How does Unloq connect to your data?
The short answer
Unloq connects to your data read-only, through paths you approve, reaching only the data and documents you point it at, from a modern warehouse to a legacy ERP. No pipeline or warehouse is rebuilt before the first answer. In our introductory calls over the past year, numerous companies asked this, most often about an older ERP or a large SAP or Oracle estate.
What are the ways in?
Unloq reaches your data through the main routes below, from most integrated to lightest. Each of them is read-only:
Deployed inside your environment. Unloq runs in your AWS or Azure tenancy, inside your VPC, and reads your systems from there.
A live read-only connection to the databases you approve. Unloq reads only the governed tables agreed for it.
A secure extract, as SQL or CSV, one-off or on a schedule, for when a live connection is unavailable or unwanted for now. Each extract is kept with its source reference, so an answer built on it still names its source.
Documents and files, such as contracts, board packs, procedures, reports and strategy memos. Unloq reads and indexes them, and every answer that uses one cites the exact document and page.
What does read-only mean here?
Read-only means Unloq reads the data you approve and does not change or delete it. Your team grants Unloq’s access, and Unloq can reach only what you have enabled. For a database, that is the governed tables agreed during onboarding. Answers reach named owners in Unloq, your CRM or Slack, and a person decides what, if anything, changes in your systems as a result.
Which systems can Unloq reach?
Beyond the four routes above, Unloq reaches whatever a business chooses to connect through integrations and MCP connectors, which use the open Model Context Protocol. The systems businesses commonly connect include CRMs such as HubSpot and Salesforce, finance systems such as Xero, QuickBooks and Sage, Google Sheets and Google Drive, Gmail and Slack, Snowflake and AWS, and Shopify. Our data team sets up each connection with yours during onboarding.
The platform is the same for every client. What we set up for you is the data connections and the metric definitions. An old ERP usually connects through a scheduled extract or a live read-only connection, and the two teams settle which during data discovery. On a large SAP estate we start with the systems the first decision needs and leave the rest until a later decision asks for them.
Does Unloq work when data is spread across acquired businesses?
Yes. Fragmented, post-acquisition data is the case the semantic layer is built for. Take two acquired businesses that count a new customer differently, one at first order and one at contract signature. Both systems stay as they are, and a definition you own decides which count an answer uses. The mapping is real work your team does with ours, and data we cannot reach is data Unloq cannot use. How definitions are agreed is on why Unloq gives the same answer every time.
Which AI models does Unloq use?
Unloq is model-agnostic and sends each task to a model that fits it. That way a simple step does not go to the most expensive model available. Unloq uses frontier models, Anthropic’s Claude among them, and models your own data scientists have built can plug in.
What this page does not cover
Storage and access are on where your data sits, and sources on how every answer is evidenced. Why connecting AI is no longer the hard part makes the wider argument. How Unloq describes what happened shows what Unloq does with your data once it is connected, pulling every system into one picture, and answers whether the data needs cleaning first.
Ask us your question
Running a system we have not mentioned? Ask us your question and name it. Our AI engineering team works the question through with our technical founders and replies with how Unloq would reach it.
Need this in writing for your review?
Request the security pack and tell us what your reviewer needs.
