, size band, need, timeframe, budget band and role are extracted from the conversation. Anything the lead did not say is marked “not given”, never guessed.

  • Task 03

    Enrich from public company facts

    Industry and size come from the company’s own website and public listings. No personal profiles are scraped and no lists are bought.

  • Task 04

    Score by your written rules

    Points per criterion come from a rule table your sales lead owns. The total is explainable line by line, so a rep can see why a lead scored the way it did.

  • Task 05

    Route to the right owner

    Territory, segment, product line or round-robin rules pick the owner. Sales-ready leads get a booking link to that owner’s calendar in the reply.

  • Task 06

    Write the CRM record once

    Contact, company, deal and a summary note are written through the CRM’s API after a duplicate check, with the rule that routed the lead attached.

  • Task 07

    Catch what is not a lead

    Job applications, supplier pitches, press and partnership requests go to named inboxes instead of the sales queue.

Step trace

One run, step by step

Sample workflow: a new lead is deduplicated and enriched, asked two qualifying questions, scored by written rules, then routed, with the CRM record written and the owner alerted.
Sample data
Sample data
Lead inForm · chat · WhatsAppDone
Dedupe + enrichMatch on email and domainDone
QualifyTwo questions, first replyDone
ScoreYour ICP rules, in codeDone
RouteOwner by territory and fitRunning
CRM recordContact · deal · summaryDone
Owner alertSummary + booking linkNeeds approval
A new lead from a form, chat or WhatsApp is checked for duplicates and enriched with public company facts, asked two qualifying questions in the first reply, scored by your rules, then routed: the CRM record is written and the owner is alerted with a summary and a booking link.
Read the trace as text (Sample data)
  1. 00:00Website formStepNew enquiry: dental group, online booking for several locations.
  2. 00:01DedupeDoneNo existing contact for this email or company domain. New record allowed.
  3. 00:01EnrichDoneCompany website read: clinic group, locations listed on the site. No personal data collected.
  4. 00:02AssistantStepFirst reply sent with two questions from your discovery script: timeframe and who signs off.
  5. 00:09LeadStepReplies: wants it live this quarter, is the operations manager.
  6. 00:09ExtractorDoneFields filled: industry, size band, need, timeframe, role. Budget left as “not given”.
  7. 00:10Scoring rulesDoneRules add the points per criterion; total above the sales-ready line.
  8. 00:10RouterApprovalTerritory and segment rule fired: senior owner. Booking link added to the reply.
  9. 00:11CRMDoneContact, company and deal written once, with the summary note and the rule that routed it.
  10. 00:11Owner alertDoneOwner notified with the summary; the first call is theirs to confirm.

Rules, not a black box

Score three sample leads with the same rules

The model reads the conversation and fills in the fields. The points come from a rule table your sales team writes and can change, so every score can be explained one line at a time.
lead-scoring · rule tableSample data
Choose a sample lead

Dental group with several locations wants online booking live this quarter. Operations manager.

Fields the model extracted

Industry
Healthcare, multi-site
Need
Appointment booking
Timeframe
This quarter
Role
Operations manager
Budget
Not given

Points from your rules

  • Fit30 / 30

    Industry and size inside your ideal customer profile

  • Need25 / 25

    Stated problem matches something you sell

  • Timing20 / 20

    Wants to start within the quarter

  • Role10 / 15

    Decision-maker or budget owner

  • Engaged10 / 10

    Answered the qualifying questions

Score

95 / 100

Sales-ready

Rule that fired

Score of 70 or more, region in your territories → senior owner

What happens next: Booking link to the senior owner’s calendar in the reply; deal created; owner alerted with the summary.

Change a weight and you can re-score past leads to see the effect before the change goes live.

Systems involved

What it connects to

Your existing tools stay the system of record; the AI reads and writes through their APIs.
Lead sources
Website forms, the chat widget, the WhatsApp Business Platform and the shared sales inbox, all feeding one intake step.
Your CRM, via its API
System of record for contacts, companies, deals and owners. Written once per lead, after the duplicate check.
Owner calendars
Booking links for sales-ready leads, so the first call lands in the right person’s diary.
Team alerts
Email or your team chat, via its API, with the summary and a link to the CRM record.
Workflow runner
n8n, Make, Zapier or custom code, chosen for your volume and where you want it hosted.
Model provider on your account
OpenAI, Anthropic Claude or Google Gemini for field extraction, replies and summaries. Never for the score itself.

Guardrails

  • The score is computed in code from extracted fields and your rule table. The model extracts and drafts; it does not decide who is sales-ready.
  • Enrichment uses company-level public information only. No personal profiles are scraped and no purchased lists are loaded.
  • Every routing decision is logged with the rule that fired, so a rep can see why a lead landed with them and challenge the rule.
  • A duplicate check runs before any CRM write, and writes are idempotent, so a retried step never creates a second contact.
  • Follow-up messages go only on the channel the lead used, within that channel’s rules and the lead’s consent, with an opt-out in every message.
  • Requests outside the rules, such as partnerships, press or job applications, go to a named inbox rather than the sales queue.

Human checkpoint

Sales owns the rules. A rep reads every sales-ready lead before the first call, and every lead marked “not a fit” appears in a weekly digest the sales lead skims to catch rules that are too strict. Changes to weights and thresholds go through a short change log, so a shift in pipeline can be traced back to a rule edit rather than guessed at.

What we plan for

Typical targets

Typical targets are planning ranges we use when scoping, each with its basis. They are not results from a client project; your own baseline decides the real number.
  • Typical target

    Under 2 minutes, any hour

    Time to first reply on a new lead

    Basis: Triggered by the form, portal or channel webhook, not by someone reading the inbox; the margin covers portal leads that arrive by email. Measured from arrival to first message on test leads from every source.

  • Typical target

    90–100% of fields the lead answered

    of new CRM records with the required fields filled

    Basis: The required-field schema is validated before the CRM write; anything the lead did not say is stored as “not given” instead of being guessed.

  • Typical target

    None in the test set

    Leads routed to the wrong owner in the rule tests

    Basis: Routing rules are unit-tested with past leads from each segment and territory before launch; any misroute after launch is logged and fixed in the rule table.

Rollout

Week by week

Typical ranges for a first release; the scoping call fixes the plan for your systems.
  1. Week 1

    Ideal customer profile, scoring criteria and routing rules written down with sales; a sample of past leads exported; CRM field list agreed.

  2. Week 2

    Intake on the first channel, field extraction, duplicate check and CRM writes against a sandbox.

  3. Week 3

    Rule table in code with tests, owner alerts, booking links for sales-ready leads, courtesy replies for the rest.

  4. Week 4

    Shadow mode: the system scores and routes alongside your team, and the differences are reviewed together.

  5. Weeks 5–6

    Go live channel by channel; the weekly “not a fit” digest and the rule change log start.

FAQ

Frequently asked questions

Who decides what counts as a qualified lead?

Your sales team. We write the ideal customer profile, the criteria and the points per criterion together in week one, and they live in a rule table you can edit. The AI fills in the fields from the conversation; the rules decide the score and the route.

What will the assistant ask a new lead?

Two or three questions taken from your own discovery script, usually timeframe, the problem in their words and who signs off. It skips anything the lead already answered in the form or message, and it never pushes for budget if your team prefers to raise that on the call.

Can sales-ready leads book a call straight away?

Yes. Leads above your sales-ready line get a booking link to the routed owner’s calendar in the same reply, with the owner’s real availability. The owner still sees the summary first and can move or decline the slot.

Is it acceptable to enrich leads with outside data?

We limit enrichment to company-level public information such as the company’s own website and public business listings. We do not scrape personal social profiles or load bought lists, and your privacy notice should say that enquiries are enriched this way.

What happens to leads that score low?

Nothing is thrown away. They get a courtesy reply with useful resources, the contact is tagged in the CRM, and they appear in a weekly digest. If the digest keeps showing good leads marked “not a fit”, that is the signal to loosen a rule.

Can it tell a returning lead from a new one?

It matches on email, company domain and phone number before writing anything. A returning lead is attached to the existing contact and stays with its current owner, and the owner is told the lead came back and through which channel.

Can we test the scoring without changing how our team works?

Yes, that is the shadow-mode week. The system scores and routes every new lead in parallel while your team works as usual, and we compare its decisions with theirs before anything is switched on.

Next step

Have a workflow in mind? Let’s scope it together.

Book a free 30-minute strategy call. Bring one process you want to automate and we will talk through scope, integrations and a sensible first step.

Free 30-minute call · no commitment

Send an RFP or security questionnaire