Contents
    Funnel of automated lead generation: Visitor, Lead, Qualified, Meeting
    From visitor to qualified meeting.

    Automated lead generation connects your website, CRM and AI so that a first signal turns into a qualified meeting with no manual effort: contacts are enriched, evaluated by fit and behaviour (lead scoring), routed to the right person or moved into nurturing. The second, often larger source of revenue is already in your CRM: existing customers who have gone dormant and will buy again with the right offer at the right time.

    This guide describes both systems the way we build them for Swiss B2B SMEs: from the ideal customer profile and a scoring model you can copy to the reactivation agent we built for a client. It also covers the questions that need answering first in Switzerland: which CRM? What do the revDSG and the UWG allow? And where does the human stay in the loop? Both systems are part of a sales funnel that connects marketing and sales from the first click to the signed order.

    What is automated lead generation?

    Instead of someone manually reviewing, researching and distributing new contacts, a system handles the routine. It recognises a relevant contact, enriches it with company data, assesses its readiness to buy and triggers the right next step: meeting, nurturing or archive.

    The distinction matters: automated lead generation does not mean generating more contacts or sending thousands of cold emails. It means making the path from the first signal to the conversation fast, traceable and measurable. The signals themselves come from channels such as SEO, visibility in AI answers, LinkedIn, trade fairs or ads in ChatGPT and Google. The system makes sure none of those signals get lost.

    What does the path from first signal to meeting look like?

    Six building blocks, in this order:

    Building blockPurposeTypical tool
    Ideal customer profile (ICP)defines who even counts as a good leadanalysis of your best existing customers
    Captureforms, meeting booking, downloads and chat land cleanly in the CRMwebsite, form tool, CRM
    Enrichmentadds company size, industry, location, rolecommercial register, website, AI research
    Lead scoringprioritises by fit and behaviourrules in the CRM
    Routingassigns, books meetings, notifiesCRM workflows
    Nurturingkeeps not-yet-ready contacts warmemail sequences with consent

    The ideal customer profile does not come from gut feeling in a workshop, but from data: take your ten to twenty best customers (by margin, not revenue) and note what they have in common. Industry, company size, region, who made the decision, what triggered the purchase. The opposite matters just as much: which customers brought a lot of effort and little return?

    Enrichment adds what the form does not ask. Industry, location and size can be derived from the company domain, legal form and registered office from the commercial register (zefix.ch). AI agents also summarise the company's website and note what is relevant for the first conversation. That keeps forms short: name, business email, company, request.

    How do I build a lead scoring model?

    A scoring model assigns points for fit (does the contact match the ICP?) and behaviour (how close are they to a buying decision?) and deducts points when something speaks against it. The model below is a starting point you can copy and adapt to your ICP. The example is an IT service provider whose core customers are fiduciary and law firms in German-speaking Switzerland.

    CategoryCriterionPoints
    FitIndustry in core market (fiduciary, law firm)+15
    FitCompany size 20 to 250 employees+10
    FitCompany size 10 to 19 employees+5
    FitLocation in service area (German-speaking Switzerland)+5
    FitRole: management, partner, head of department+15
    FitRole: specialist or IT lead+8
    FitBusiness email domain+5
    BehaviourContact form or meeting request+30
    BehaviourAttended a webinar or event+10
    BehaviourVisited pricing page+10
    BehaviourCase study or service page several times within 14 days+5
    BehaviourDownloaded a checklist or guide+5
    BehaviourNewsletter click (counted at most three times)+2
    DeductionFree email address (gmail, bluewin etc.)−5
    Deduction30 days without activity (score decay)−10
    ExclusionJob application, students, competitors, suppliersset to 0

    Thresholds: from 70 points the lead goes straight to sales, from 40 to 69 points into nurturing, below 40 points it is watched. One rule overrides all points: anyone who actively asks for a conversation always gets a person, regardless of score.

    A worked example: the head of finance at a fiduciary firm with 35 employees in Winterthur downloads a checklist and visits the pricing page. Fit: 15 + 10 + 5 + 15 + 5 = 50 points. Behaviour: 5 + 10 = 15 points. Total 65, so nurturing. If she signs up for the webinar two weeks later, the score rises to 75 and the responsible salesperson gets a notification with a summary of every touchpoint.

    The first model is a hypothesis. After a few months, compare the score bands with actual deals: if high-scoring leads do not win more often than low-scoring ones, the weights are off.

    Routing, speed-to-lead and nurturing

    Routing means the right lead lands with the right person, with context. Rules by region, industry or product line are quick to set up in the CRM. What matters is what the person receives: company profile, score with reasons, pages visited, previous contacts and a suggested opening for the conversation.

    Speed-to-lead is the time between an enquiry and the first genuine response from a person. In many SMEs it is the biggest weak spot, because enquiries sit in a shared inbox. Automation helps in three ways: instant confirmation with a link to book a meeting, notifying the responsible person on the channel they actually read, and an escalation if nobody has responded within a set deadline.

    // Enrich, score, route: sales only sees what matters.
    async function onNewLead(lead: InboundLead) {
      const enriched = await enrich(lead);            // company size, industry, role
      const { score, reasons } = scoreLead(enriched); // fit + behaviour - deductions
    
      if (lead.requestedMeeting || score >= 70) {
        await notifyOwner(enriched, reasons);         // with summary
        return escalateIfNoResponse(enriched, "4h");  // deadline per your own rule
      }
      if (score >= 40 && lead.hasConsent) return addToNurture(enriched);
      return watch(enriched);                         // not (yet) a fit
    }
    

    Nurturing keeps contacts warm who fit but are not ready to buy. Good nurturing sequences are short and useful: a case study from the same industry, a checklist, an invitation to an event. They require consent (more on that below) and end as soon as the contact responds or the score crosses the threshold.

    The second revenue source: reactivating existing customers

    Winning new customers takes visibility, trust and time. Existing customers already know you, have had good experiences, and their purchase history sits in your data. Even so, many SMEs barely look after them systematically: anyone whose last order was months ago slips through the cracks. Not out of disinterest, but because no one has the time to go through hundreds of customers one by one.

    Detecting dormant customers

    A fixed deadline such as "twelve months without a purchase" rarely fits. A customer who orders consumables monthly is a warning sign after three quiet months; a customer who replaces equipment every two years is not. Rules based on each customer's own rhythm work better:

    1. Calculate the reorder cycle: the typical gap between two orders per customer, or per product category if there are few orders.
    2. Mark as dormant: when clearly more time has passed since the last purchase than that cycle, with one and a half times the cycle as a starting value.
    3. Add early signals: shrinking order volumes, expiring maintenance contracts or licences, a change of contact person, declining usage.
    4. Respect suppressions: open complaint, overdue invoices, recent contact, unsubscribe.

    Matching upsell and cross-sell potential

    For every dormant customer the question is: which offer fits now? Three sources provide the answer. First, your own purchase patterns: which products do similar customers typically buy together or one after another? Second, fixed rules from sales, such as "anyone who has used module A for a year is a candidate for module B". Third, the lifecycle: replacement, expansion, maintenance. The result is a score made of potential (what is the deal worth?) and likelihood (how well does the offer fit the history?).

    A project from practice: the reactivation agent

    For a client with a large customer base, we built a reactivation agent. It runs continuously in the background, identifies dormant customers from their purchase cycles, assesses their up- and cross-selling potential and drafts a personal message for each one that refers to their actual purchase history. All that reaches the sales team is a prioritised list: customer, reason for the outreach, suggested offer, finished draft.

    // Who is ready for which offer? The agent prepares, the human approves.
    async function reactivate(customer: Customer) {
      if (!isDormant(customer) || isSuppressed(customer)) return;
    
      const signals = analyzeHistory(customer);              // purchase cycles, products, behaviour
      const offer   = await matchUpsell(signals);            // fitting up-/cross-sell
      const draft   = await draftMessage(customer, offer);   // personal draft
    
      return queueForApproval({ customer, offer, draft, reason: signals.why, priority: offer.score });
    }
    
    StepWho handles itResult
    Detect inactivityAgentfiltered, relevant customers
    Assess potentialAgentscore for up- and cross-selling
    Match the offerAgentoffer that fits the history
    Write the draftAgentpersonal draft message with reasoning
    Review, edit, approveHumanfinal text, timing, relationship
    Conversation and closeHumanthe order

    The approval workflow

    The agent sends nothing on its own. Every draft lands in an approval queue in the CRM, and the responsible person has three options: approve, edit, or reject with a reason. The rejection reasons are valuable, because they show where rules or offer matching need sharpening. Fixed limits have proven useful, such as at most one outreach per customer per quarter, no message while a complaint is open, no repeat after a no.

    The result

    • Additional revenue from existing customers that used to be left on the table, with no new acquisition budget and no extra sales hours.
    • Sales only works on warm, prioritised contacts instead of sifting through cold lists.
    • Personal at scale: every outreach is based on real purchase history, not a mass newsletter.

    Which CRM suits a Swiss B2B SME?

    The CRM is the central data foundation for both systems. In Swiss SMEs we mainly see four setups:

    SystemTypical useRole in the setup
    HubSpotmarketing and sales in one systemforms, scoring, email, workflows, pipeline
    Pipedrivesmall to mid-sized sales teams, pipeline-focusedpipeline and activities, automation via workflows and integrations
    Salesforcelarger organisations, complex processes, several teamscomprehensive platform with wide room for customisation
    Bexioaccounting and invoicingsource of purchase history, payment status and reorder cycles

    The choice depends less on features than on who maintains the system every day. A simple CRM that sales actually uses beats a powerful platform full of outdated records. For reactivation, the link to accounting is decisive: invoice data in Bexio or another ERP is the most reliable source of who bought what and when. Pulling that data into the CRM via an integration is usually the first technical step. How to approach such connections in general is covered in our article on process automation for Swiss SMEs.

    Where does AI help and where does the human stay?

    AI is strong at volume and preparation, people at judgement and relationships. The division of labour is the same as with all AI agents in the company:

    AI handlesPeople handle
    researching and summarising companiesdeciding whether a contact really fits
    calculating and explaining the scoresetting and calibrating scoring rules
    classifying and assigning enquiriesrunning the first conversation
    analysing purchase history, finding dormant customersjudging relationship and timing
    writing draft messagesreviewing, editing, approving drafts
    reminding of deadlineshandling objections, negotiating, closing

    What do the revDSG and the UWG allow?

    Two laws determine what is permitted in Switzerland. This is a practitioner's view, not legal advice.

    UWG: marketing email needs consent. Art. 3(1)(o) of the Federal Act against Unfair Competition lists as unfair conduct to "send or arrange to be sent mass advertising without direct connection with any requested content by telecommunication and in doing so fail to obtain the prior consent of the customer, or to indicate the correct sender or a simple and free of charge option of refusal". The exception follows in the same letter: "any person who receives contact details of a customer when selling goods, works or services, and who indicates the option of refusal when doing so, does not act unfairly if they send that same customer mass advertising for their own similar goods, works or services without the customer's consent". The rule protects not only private individuals but also companies as customers.

    In practice this means:

    • Cold email sequences to bought, scraped or business-card-sourced addresses are not lead generation, they are a violation. An AI that rephrases every email slightly does not change that: an automated series is still a series.
    • Inbound comes before outbound. Anyone who sends an enquiry or requests a guide has created a direct connection. For nurturing sequences beyond that, ask for consent actively at the download.
    • Existing customers may be emailed about your own similar offers if you point out the opt-out when collecting the address and in every message. Clearly different products or partner offers need consent. The reactivation agent checks this as a suppression rule.
    • Phone: an asterisk notice in the telephone directory must be respected under letter u, numbers without a directory entry are treated the same way, and under letter v marketing calls need a displayed, registered number.

    How to implement consent and unsubscribes cleanly in a newsletter tool is shown using the example of Mailchimp and the revDSG.

    revDSG: transparent, purpose-bound, secure. Under the Data Protection Act, lead scoring is profiling, meaning an automated evaluation of personal aspects. That is permitted, but requires transparency: the privacy policy mentions enrichment and scoring, under Art. 19 DSG also when data is not collected from the person themselves. Further points are data processing agreements with CRM and AI providers, reviewing transfers abroad for providers outside Switzerland and a deletion rule for contacts who never become customers. Because a person approves every message in the reactivation agent, there is no decision based solely on automated processing within the meaning of Art. 21 DSG. What else to consider when using AI models with customer data is covered in our article on AI and data protection in Switzerland.

    Get started in 5 steps

    1. Clarify ICP and data foundation: analyse your best customers, export invoice data from Bexio or your ERP, clean up duplicates.
    2. Set up the CRM as the single source: all forms, meeting bookings and downloads go straight into the CRM, with a consent field and source per contact.
    3. Define scoring and routing: adapt the model above with sales, set thresholds, configure notification and an escalation deadline.
    4. Set up nurturing: two to three short sequences per main segment, only for contacts with consent.
    5. Start reactivation as a pilot: one customer segment, a clear dormancy rule, an approval queue; after a few weeks review the rejection reasons and expand.

    How do I measure success?

    The number of leads is the weakest metric, because it can be inflated at will with poor contacts. These are meaningful:

    MetricWhat it shows
    Speed-to-lead (median)how quickly a person responds to enquiries
    Lead to meetingwhether the right contacts are coming in
    Meeting to proposalwhether qualification and first conversation work
    Close rate by sourcewhich channels actually bring revenue
    Close rate by score bandwhether the scoring model is right
    Pipeline value and cycle lengthhow sales is developing overall
    Approval rate of draftshow accurate the reactivation agent is
    Reply rate and reactivated customerswhether the outreach lands
    Revenue from reactivationthe contribution of existing customers
    Unsubscribeswhether frequency and relevance are right

    Common mistakes

    • Tool before process: a new CRM does not fix an unclear ICP or missing responsibilities.
    • Scoring without sales: if marketing assigns the points alone, sales will not trust the scores and will ignore them.
    • No score decay: a download from a year ago is no longer a buying signal.
    • Bought lists and email sequences: not permitted and harmful to your own domain's deliverability.
    • Fully automated sending to existing customers: one ill-fitting message to a good customer costs more trust than ten fitting ones build.
    • No suppression rules: offering an upgrade to a customer with an open complaint shows that nobody looked.
    • Forgetting existing customers: the entire budget goes into new customers while the purchase history sits unused in accounting.

    Conclusion

    Automated lead generation and reactivation follow the same principle: the system screens, scores and prepares, the person decides and runs the conversation. Build both properly and sales only works on contacts that fit, with no revenue left sitting in your own customer database.

    As a full-funnel marketing agency, we build these systems together with the Google Ads and performance marketing that fill the pipeline and the AI automation that works it. Marketing and engineering sit in the same team. Book an intro call.

    Tenzin Langdun

    About the author

    Co-Founder of Hierarchy · SEO, GEO & AI Expert

    Tenzin Langdun is a co-founder of Hierarchy, where he leads SEO and GEO: how visible companies are in Google, in Google's AI answers and in assistants such as ChatGPT, Gemini and Claude. He also leads the agency's own data studies, including the Zurich AI Visibility Study of 1,515 AI answers about 5,110 companies. He brings over 10 years of marketing experience across strategy, paid acquisition and SEO, and has worked at KPMG, EY, Siemens, Adnovum and LGT, in AI, cybersecurity, audit and consulting, and in software engineering for the insurance sector and for banking and finance. He holds a BSc from ZHAW and an MSc in Artificial Intelligence from the University of Bath. His research on AI-based thyroid cancer detection was published in Scientific Reports (Nature Portfolio) in 2024 and won the Siemens Excellence Award 2024 and the ZHAW Lab Sciences Award 2024.