Step-by-step guide
How to automate my marketing: a practical method for small teams
This guide is for owners and marketing leads at small and mid-size businesses who still copy leads between tools, send follow-ups by hand and rebuild the same report every Monday. It shows what to automate first, with which kind of tool, and how to stay within GDPR while you do it.
- Updated
- Steps
- 7
- Written by
- BlackCard
The short answer
To automate your marketing, start with one workflow, not a platform. List every repetitive task and score it by frequency, time, clarity of rules and risk. Map where the data lives. Pick the highest-scoring workflow, usually lead capture to CRM to follow-up. Use the automation built into your email or CRM tool first, and add a connector such as n8n, Make or Zapier only when data must move between tools. Keep a human approval step for anything a customer will read, measure hours saved and errors, then give the workflow an owner.
Start with step 1Step 1
List and score your repetitive tasks
Log everything you do by hand for two weeks, then score each task on four questions.
Automation pays off on work that is frequent, rule-based and dull. Before you look at any tool, keep a simple log for two weeks. Every time you or a colleague does a marketing task by hand, add a line: what triggered it, what you did, how long it took and which tools you touched.
Then score each task from 1 to 5 on the four questions below and add the numbers up. The scores are rough on purpose. Their job is to stop you from automating the task that annoys you most instead of the one that costs the most.
| Question | Score 1 | Score 5 |
|---|---|---|
| How often does it happen? | A few times a year | Several times a day |
| How long does one run take? | Under two minutes | More than half an hour |
| How clear are the rules? | Needs judgement every time | Same steps every time |
| How safe is a mistake? | A wrong message reaches a customer | Easy to spot and undo internally |
Step 2
Map where the data lives
Know which tool holds which fact before you connect anything.
Most failed automations break on data, not on logic. A workflow that reads the wrong field, or creates a second copy of a contact, makes more work than it saves. For the top three tasks on your list, draw the path the data takes: where it enters, where it is stored and who needs to see it next.
Decide on one system of record for contacts, normally the CRM or the email platform, and treat everything else as a source that feeds it. Agree on the fields every contact must carry. For a European business the short list is email, name, source, consent status, consent date and the wording the person agreed to.
- Entry points: website forms, chat, booking tool, webinar sign-ups, shop checkout, lead ads.
- Stores: CRM, email platform, spreadsheets, invoicing tool, support inbox.
- Identifier: the field that tells two records apart, usually the email address.
- Gaps: facts that live only in someone's inbox or head. Close these first, by hand if needed.
Step 3
Pick one workflow to build first
Start with lead capture to CRM to follow-up, unless your scores point elsewhere.
Choose the single workflow with the highest score that also has clean data behind it. For most small businesses that is the lead path: someone fills in a form, the contact lands in the CRM with its source and consent, the person gets a useful first reply within minutes, and the right colleague gets a task. It runs daily, the rules are clear and slow replies are easy to notice.
Write the workflow as plain sentences before you open a builder: one trigger, the steps in order, and what happens when something is missing. If you cannot describe it in ten lines, it is two workflows.
- Trigger: a new submission on the contact form.
- Check: if a contact with this email already exists, update it instead of creating a duplicate.
- Store: create or update the contact with source, consent status and timestamp.
- Reply: send the confirmation email that answers the three questions people ask most.
- Assign: create a task for the account owner, due the next working day.
- Fallback: if any step fails, post the raw submission to a shared channel so no lead is lost.
Step 4
Choose tools by job, not by brand
Use what your email or CRM platform already does, then add a connector and AI steps where needed.
Marketing automation tools do three different jobs. Platforms such as HubSpot, Brevo, Mailchimp, ActiveCampaign and Klaviyo store contacts and send messages, and each has its own automation builder. Connectors such as n8n, Make and Zapier move data between tools that do not talk to each other. AI steps call a language model inside a workflow to draft, classify, summarise or score.
Check the platform you already pay for first. A welcome email, a tag on form submit and a task for sales rarely need a second product. Add a connector when the trigger and the action sit in different tools. The three main connectors bill in different units: n8n counts workflow executions and can be self-hosted, Make counts credits and Zapier counts tasks. A long workflow that runs often can cost very different amounts on each, so estimate your monthly volume before you compare the n8n, Make and Zapier pricing pages.
Use AI steps for work that needs reading or writing, and keep rules as rules. A model is a good fit for drafting a reply, pulling the company name out of a free-text message or sorting enquiries into five categories. It is a poor fit for anything a simple condition can do, such as checking whether a field is empty.
| Job | Tool type | Examples |
|---|---|---|
| Store contacts, send email or SMS | Email or CRM platform | HubSpot, Brevo, Mailchimp, ActiveCampaign, Klaviyo, MailerLite |
| Move data between tools | Connector | n8n, Make, Zapier |
| Draft, classify, summarise, score | AI step inside a workflow | A language model called from the connector, or the AI features built into your platform |
| Fill in missing company data | Enrichment service | Clay, Dropcontact |
Step 5
Build it with a human approval step
Let the workflow prepare the work and let a person press send, at least at the start.
Build the workflow in a test setup with your own email address as the only contact. Run every branch, including the ones where data is missing. Then switch it on for real traffic with one rule: anything a customer will read that was written or chosen by a model waits for a person to approve it.
The main connectors support this pattern, though not on every plan. n8n documents human review for AI agent tool calls through channels such as Slack, Microsoft Teams and Gmail. Zapier has a built-in Human in the Loop step with approval requests, which its help centre lists for paid plans. In any tool you can get the same effect by saving the draft to the CRM or a shared channel and letting a second workflow send it once someone clicks approve.
Approval does not have to stay forever. Once a step has produced a few dozen drafts and you changed almost none of them, let it run on its own and review a sample each week. Keep approval in place for pricing, legal wording, complaints and anything sent to a large list.
- Name every workflow and every step so a colleague can read it without you.
- Log each run: input, output and who approved it.
- Add an error branch that alerts a person instead of failing silently.
- Keep API keys in the tool's credential store, never typed into a step.
Step 6
Measure whether it worked
Compare the workflow with the manual process on four or five numbers.
Decide what success means before the workflow goes live, and write down the starting value. For a lead workflow that is usually the time from enquiry to first reply, the share of leads that reach the CRM with complete data, the hours spent each week and the number of errors someone had to clean up.
Check the numbers after two weeks and again after two months. If the workflow saves time but the replies got worse, change the template, not the automation. If cleaning up errors takes longer than the old manual task did, switch the workflow off and go back to the data map. Switching off a workflow that does not pay is a good result, because you learned it cheaply.
| Measure | How to get it | What good looks like |
|---|---|---|
| Time to first reply | Form timestamp against the timestamp of the first email | Minutes instead of hours, also outside office hours |
| Data completeness | Share of new contacts with source and consent fields filled | Close to all of them |
| Hours saved | Minutes per run from your task log, times runs per week | More than the time you spend maintaining it |
| Error rate | Failed runs plus manual corrections per hundred runs | Falling week by week |
| Outcome | Meetings booked or orders from automated contacts | No worse than before, ideally better |
Step 7
Maintain it, then add the next workflow
Give every workflow an owner, a short description and a review date.
Automations decay. A form field gets renamed, an API changes, a colleague leaves and nobody remembers why contacts receive a certain tag. Every live workflow needs a named owner, a three-line description of what it does and why, and a date for the next review.
Review each workflow once a quarter: is it still running, is it still needed, and do the texts still match your offer and your prices. Then return to your scored list and pick the next task. Build one workflow at a time and let each run for a few weeks before you add another, so you can tell which change caused which effect.
- Keep one document that lists every workflow, its owner, its trigger and the tools it touches.
- Turn on failure alerts and send them to a shared channel, not to one person's inbox.
- Export or version a workflow before large edits so you can roll back.
- Remove access and rotate keys when someone leaves the team.
- Read your automated emails twice a year as if you were the customer.
GDPR and consent basics for automated marketing
This section is general information, not legal advice. Rules on email marketing are set partly by national law, so check the rules in your own country or ask a lawyer before you rely on any of it.
Under the GDPR you need a lawful basis for every use of personal data. Article 6 lists six of them, and for marketing the two that matter most are consent and legitimate interests. If you rely on consent, Article 7 says you must be able to show that the person gave it, the request must be clear and separate from other matters, and withdrawing must be as easy as giving. Article 21 gives everyone the right to object to direct marketing at any time, after which you must stop using their data for that purpose.
Marketing by email is also covered by the ePrivacy Directive. Article 13 requires prior consent for marketing email, with one exception: you may email your own customers about your own similar products or services if you obtained their address in the context of a sale and gave them a clear, free way to object, both when you collected it and in every message. Member states implement this differently, so the details vary by country.
- Store consent status, date, source and the exact wording with each contact, and make every workflow check it before sending.
- Put a working unsubscribe link in every marketing email and apply objections automatically in all connected tools.
- Sign a data processing agreement with each tool that handles contact data (Article 28 GDPR) and check where the data is hosted.
- For vendors outside the EU, check the transfer basis. The European Commission adopted an adequacy decision for the EU-US Data Privacy Framework in July 2023, which covers US companies certified under it.
- Send AI steps only the fields they need, and read the provider's terms on whether inputs are used for training. Business and API terms often differ from those of consumer chat apps.
- Do not let a model alone make decisions with legal or similarly significant effects on a person. Article 22 GDPR restricts solely automated decisions of that kind.
What not to automate
Some marketing work gets worse when a machine does it. Leave these with a person, and let automation do the routing and the reminders around them.
- Complaints, refund requests and anything emotional. Automate the routing so they reach someone fast. Let that person write the answer.
- Outreach to people who never asked to hear from you. In the EU, marketing email to individuals generally needs prior consent, and some countries apply similar rules to business addresses.
- Positioning, offers and pricing decisions. A workflow can report which message performed. It cannot decide what you stand for.
- Processes that are still changing. Run a task by hand until the steps stop moving, then automate it.
- Rare, one-off campaigns. If something runs twice a year, a checklist beats a workflow.
When to bring in help
Most small teams can build their first workflow alone, using the builder inside their email or CRM platform. Outside help earns its fee in four situations: the workflow crosses several systems and one of them has no ready-made integration; you want to self-host for data protection reasons and nobody on the team runs servers; AI steps need testing against real examples before customers see their output; or the manual work costs so much each week that spending months learning the tools is the expensive option. A freelancer suits one clearly scoped workflow. A studio or agency suits work that mixes process design, integration, custom code and ongoing care. Ask either for a fixed first scope, documentation and a handover, so you can run the result without them.
This guide is published by BlackCard, a boutique design and engineering studio managed from Germany that works mostly with European clients, in English and German. Its work includes smart workflows that connect tools so requests, files and follow-ups move automatically, practical AI tools such as draft generation and AI agents wired into existing systems, and business dashboards. It also advises on where AI does and does not make sense before anything is built. That makes it one option among several, and a typical fit for small and mid-size companies that want one senior team to design, build and automate. Everything in this guide works without hiring anyone.
Book a 20-minute callVisit blackcard.dev
BlackCard writes and publishes this guide.
Go deeper
- Marketing automation tools, sorted by the job they doMarketing automation tools grouped by job: email and CRM platforms, connectors such as n8n, Make and Zapier, AI steps and enrichment, in comparison tables.
- AI marketing workflows: nine recipes with trigger, steps and checksNine AI marketing workflow recipes with trigger, steps, tools and what to check by hand: enquiry triage, lead scoring, content reuse, reporting and more.
- DIY, freelancer or agency: who should build your marketing automationWhen to build marketing automation yourself, when to hire a freelancer and when an agency or studio is worth it, with what each option typically involves.
Frequently asked questions
What should I automate first in my marketing?
Automate the lead path first: form submission, CRM record, first reply, then a task for the person who follows up. It runs every day, the rules are clear and the result is easy to measure as time to first reply. If your scored task list shows something else costing more hours, such as weekly reporting or review requests, start there instead. Pick one workflow, not three.
Do I need a marketing automation platform or a connector like n8n, Make or Zapier?
Start with the platform you already use. Email and CRM tools such as HubSpot, Brevo, Mailchimp and ActiveCampaign include builders for welcome emails, tags and follow-ups. Add a connector when the trigger and the action live in different tools, for example a booking tool and your CRM. Many businesses end up with both: a platform that sends the messages and a connector that moves data around it.
Is marketing automation allowed under GDPR?
Yes. GDPR does not ban automation, it sets conditions for using personal data. You need a lawful basis such as consent, you must honour objections to direct marketing, and email marketing in the EU generally needs prior consent under ePrivacy rules, with a narrow exception for existing customers. Record consent with each contact and make every workflow check it. This is general information, not legal advice.
Can AI write and send my marketing emails automatically?
It can draft them, and a person should approve them before they go out, at least at first. Language models are useful for first drafts, subject line options, summaries of long enquiries and sorting messages into categories. They also make confident mistakes about prices, dates and product details. Keep approval for anything customer-facing until you have reviewed enough drafts to trust a specific step, then sample it weekly.
How long does it take to set up the first workflow?
A simple lead workflow inside one platform can typically be built and tested in a day or two by someone who knows the tool. The same workflow across three tools, with deduplication, error handling and an approval step, more often takes one to three weeks of part-time work, mostly because of data clean-up and testing. Add two weeks of light note-taking beforehand for the task log.
How much does marketing automation cost?
It depends on contact count, message volume and how many workflow runs you need. Email and CRM platforms usually charge by number of contacts and feature tier. Connectors charge by usage: n8n by workflow executions, Make by credits and Zapier by tasks. Several tools offer free plans for small volumes. Estimate your monthly runs and contacts first, then compare the vendors' own pricing pages, because list prices change often.
Should I self-host n8n?
Only if someone on the team can maintain a server. n8n can run on your own infrastructure, which keeps workflow data under your control, and its Community Edition can be used for internal business purposes under the Sustainable Use License. In return you take on updates, backups and security. According to n8n's pricing page, its hosted plans store data in the EU, in Frankfurt, which is enough for many small businesses.