AI agents

AI agents for business start with secure data.

AI agents for business are AI systems that carry out tasks on their own, such as retrieving data, drafting a report or triggering an action in another system. How secure they are depends mainly on the data they can access. ENABLE doesn't run agents itself, but provides the controlled data layer: published, read-only query endpoints with an API key, so an agent gets exactly one scoped dataset.

Book a demo Explore the platform

enable.dataanalytics.nl/acme

EnglishClient adminSV

Publish as API

Saved query · revenue-by-region · version 3

Endpoint

Endpoint slugrevenue-by-region

URLGET/api/v1/queries/revenue-by-region

X-API-Keyak_••••••••••••3f9a

Rate limit60 requests / min

AccessRead-only

View OpenAPI specification

Test call 200 OK · 184 ms

{
  "query": "revenue-by-region",
  "version": 3,
  "rows": [
    { "region": "South", "revenue_2026": 1421300 },
    { "region": "North", "revenue_2026": 1284900 },
    …
  ]
}

AI agents for business are AI systems that don't just answer, but take steps themselves: looking up information, calling a tool, drawing a conclusion and sometimes carrying out an action. An agent that prepares the stock report every Monday, for example, needs to be able to retrieve stock and sales figures.

That is exactly where the biggest risk lies. An agent with broad access to your database or systems can see and do more than necessary, and afterwards it is hard to trace why it did something. Working responsibly with AI agents therefore starts with limited, verifiable access to data.

ENABLE does not run AI agents itself and is not an agent platform. What ENABLE does do: publish a query from your data warehouse as a secured REST endpoint. An agent or automation you build elsewhere then gets exactly that one read-only dataset, with its own API key, an OpenAPI specification and rate limits.

Key points

AI agents for business at a glance

One dataset per endpoint

The endpoint always runs the same published query. An agent cannot send its own SQL and therefore cannot go searching through your warehouse.

Read-only

Every call runs in a read-only transaction. Through ENABLE, an agent cannot change, add or delete anything.

Its own API key

Each endpoint has its own key in the X-API-Key header. If you switch off the publication, the endpoint stops working.

OpenAPI specification

ENABLE describes your published endpoints in an OpenAPI specification. Many agent and automation tools can import such a specification as a tool description.

Only your parameters

You decide which parameters an endpoint accepts, such as a period or branch. An agent cannot add other filters.

Rate limits and maximum rows

The number of calls per endpoint and caller is capped, as is the number of rows per response. That limits the damage if an agent gets stuck in a loop.

Version and publication log

An endpoint serves the published version of the query. Changes only go live when a customer admin republishes, and that is recorded in the activity log.

What are AI agents, and what do they do in a business?

An AI agent combines a language model with tools, such as an API, a database or a mailbox, and lets the model decide for itself what the next step is. Businesses use agents or "digital workers" to prepare reports, sort incoming requests or enrich data, for example. This is also known as agentic AI.

The difference from classic workflow automation is that an agent doesn't follow a fixed sequence of steps but decides for itself. That makes it flexible, but also less predictable.

	Chatbot	Workflow automation	AI agent
What it does	Answers questions	Carries out fixed steps	Chooses its own steps and tools
Predictable	Fairly	Yes	Less so
Access to systems	Usually none	Configured per step	Through the tools you give it
Typical risk	Incorrect answer	Error in the rule	Overly broad access, unexpected actions

The risks of AI agents with overly broad access

Most problems with agents arise not from the model, but from what the model is allowed to do.

Too much data: an agent with a database password can read everything that account is allowed to read.
Untraceable actions: if an agent composes its own queries or API requests, it is hard to establish afterwards what it retrieved and why.
Prompt injection: text in an email, document or web page can give an agent instructions you never intended.
Errors that compound: a wrong assumption in the first step becomes a decision by the fifth.
Load and costs: an agent stuck in a loop can make thousands of requests.

AI agents for business: connect them to data securely with the Data API

You reduce most risks by giving the agent less: only the data it needs, read-only and through a channel you can shut off. Take the fictional Acme Groothandel. An automation that drafts a stock report every Monday calls one endpoint listing items below their minimum level: item, location, stock and minimum. Purchase prices and customer details are not in that query, so the agent can't see them either.

	Agent with database access	Agent via the ENABLE Data API
What the agent can read	Everything the database account can	Only the result of one published query
Who decides the query	The agent	A customer admin, when publishing
Writing	Depends on the permissions	Not possible, read-only
Shutting off	Change the password	Switch off the publication
Documentation	Describe it yourself	OpenAPI specification

Automating business processes with AI: where do you start?

With AI automation, the rule is: start small and keep a human in the loop until you know how the agent behaves. You build, host and monitor the agent itself in the tool of your choice; ENABLE supplies the data.

Choose one well-defined process with a clear outcome, such as a weekly report.
Determine which data it really needs, and which it doesn't.
Publish exactly that data as an endpoint, with only the parameters the agent needs.
Have an employee check the outcome before anything happens.
Only expand once the first process works predictably.

In ENABLE

How to give an AI agent data through ENABLE

01
Write the query
In the SQL Explorer, write a query, with AI assistance if you like, containing only the columns the agent needs.
02
Define the parameters
Specify which values the agent may pass, for example a period or branch.
03
Publish as an endpoint
A customer admin publishes the query as a REST endpoint. You get an API key and the OpenAPI specification.
04
Connect the agent
Set up the endpoint as a tool in your agent or automation environment and store the API key there as a secret.
05
Monitor and shut off
Rate limits cap usage. If something doesn't work as intended, you switch off the publication.

Book a demo How the platform works

FAQ

Frequently asked questions

What are AI agents for business?

AI agents are AI systems that independently take steps to complete a task, such as retrieving data, drafting a report or forwarding a request. They combine a language model with tools such as APIs and databases.

Does ENABLE have AI agents?

No. ENABLE does not run agents itself and is not an agent platform. ENABLE can, however, publish a query from your data warehouse as a secured, read-only REST endpoint that an agent you build elsewhere can use.

How do I give an AI agent secure access to company data?

Give the agent only the data it needs, read-only and through a channel you can shut off. With the ENABLE Data API, an agent gets the result of one published query, with its own API key and rate limits, instead of a database password.

Can an agent change data through ENABLE?

No. Published endpoints run in a read-only transaction and only execute the fixed, published query. An agent cannot send its own SQL.

What is the difference between an AI agent and workflow automation?

Workflow automation follows a fixed sequence of steps that you set up in advance. An AI agent decides for itself which step and which tool to use. That makes an agent more flexible, but also less predictable.

Which agent or automation tools work with the Data API?

Any tool that can send an HTTPS request with a header. The endpoint expects the API key in the X-API-Key header and returns JSON. Tools that support OpenAPI can use the specification as a description of the endpoint.

Share securely, live fast, no headache

In an online demo we show you the portal: dashboards per role, row-level security, plain-language questions and how we set it up and manage it for you.

Book a demo info@enableyourdata.nl