n8n automation for companies

n8n automation for a company is a workflow that fires on its own, writes the result into the right system, and tells a person when it fails. A workflow that only runs while someone has the editor open is a prototype. This page is the work of putting the workflow into production.

The usual targets are a CRM, WhatsApp, an ERP, a store, and a spreadsheet. Internal routines fit too: a collection nudge, a lead handed to sales, a file that leaves one system and lands in another. The cost comparison between self-hosted and Cloud, with the official prices and the things that break, is in n8n in 2026: self-hosted vs Cloud.

Who this is for

It is for a company that already knows which repeated job should leave someone's hands. The trigger exists: a form, a message, a clock, a new spreadsheet row, a status in the ERP. What is missing is the workflow that does the middle without an analyst retyping fields.

It is also for a team that already has an n8n experiment and has hit the wall. The credential lives on one laptop. Nobody saw yesterday's failed execution. An upgrade moved the server and the saved passwords no longer open.

It is not for a team that wants to "look at the tool." Community Edition is free to install, and n8n documents what the license allows Source [2] Source [3]. Trying it is an afternoon. Production is a different conversation.

Buyers in the US, Europe, Israel, and Latin America hit the same shape. The difference is the channel. In much of Latin America, Spain, and Israel, WhatsApp is how the sale actually happens. A workflow that updates a CRM and never reaches that chat is half a workflow.

What we deliver

Workflows that run in production, with a named owner on your side. Where AI shows up, it shows up as a step with a defined input and a defined output: classify a message, pull a field, draft a reply a person approves. The model is not left loose in the middle of the process with nowhere to land.

Integrations you already run: CRM, WhatsApp Business through the official API, ERP, e-commerce, and spreadsheets. Data stays in your accounts. The access we ask for is the minimum the workflow needs. Client data is not used to train public models. Handling follows LGPD and GDPR practice.

A decision about where n8n lives. n8n Cloud, on Starter and Pro, is hosted by n8n, with data in the EU, in Frankfurt. On the pricing page as read on 2 October 2026, Business is self-hosted only (n8n's own cloud for that plan is a waitlist). Enterprise can be Cloud or self-hosted Source [1]. Self-hosted Community is free for internal use. The license is not "do whatever you like" Source [2].

Failure monitoring. A broken execution has to show up the same day, not at the next monthly meeting. The pricing page lists "Error workflows" under debugging. Check that the item is on the plan you buy. What each plan card does say: Starter and Business include forum support. Dedicated support with an SLA is Enterprise Source [1]. On a self-hosted instance, logs are part of operations, and the backup guide asks for a full backup before you update Source [4].

Upkeep. A production workflow changes because the CRM renamed a field, WhatsApp changed a template, or n8n shipped a version. The agreement names who upgrades, when, and how the encryption key survives a server move.

How it works

The sequence is the studio's usual one: diagnosis, strategy, execution, evolution. Two-week sprints, with a delivery each cycle.

Diagnosis looks at the repeated job, the systems, and who owns each credential. Strategy picks Cloud or your own server, what ships in the first workflow, and what waits. Execution builds, tests against a small slice of real data, and only then turns the trigger on. Evolution is the following month: the workflow that broke, the new field, the schedule that moved.

While we build, AI agents help write, test, and review the workflow. A person reads it before it is activated. That does not replace the process owner on your side. Someone there has to know what the workflow does when a vendor changes an API.

Most automations go live in two to four weeks. A full project, with several systems and a designed exception path, takes four to eight weeks. A sprint is US$1,200 and you can cancel when you want. A project, with scope, timeline, and price fixed, starts at US$3,000, with a delivery each sprint and support after delivery. An ongoing partnership, with a roadmap and a monthly check-in, starts at US$900 a month. The three shapes are on pricing. The first conversation is 30 minutes and does not commit you.

The studio is based in São Paulo and already works with clients in São Paulo, New York, Madrid, and Tel Aviv. Delivery is in English and in Portuguese.

Self-hosted, Cloud, and the license

The community license, the Sustainable Use License, lets you use and modify n8n for your own internal business purposes, or for non-commercial and personal use. You may distribute the software only free of charge, for non-commercial purposes. Files marked .ee. are not under that license Source [2].

The license FAQ applies to self-hosted n8n, not to Cloud, and it is specific about consulting. You may build automations for a client on your instance as long as the client cannot create or edit them. You may charge for creation, setup, maintenance, training, and consultancy. You may install n8n on the client's own server as long as you are not also hosting it for them. You may not host n8n as a service so clients build workflows. You may not let end users build or configure their own workflows through a custom UI, the API, MCP, or an AI agent acting for them. You may not remove the branding and resell n8n as another product Source [3].

The work samambai does fits what that FAQ allows. The client receives the output of the workflow (the message sent, the record created, the alert). The client does not receive the editor on the studio's instance to build whatever they want. If the company wants n8n on its own server, the install is on that server.

Credentials are what break when the machine changes. n8n creates a key on first launch and stores it in ~/.n8n. That key encrypts credentials before they are written to the database. Without it, a restored database cannot open the passwords. The docs say to set N8N_ENCRYPTION_KEY on purpose. In queue mode, every worker uses the same key as the main instance Source [5]. A complete backup is the .n8n folder plus external Postgres if you use it. The command-line export carries workflows and credentials. It does not carry users, roles, execution history, variables, or the key Source [4].

WhatsApp, where WhatsApp is the sales channel

In Brazil, Mexico, Spain, Israel, and a large part of Latin America, the customer does not go to a portal to answer. They answer in WhatsApp. Connecting n8n to that chat goes through the official API, not through a personal phone left on a desk.

Meta's phone-number documentation asks you to send the customer number with a plus sign and a country calling code. If the plus sign is missing, the API prepends the business number's country code, and the message can be misdelivered. For Brazil and Mexico, Meta states that the extra prefix on the phone number may be modified by the Cloud API, and that this is standard behavior, not a bug Source [6]. Operationally: the number sitting in the CRM, written the way a local person dials it, is not always the number the API accepts. The workflow needs a normalization step before send.

A number already on WhatsApp Messenger has to be deleted there before it can be registered on the API. Registration requires a two-step verification PIN. A new business portfolio starts capped at two numbers, until the business is verified or it reaches a messaging limit of 2,000 Source [6]. An agent that actually talks to the customer on that channel is a separate page: WhatsApp AI agents.

What we do not do

We do not host an n8n so your end customer can build flows. The community license forbids it, and the paid plan for a product where the user assembles automations is a conversation with n8n, not a shortcut Source [3].

We do not strip n8n's branding to ship another product on the same code Source [2] Source [3].

We do not keep your operation's data in our accounts. The CRM, the spreadsheet, WhatsApp, and the ERP stay yours. The credential lives in the instance's store, encrypted, not in a shared document.

We do not sell n8n SSO, multiple environments, queue-mode scaling, or an n8n SLA as if they were included in a US$1,200 sprint. SSO, SAML, LDAP, environments, queue mode, and Git are on n8n's Business card. A dedicated SLA is Enterprise. Those are licenses your company buys from n8n Source [1]. The sprint is the work of building and running the workflow.

We do not promise that an AI step "handles the customer alone" with no path to a person. A sales or booking message has a handoff for the case the model does not know.

When not to hire samambai

Do not hire us if the job is a single, well-specified step and someone inside already operates n8n. A freelancer, or that person, is the right shape. A studio is the right shape when production, credentials, WhatsApp, and upkeep have no owner.

Do not hire us if the goal is a product in which your customer builds their own automation. That leaves the community license Source [3].

Do not hire us if you want the studio to own the data. We run the workflow. The account is yours.

Do not hire us if the deadline is tomorrow morning and the workflow crosses three systems nobody has documented. A prototype can exist on that clock. A monitored production flow cannot.

Do not hire us if automation is the company's product and the volume justifies an internal on-call engineer plus a Business or Enterprise license. At that point a studio in the middle gets in the way. We can help you get there. We do not replace the team once the workflow is the business.

One server, no client names

The studio runs its own n8n server, with dozens of workflows in production: support, integrations, and routines. It is not a demo. It is the same kind of piece we deliver: a credential with an owner, a failure you can see, an upgrade that does not lose the key.

We do not publish client names or message counts here. If you need an execution estimate to pick a plan, use your own volume and n8n's official table, not a marketing case study Source [1].

If the next step is the workflow rather than the reading, the first call is 30 minutes. If you want the Cloud-versus-server arithmetic first, start with the cost article and with pricing.