What Is a Connector? How to Connect Your Business Systems
A connector is the building block that lets two systems talk to each other: CRM, accounting, email and calendar. Here you get the concept explained in plain terms, the difference between connector, integration and API, and a concrete checklist for security and GDPR when you connect the business's systems.
Table of contents
A connector connects two systems, for example an AI agent and your CRM, calendar or email account, so that they can exchange data. Access has to be delimited and secured in the specific set-up.
A connector is a piece of ready-built software that connects two systems so that they can exchange data automatically, without a person having to copy information back and forth. When your CRM automatically knows the latest invoice from the accounting system, or when an appointment in the calendar automatically lands as an activity on the right customer, it is a connector doing the work. The connector handles the technical side (access, translation of data formats and security) so you don't have to build the connection yourself from scratch.
So the short answer is: a connector is the bridge between two systems. The rest of this article explains why that bridge has become so important, how it actually works (APIs, OAuth and tokens: explained so a non-technical director can follow), what the difference is between a connector, an integration and an API, and what you should demand in terms of security and GDPR before you connect anything at all.
What is a connector? The definition: short and precise
Let's pin down the definition so that it can stand on its own:
A connector is a reusable software component that translates between one particular system (for example a CRM, an accounting system or an email account) and the rest of your IT landscape. It knows the system's language (its API, its login method and its data formats) and provides a uniform connection that other tools can use.
Think of a travel adapter. Your French charger doesn't fit an English socket, but the adapter translates between the two standards so the power flows. A connector does the same for data: Economic speaks one language, HubSpot another, Outlook a third, and the connector translates so that data can flow between them.
The three things a connector always handles
A well-built connector solves three tasks that would otherwise land on your own desk:
- Access: It logs into the system on your behalf (typically via OAuth (more on that later)) and keeps the access up to date, without anyone having to store passwords in a spreadsheet.
- Translation: It knows the system's data model. It knows that what one system calls a "contact" is called a "person" in another, and that date formats, currency and fields have to be mapped correctly.
- Robustness: It handles the boring stuff: what happens when the system is down, when too many requests come in, or when the access expires? A good connector fails gracefully and retries, instead of dropping data on the floor.
Why the word "connector" is turning up everywhere right now
The concept isn't new; integration platforms have used it for years. But it has made a significant comeback, because AI agents need access to the business's systems to be able to do real work. An AI assistant that can't see your calendar, CRM or accounting data can only talk. An AI agent with connectors can act: look up the customer, find the unpaid invoice, write the draft of the follow-up email and put the meeting in the calendar. The connector has thus become the precondition for AI moving from "chatbot" to "colleague". If you want the background to what this kind of automation means in practice, we've written a thorough introduction in what is CRM automation.
The problem: the business's data lives in silos
Before we go deeper into the technology, it is worth understanding why connectors are necessary at all. The answer is data silos, and virtually every business has them.
How silos arise
Nobody decides one morning to build data silos. They arise gradually and for perfectly sensible reasons:
- Sales chooses a CRM, because it is good for pipeline and customer dialogue.
- The bookkeeping department chooses an accounting system, because it is good for invoices and VAT.
- Everyone uses email and calendar, because that is where the day-to-day happens.
- Warehouse or operations may run in an older specialist system or a spreadsheet, because "it works, doesn't it".
Each choice is right in isolation. The problem is that the systems were never asked to talk to each other. The customer's name now exists in four places, in four spellings, with four different update dates, and none of the systems knows what the others know.
What silos cost in day-to-day work
The cost rarely shows up as one big figure on the bottom line. It shows up as friction that eats time and creates errors:
- Double entry: The same order is keyed into the CRM, the accounting system and perhaps a spreadsheet. Every entry is a chance for an error.
- Out-of-date information: The salesperson rings a customer who has an unpaid reminder outstanding, but it only shows in the accounting system, which the salesperson doesn't look in.
- No single customer history: What has the customer bought, asked about, complained about and paid? The answer is spread across inboxes, systems and memories.
- Decisions on gut feeling: When data can't be brought together, management reporting becomes manual work, or guesswork.
The silo problem is made worse by poor data discipline: duplicates, outdated contacts and half-empty fields make every integration harder. That is why connectors and data quality are closely linked; we've gathered the practical advice in our guide to CRM data hygiene. Do clean up before you connect the systems; otherwise you're just syncing the mess around faster.
How a connector works: APIs, OAuth and tokens explained in plain terms
Now to the engine room. You don't need to be able to program to understand how a connector works; three concepts are enough: API, OAuth and tokens.
The API: the system's staffed entrance
An API (Application Programming Interface) is the official, staffed entrance to a system, built for machines instead of people. Where you yourself log into the accounting system via a website and click around, another program can use the API to ask precise questions and get precise answers: "Give me all invoices for customer X" or "Create this contact".
As an image, the website can be the shop's entrance for customers and the API a goods-receiving bay with fixed rules. Many business systems offer an API, but access and functions depend on the system, version and subscription. The connector has to fit exactly that access.
OAuth: the key card instead of the master key
When a connector needs access to your systems, the access method has to be assessed. A shared username and password can make it hard to delimit and revoke the access. Permissions, expiry and options for separate accounts depend on the system; a password is not necessarily access to everything.
OAuth is used to give a program delegated access without sharing the user's password with the program. Logins such as "Sign in with Google" typically also build on OpenID Connect; OAuth is about authorisation. As a simplified image, delegated access can be compared to a key card:
- You approve or manage the access the connector is to have, according to the system's options.
- The permissions can be delimited with the access scopes the system offers, for example reading the calendar.
- Find out how the access is revoked, and when the revocation takes effect.
- With OAuth the connector gets a token for the agreed access; it doesn't get your password via the OAuth permission.
Tokens: access passes with an expiry date
An access token is a digital access pass. It can be an opaque string or have a structured format; the connector uses it according to the system's rules. Two things in particular need to be looked into:
- Permissions: Limit access to the necessary functions and data with the scopes and other access controls the system supports.
- Lifetime and renewal: A short lifetime on an access token can limit exposure in the event of a leak. A refresh token can issue new access tokens and therefore has to be protected separately. Look into the vendor's rotation or binding to the client, as well as options for revocation; a token leak does not automatically mean only a short-lived risk.
In this example the connector uses an API with an access token via OAuth. Other systems may require other access methods. Clarify the specific method, the permissions and the maintenance before the connection is put into use.
What happens when something goes wrong?
An underrated part of a connector's work is error handling. Systems go down from time to time, APIs set limits on how many requests you may send per minute, and tokens expire. A professional connector handles all of this: it waits and retries on temporary errors, it respects the system's limits, and it says clearly if an access needs to be renewed. That is exactly the kind of work nobody wants to build and maintain themselves for every single system.
Connector vs. integration vs. API: the terms separated
The three words are often used interchangeably, and that creates confusion in both purchasing and management meetings. Here is the clean version: the API is the system's entrance, the connector is the reusable building block that uses the entrance, and the integration is the complete, concrete flow you build with the building blocks.
Connector, integration and API: what is what?
| Term | What is it? | Everyday image | Who supplies it? |
|---|---|---|---|
| API | The system's official, machine-facing entrance: fixed rules for fetching and sending data. | The goods-receiving bay behind the shop, with fixed rules for delivery. | The system vendor (the CRM, the accounting system, the email provider). |
| Connector | A reusable component that knows a system's API, access method and data model. | The travel adapter that fits exactly that socket. | The integration platform or AI platform (for example a platform with a curated selection of connectors). |
| Integration | The concrete, complete flow between two or more systems, built on top of connectors. | The whole electrical installation: adapters, wiring and the lamp that actually lights up. | You yourself, a partner or the platform, configured to your needs. |
A practical consequence of the distinction: when a vendor says "we have an integration with X", ask whether they mean a connector (the building block exists, but the flow still has to be built) or a ready-made integration (the flow works from day one). The difference can be weeks of project work.
Examples of everyday flows that require connectors
Theory is good, but connectors only become interesting when you see what they make possible in an ordinary working week. Here are three recognisable flows.
From customer meeting to updated CRM
A classic gap in most businesses: the meeting is held, but the CRM never gets told. With connectors to calendar, email and CRM, the flow can look like this:
- The meeting is in the calendar with the customer's email address as an attendee.
- The calendar connector registers that the scheduled end time has passed. An employee or an actual meeting data source confirms whether the meeting was held.
- The CRM connector finds the customer and logs the meeting as a held activity once it has been confirmed as held.
- The email connector finds the latest correspondence, so the follow-up has context.
- A draft follow-up email is ready; a person approves and sends it.
None of it is magic. It's three connectors and a well-defined flow. If you want to see how this kind of flow fits in with the whole sales process, we go into depth in our complete guide to automating sales processes.
Invoice follow-up across accounting system and email
If an invoice is overdue, a flow with connectors to the accounting system, email and CRM can support the follow-up:
- The accounting connector sees that invoice 1042 is overdue.
- The CRM connector looks up the customer: is there an active dialogue or an open case that you should know about before chasing?
- A friendly reminder draft is generated with the right invoice details.
- A person approves, and the follow-up is logged on the customer in the CRM.
Note the pattern: the connector fetches and compiles; the person decides. That is the division of labour that makes automation safe.
Stock status, orders and customer service
The customer rings: "Where's my order?" Without connectors that means three systems open on the screen and long keyboard pauses down the line. With connectors to the order system, warehouse and email, an employee (or an AI agent preparing the answer for the employee) can pull the picture together in seconds: the order exists, the item is in stock, the shipment is on its way, and the track-and-trace number is in the email that was sent on Tuesday. Incidentally, the overall view across systems is the whole basic idea of modern revenue operations, a discipline we've described in the Revenue Operations guide.
Security: what you should demand before you connect anything
Connectors open doors into your most sensitive data. That is why the security requirements aren't an appendix; they are half the decision. Here are the four requirements you should put to every vendor.
Encryption of tokens and data
Tokens are key cards, and key cards should be kept in a safe, not in a drawer. Demand that:
- Tokens are stored encrypted (not in plain text in a database or a configuration file).
- All traffic between connector and system takes place over encrypted connections (TLS/HTTPS).
- The vendor can answer clearly where tokens are stored, who can access them, and how they are rotated.
If the vendor can't answer the three questions precisely, that is a warning sign in itself.
Least privilege
The principle is simple: a connector should have access only to exactly what it needs, neither more nor for longer. In practice that means:
- Delimited scopes: If the flow only needs to read the calendar, the token should only be able to read the calendar. "Full access, it's easiest" is a trap.
- Access per purpose: Avoid a single catch-all account with access to everything. Separate the accesses, so that they can be withdrawn individually.
- Visibility: You should be able to see a list of active accesses (which connectors have access to what) and be able to block each one immediately.
GDPR and data processing agreements
When a connector processes information about identifiable persons, the data protection rules have to be assessed. Customer data can contain personal data, but isn't always personal data. Two roles are crucial to get in place: you are the data controller if you determine the purposes and means of the processing; the vendor is the data processor when it processes on your behalf. The Danish Data Protection Agency explains the allocation of roles and the requirement for a data processing agreement in its guidance on data controller and data processor. In a data processor relationship you must also exercise relevant oversight of the processing.
The overall GDPR requirements for businesses: when the rules apply, what counts as personal data, and which rights data subjects have, are collected in the EU's official review of the data protection rules for businesses. The minimum for you before a connector is put into operation:
- Data processing agreement in place, when the vendor processes personal data on your behalf.
- Clarity about data flows: Which information is moved, to where, and are copies kept along the way?
- Location of data: Is data processed in the EU/EEA, or is it transferred to third countries, and on what basis?
- Deletion: What happens to data (and tokens) when the collaboration ends?
If you work with AI on top of your customer data, requirements under the EU's AI Act may also apply, depending on the system, the use and your role. We've gone through this in the article on the EU AI Act and sales in Danish businesses.
Control and audit
The last requirement: traceability. You should be able to answer "who did what, when?" even when the "who" is a connector. Demand logging of which actions were carried out through which accesses, and the ability to review it afterwards. Without a log, troubleshooting is guesswork, and compliance conversations become unpleasant.
iPaaS and integration platforms vs. AI agent with connectors
Once you've decided that the systems should be connected, the choice typically lies between two architectures, and it is worth understanding both, because they aren't mutually exclusive.
The classic route: iPaaS and integration platforms
An iPaaS (integration Platform as a Service: for example Zapier, Make or Workato) is a platform where you build flows of the type "when X happens in system A, do Y in system B". The strengths are real:
- Maturity: The category has existed for a long time, and the platforms have connectors to hundreds of systems.
- Predefined logic: A flow describes the rules in advance. That can suit fixed tasks, but it still requires handling of errors and exceptions.
- Low barrier to entry: Simple flows can be built without developers.
Flows require maintenance, oversight and handling of changes in the connected systems. Integration platforms can support branching, retries and other error handling. Look into whether the specific platform and set-up handle your exceptions.
The new route: AI agent with connectors
The alternative (or supplement) is an AI agent that has connectors as its "hands". The difference in thinking is significant:
- An iPaaS carries out predefined flows. An AI agent is given a goal ("follow up on overdue invoices with active customers") and itself uses the relevant connectors to gather information and propose actions.
- An iPaaS can handle errors with, for example, retries and alternative paths. An agent can choose tools dynamically, but can also fail and needs limits and the option of human follow-up.
- Both models require maintenance of connections, rules and access. An agent doesn't automatically gather all logic or control in one place.
A fixed, rule-based task can be suited to a classic integration or iPaaS. An agent can be tried out when the task requires dynamic tool choices or assessment of context. Compare quality, costs and errors on your specific task; the models can also be combined.
How we can work at Agent360
We can build the solution in your own systems. We can use the integrations the systems make available, and agree a division of labour as in the examples above: the agent can gather and prepare via connectors, while a person approves what matters.
Checklist: when you assess integration options
Finally, the practical part. Print this list, and use it the next time a vendor says "we integrate with everything":
- Does the connector exist for exactly your systems? Not "a similar system": yours. Also check your version/edition (cloud vs. locally installed makes a difference).
- Connector or ready-made integration? Is it a building block, or does the flow work from day one? What does it take to get from one to the other?
- How is access granted? Look into OAuth with the necessary scopes, where the system supports it, or the alternative access method. Demand an explanation of permissions, protection and revocation.
- How are tokens stored? Encrypted, with clear answers on where, who and how they are rotated and blocked.
- Can you see and control accesses centrally? One overview of all active connectors and their permissions, with the ability to withdraw each one immediately.
- Is the GDPR foundation in place? Clarify purpose, legal basis for processing, necessary data, information, rights, data flows and deletion routines. Establish the roles, and enter into a data processing agreement when the vendor processes on your behalf.
- What happens in case of errors? Does the connector retry? Do you get notified? Can data be lost along the way, and how would you find out?
- Is there a log and traceability? "Who did what, when" must be answerable, also for automated actions.
- Who maintains the connector? APIs change. Is it the vendor's responsibility to keep up, or your problem when something breaks?
- Is there an exit? Clarify export, migration, licence and technical dependencies. Self-hosting and access to source code can be relevant, but don't guarantee that the whole solution can be run on without the vendor.
If a vendor can answer all ten points clearly, you're in good hands. If they can't, you've just saved yourself an expensive project.
Frequently asked questions about connectors
Q: Is a connector the same as an app or a plugin?
Not quite. A plugin extends one program with new functionality inside the program (for example a spell checker in your browser). A connector connects two separate systems so that data can flow between them. In some platforms connectors are admittedly delivered as "apps" that you install, but the function is the bridge between systems, not a new button in one system.
Q: Do we need to be able to program to use connectors?
No, that's exactly the point. The connector is built so that the technical side (API calls, OAuth, data formats, error handling) is packed away. Your job is to approve the access (typically a couple of clicks in a login window) and decide which flows should use it. The complexity moves from your IT department to the vendor of the connector, and that is where it belongs.
Q: Is it safe to give a connector access to our systems?
Security depends on the specific set-up. Look into access via OAuth, protection of tokens, minimum necessary permissions, logging and how the access is revoked. Also clarify roles and the data processing agreement, where one is required. Delimited permissions can limit the consequences of misuse, but there is no general guarantee that a connector is safer than a particular manual workflow.
Q: What happens if we terminate the collaboration with a connector vendor?
Clarify the termination process before the agreement is entered into: how accesses are revoked, which data is returned or deleted, and how deletion is documented. In a data processor relationship, the agreement must regulate the return or deletion of personal data, subject to any legal requirements to retain it. The right to configuration and to continue running things yourself also depends on the specific terms.
Q: Our systems are old, can they be connected at all?
Often yes. Most systems, including older ones, have some form of machine access: an API, a database, file exchange or export/import. A connector can be built on top of several of these routes, although modern APIs with OAuth are clearly the best starting point. The honest answer, though, is: the older and more closed the system, the more manual work the connection requires, and in some cases replacing the system is the right investment rather than a fragile bridge to a dying system.
Q: How many connectors do we need?
Fewer than you think. Start with the three or four systems where the day-to-day friction is greatest: typically CRM, accounting system, email and calendar. Those four alone cover most of the flows described in this article: meeting follow-up, invoice follow-up and a unified customer history. Then expand as needed instead of connecting everything from day one. Every connector is an access that has to be managed, so connect with care, not by reflex.
Conclusion: the connector is the bridge, and bridges have to be built properly
A connector is a software component that connects systems so that they can exchange data. Access has to be delimited, secured and maintained in the specific set-up. Compare the solution's costs and risks with your current workflow, including manual copying and any sharing of passwords.
The most important things to take away from here: understand the difference between API, connector and integration when you talk to vendors; look into OAuth, protected tokens and least necessary permissions; clarify GDPR roles and the relevant agreements before personal data is processed; and choose architecture according to the task. Use the checklist's questions to get concrete answers from the vendor.
Tags: #connectors #integrations #API #OAuth #system-integration #GDPR #AI-agent
Contents
Unsure about the rules for your own set-up?
We look at what you use today and what it requires: documentation, consent and what the EU AI Act means for you. 20 minutes, no commitment.
Book a reviewRelated articles

MFA: multi-factor authentication for your business – how to get started
Learn how to implement MFA securely in your organisation. Guidance on methods, costs and best practices for protecting company data.

English-language invoices: requirements for invoices to foreign customers
Invoice in English to foreign customers legally. Danish regulations, EU and international VAT rules, zero-rating, reverse charge, and automation.

Public Procurement in Denmark: how to bid for government contracts
How Denmark's Public Procurement Act works for suppliers. Threshold values, tender procedures, exclusion grounds and AI solutions for proposal management.