Automating lead qualification and routing means letting your CRM handle the repetitive work (collecting enquiries, completing records, scoring, assigning an owner, scheduling follow-ups) while a person handles the moments where the relationship is actually built: the first real conversation, the proposal, the objection. Done well, automation shortens response times. Done badly, it makes prospects feel they are talking to a machine.
The working rule is short: automate what happens before and after the conversation, never the conversation itself. The rest of this guide shows how to apply it with a mainstream CRM (HubSpot, Brevo, Zoho, Odoo or similar) and, when needed, an orchestration tool such as n8n or Make.
The problem we are solving
In many small businesses, in Tunisia and elsewhere, enquiries come through five or six channels: the website form, Facebook and Instagram messages, WhatsApp, phone calls, email, trade fairs. They end up in notebooks, personal inboxes and screenshots. The usual results:
- enquiries left unanswered for days;
- two salespeople calling the same prospect;
- no visibility on where paying customers come from;
- a manager who cannot say how many enquiries are waiting.
Automation does not fix an organisational problem on its own. It makes a rule the team has already agreed on visible and consistent.
Prerequisites
- One CRM, chosen and actually used, even a simple one. A shared spreadsheet works for a while, then breaks.
- Written pipeline stages: new, contacted, qualified, proposal sent, won, lost.
- Qualification criteria shared by marketing and sales: what makes a lead ready for a call?
- An owner for the configuration, able to change a rule without breaking everything.
- A legal frame: a prospect's contact details are personal data. In Tunisia, processing them falls under Organic Law No. 2004-63; tell people how their data is used, collect only what you need, and check with the INPDP which formalities apply to you. Other jurisdictions, such as the GDPR in Europe, may apply depending on who you target.
Step 1: capture every entry point
Each channel should create or update a CRM record with at least: name, contact method, source, date and original message. Website forms usually connect natively or through a plugin. For harder channels (Meta lead ads, a shared inbox), an n8n, Make or Zapier scenario can bridge the gap.
Watch deduplication. Pick a key, such as email or a phone number normalised to international format (+216…), so one person does not end up with three records.
Step 2: enrich without over-collecting
Long forms get abandoned. Ask for little, then complete the record as the conversation progresses. Automatic enrichment can add factual data: country from the dialling code, landing page, campaign from UTM parameters. Avoid scraping third-party data without a clear legal basis.
A generative AI model can help summarise a free-text message or extract the request ("quote for three workstations", "question about lead times"). Treat the summary as an aid: it can be wrong, and you should not send sensitive data to an external service before checking how that service processes it.
Step 3: qualify with readable rules
Qualification answers two questions: does this prospect fit our target, and are they ready to move forward? There are three approaches, which can be combined.
| Approach | How it works | Strengths | Weaknesses | Best for |
|---|---|---|---|---|
| Manual review | A salesperson reads each enquiry and decides | Human judgement and nuance | Slow once volume grows | A handful of enquiries a day |
| Rule-based scoring | Points for criteria (industry, size, pages viewed, explicit quote request) | Transparent, easy to explain and adjust | Needs regular tuning; rigid | Most small businesses |
| AI-assisted classification | A model suggests a category or priority from the message | Handles varied free text | Can be wrong, harder to explain, cost per call | High volumes of free-text messages |
To start, rule-based scoring is almost always enough. An example for a B2B services company:
- explicit request for a quote or meeting: +30;
- company in a target industry: +20;
- visited the pricing or services page: +10;
- personal email address and empty message: −10.
Above a threshold the team agrees on, the lead moves to "to call". These values are illustrations, not benchmarks: tune them after comparing scores with deals actually won.
Step 4: route to the right person
Routing gives each qualified lead an owner. Common rules:
- by region: capital area, coastal cities, south, international;
- by offer: training, services, products;
- by language: Arabic, French, English;
- round robin among available salespeople;
- by existing account: a known customer always goes back to their usual contact.
Always keep a default rule (a manager who receives anything unmatched) and handle absences. A lead assigned to someone on holiday is a lost lead.
Step 5: keep humans where it counts
This is the core of the method:
- Acknowledgements can be automatic, as long as they are honest: "We have received your request; a member of our sales team will get back to you within one working day." Do not dress an automated message up as a personal one.
- The first real contact is human: a call, a written email or a WhatsApp message from the assigned person. The CRM can suggest a template; a person reviews and adapts it.
- Follow-ups are reminders for the team, not necessarily automated messages to the prospect. A "call back in 3 days" task often works better than a generic email.
- A human can always take over: any automated sequence stops as soon as the prospect replies.
- AI-drafted replies are never sent without review in sales conversations.
Step 6: alert on response times
Set a target first-response time (for example, within one working day) and alert when it is missed: first the owner, then the manager. A simple weekly dashboard shows incoming leads, median first-response time and the breakdown by stage and source.
A complete example
Take a training centre that receives a request through its "corporate training" form:
- A record is created with the source (a LinkedIn campaign, via UTM) and the message.
- Rules add +30 (company request) and +20 (declared team size above 10): the lead moves to "to call".
- Routing assigns it to the person in charge of corporate training.
- The prospect receives a transparent acknowledgement.
- The owner gets a task, calls, and records the need in the CRM.
- If no activity is logged by the end of the day, their manager is alerted.
Nothing in this flow replaces the conversation; everything around it becomes reliable.
Common mistakes
- Automating a process that does not yet exist on paper.
- Stacking automated email sequences until prospects tune out.
- Never revisiting scoring rules.
- Letting salespeople bypass the CRM "because WhatsApp is faster".
- Connecting tools without error logs: a broken scenario can silently drop leads.
What automation will not do
It will not make up for a poorly positioned offer, an understaffed team or a broken promise to call back. Nor does it replace your salespeople's knowledge of the market. Its job is to make sure no enquiry is lost and that everyone has the right information at the right time.