AI due diligence can run where the data room already has to stay. There are three ways to deploy it: in the vendor's shared cloud, in a single-tenant instance the vendor operates for you, or with the whole platform installed inside your firm's own AWS or Azure account or on your own servers. Which one you need depends on what the NDA allows and what your security team will sign off.
This article sets out the three models, what "your own cloud" means in practice, the questions to put to any vendor, and how Needl.ai deploys.
Why deployment decides the shortlist
A data room arrives under an NDA, and many NDAs restrict where the documents may be stored and processed. A platform can be accurate and still be unusable on a live deal if the files have to leave an environment the seller, the lender group or your own compliance team has approved. For that reason deployment is usually the first question to settle, before features. The same issue applies to general assistants, which is covered in Can You Upload a Data Room to ChatGPT or Claude?
The three deployment models
Shared cloud
- Where it runs: the vendor's managed cloud, with each customer's data kept separate, encrypted and access-controlled.
- Who operates it: the vendor.
- Fits: trials and fast starts, and deals where the NDA allows processing by an audited third party.
Single-tenant instance
- Where it runs: infrastructure reserved for one customer, with no compute or storage shared with anyone else.
- Who operates it: the vendor.
- Fits: firms that need isolation and do not want to run the software themselves.
Your own cloud account, or on premise
- Where it runs: inside your AWS or Azure account, or on your own servers.
- Who operates it: your team, with the vendor's engineers alongside.
- Fits: data that must stay inside your network, under your keys, your network rules and your logs.
What "your own cloud" means in practice
The phrase is used loosely, so it helps to pin it down. On Needl.ai, a deployment in your own AWS account means the following.
- The documents do not move. They are processed where they already sit, under the IAM, VPC and KMS controls your team already operates.
- Model calls stay inside the boundary. Models run through Amazon Bedrock inside your AWS account, so a question about the data room is not sent to an outside API.
- You hold the account, the keys, the network boundary and the audit trail. Infrastructure activity lands in your own logs.
- Permissions follow your source systems. A person sees answers only from files they could already open.
The same platform installs in an Azure account or on premise. The full list of controls is on the trust and security page, and the AWS architecture is described on Needl.ai on AWS.
Questions to put to any vendor
- Where are the documents stored and processed? Ask for the cloud, the region and the account owner.
- Where does model inference run? A platform can store files in your account and still send text to an outside model API. Ask which.
- Who holds the encryption keys?
- Is the deployment single-tenant, or is separation logical only?
- Is any customer data used to train a model, the vendor's or a provider's?
- What happens to the room after the deal? Ask for deletion on request, with written confirmation.
- Which independent audits exist, and can you read the reports? SOC 2 Type II and ISO 27001 are the usual baseline.
- Who patches and upgrades it, and how?
- How is it bought? A cloud marketplace listing can avoid onboarding a new vendor.
Where the main platforms stand on these points, as each one documents publicly, is compared in How to Choose an AI Due Diligence Platform in 2026.
How long a private deployment takes
Running in your own account adds a security review and an installation. It does not have to add quarters. Needl.ai's AWS architecture is a pattern AWS has already reviewed, so the templates and the reference design exist before your project starts. A typical engagement is a two-week proof of concept on two or three past deals, a one-month pilot inside your environment, and go-live one to two months after that. The stages are set out in how we work.
Firms that are not ready for that on day one can start in Needl.ai's cloud and move into their own environment later.
How Needl.ai deploys
- Three options, the same controls: Needl.ai's cloud, a dedicated single-tenant instance, or your own AWS or Azure account or servers.
- Audited: SOC 2 Type II report and ISO 27001 certification, GDPR compliant and CASA assessed, with an ISO 42001 audit in progress. Reports are available under NDA.
- Never trained on: your documents, questions and outputs are never used to train any model.
- In use at scale: 27,000 users at a Big Four firm work with Needl.ai behind the firm's firewall, and a leading credit rating agency runs it as a dedicated instance after its information security assessment.
- Bought through your cloud: Needl.ai is listed on AWS Marketplace and Microsoft Marketplace.
The product itself, from loading the room to the cited investment committee memo, is described on the AI due diligence page.
FAQs
Can AI due diligence run in a single-tenant deployment?
Yes. Needl.ai offers a dedicated single-tenant instance that it operates for you, with its own infrastructure and no compute or storage shared with other customers. It also installs inside your own AWS or Azure account or on premise.
Does the data room leave our network in a private deployment?
No. In your own AWS account the documents are processed where they sit, under your keys and network rules, and models run through Amazon Bedrock inside the same boundary.
Is our data used to train AI models?
Not on Needl.ai. Documents, questions and outputs are never used to train any model, in any deployment.
Can we start in the vendor's cloud and move later?
Yes. The same security controls apply in each option, so a firm can start in Needl.ai's cloud and move into its own environment when it is ready.
How is a private deployment bought?
Directly, or through AWS Marketplace or Microsoft Marketplace, where it arrives on the cloud bill your firm already receives.

Can You Upload a Data Room to ChatGPT or Claude?
Build or Buy AI Due Diligence? What an In-House Build Takes
How to Run an AI Due Diligence Proof of Concept