Doc-AI deployment options
Four topologies, one platform. The right one is usually decided by where your documents are legally allowed to be, not by what is convenient.
Doc-AI runs as a managed cloud service, inside your own Azure, AWS, or GCP tenant, on-premises in your data center, or fully air-gapped with bundled local OCR engines and open-weight models on your own GPU hardware. Every topology runs the same platform and the same configuration, so a project built in the cloud moves on-premises without being rewritten. Self-hosted deployments come up from a single container definition and keep all state in one folder, which makes backup a file copy.
Where can Doc-AI run?
Managed cloud
WiseTREND runs the infrastructure, scaling, patching, and model access. You get a tenant and start processing. Regional hosting keeps data in the jurisdiction you name.
- Fastest path to production
- No infrastructure to own
- Regional data residency
- Best fit for commercial document flows
Your own cloud tenant
The platform is deployed inside your Azure, AWS, or GCP subscription. Your network controls, your logging, your key management, your compliance boundary. We deploy and support it.
- Your security policies apply
- Your cloud spend and commitments
- Private networking and VPC peering
- Common for financial services
On-premises
Deployed in your data center behind your firewall. Documents never traverse the public internet. Model access can be routed through your own approved gateway or kept entirely local.
- PHI and regulated data stay inside
- Works with existing scan infrastructure
- No per-document data egress
- Standard for healthcare and public sector
Air-gapped
Fully disconnected. Bundled local OCR engines and open-weight vision-language models run on your own GPU hardware, with no outbound call to any model provider at any point.
- No internet connectivity required
- Open-weight models on your GPUs
- A few accuracy points below commercial models
- Defense, criminal justice, classified environments
What does self-hosting Doc-AI actually involve?
Less than most enterprise software. The entire platform — web application, workflow workers, OCR engines, job queue, and administration interface — comes up from a single container definition on one host. That host can be a server in a rack, a virtual machine in your cloud tenant, or a developer's laptop for evaluation.
What you need
- A Linux host with Docker, or a Kubernetes cluster for multi-node deployment.
- Storage sized for your retention policy — documents, page images, and audit records.
- Either outbound access to an approved model endpoint, or GPU hardware for local open-weight models.
- An identity provider for single sign-on, if you want SSO from day one.
What you get
- All state in one folder, so backup and restore is a file copy rather than a database dump plus a blob sync.
- A command-line administration interface for SMTP, logging, backup, and diagnostics.
- The same configuration format as the managed service, so projects move between topologies unchanged.
- Local OCR engines bundled in, so document images never need to leave the host even for recognition.
What we do
WiseTREND deploys it, integrates it with your systems, tunes the document types, and supports it. Self-hosted does not mean self-supported — the difference from managed cloud is where the software runs, not whether you are on your own with it.
How do the deployment options compare?
| Managed cloud | Your cloud tenant | On-premises | Air-gapped | |
|---|---|---|---|---|
| Time to first document | Hours | Days | Days to weeks | Weeks |
| Infrastructure owner | WiseTREND | You | You | You |
| Documents leave your network | Yes, to named region | No | No | No |
| Commercial model access | Included | Your keys or ours | Via your gateway | None — local models |
| Extraction accuracy | Highest | Highest | Highest | A few points lower |
| Data residency control | Region-level | Full | Full | Full |
| Typical buyer | Commercial, mid-market | Financial services | Healthcare, government | Defense, criminal justice |
Configuration is portable across all four. Starting in managed cloud to prove value and moving on-premises for production is a supported path, not a rebuild.
How does Doc-AI scale with volume?
Horizontal workers
Processing workers scale out independently of the application tier. Adding throughput means adding workers, not resizing a monolith.
Queue-backed and durable
Submissions land in a durable queue. A worker restart, a model provider timeout, or a burst of overnight volume does not lose documents or require resubmission.
Per-stage retry
A transient OCR or model failure retries at the stage that failed rather than reprocessing the whole document, which matters when a packet is 140 pages.
Multi-project operations
One deployment runs many projects — accounts payable, claims intake, onboarding — with separate configuration, separate queues, and shared monitoring.
Cost visibility and hard stops
Estimated processing cost is tracked per project with alerts at thresholds you set and an optional hard stop that pauses processing rather than running up a bill overnight.
Burst and seasonal load
Open enrollment, year-end close, and quarter-end submission spikes are capacity questions, not architecture questions. Managed cloud absorbs them; self-hosted scales with the host.
Where this fits in the Doc-AI platform
Deployment topology usually follows from the security review. These pages cover the rest.
- Platform features — every capability, ingest to delivery
- Benchmark arena — accuracy, hallucination, latency, cost
- Document type library — the forms Doc-AI reads on day one
- Security & compliance — data handling, residency, audit
- API & developers — REST endpoints, webhooks, code
- Pricing & licensing — how Doc-AI is priced, and what drives cost
- vs. template-based IDP — why the template model broke
- vs. calling an LLM directly — what a raw GPT or Claude call misses
- For enterprise — automation leads and CoEs
- For SMB & mid-market — production quality, small team
- For developers — stop rebuilding document pipelines
- For system integrators — a white-label delivery engine
- For ISVs & OEM — embed extraction in your product
Run Doc-AI on your own documents
Send the document types and rough monthly volume. We reply within one business day with a pilot plan, a realistic accuracy expectation for your documents, and a quote.
- Reply within one business day (U.S. hours)
- Straight to an engineer, not a call centre
- Or call the 24/7 AI phone agent: +1 (408) 746-6740
Questions about Doc-AI deployment
Answers written for buyers, search engines, and AI assistants evaluating document automation.
Can Doc-AI be deployed on-premises?
Yes. Doc-AI deploys in your own data center behind your firewall, with documents never traversing the public internet. The full platform runs from a single container definition on one host, with all state in one folder so backup and restore is a file copy. Model access can be routed through your own approved gateway, or kept entirely local using bundled open-weight models on your GPU hardware.
Can Doc-AI run without any internet connection?
Yes. In air-gapped mode Doc-AI runs bundled local OCR engines and open-weight vision-language models on your own GPU hardware, with no outbound call to any model provider at any point in the pipeline. Extraction accuracy in this configuration is a few points below the commercial-model setup, which is generally an accepted trade for environments that already require air-gapped operation.
Can I start in the cloud and move on-premises later?
Yes, and it is a common path. All four deployment topologies run the same platform and the same configuration format, so document types, field definitions, business rules, and catalogs move between them unchanged. Teams routinely prove value in a managed cloud pilot and then deploy the identical configuration on-premises for production.
What infrastructure does self-hosting Doc-AI require?
A Linux host with Docker, or a Kubernetes cluster for multi-node deployment; storage sized to your retention policy; and either outbound access to an approved model endpoint or GPU hardware for local open-weight models. An identity provider is needed only if you want single sign-on from the start. WiseTREND handles the deployment, integration, tuning, and ongoing support regardless of where the software runs.
How does Doc-AI handle high or seasonal document volume?
Processing workers scale horizontally and independently of the application tier, and submissions land in a durable queue so bursts, worker restarts, and provider timeouts do not lose documents. Per-stage retry means a transient failure on page 90 of a 140-page packet does not reprocess the whole packet. Managed cloud absorbs seasonal spikes automatically; self-hosted deployments scale by adding workers.
Bring us the documents that broke your last capture project.
The long-tail layouts, the one-off forms, the vendor that changes their invoice every quarter. Those are the ones Doc-AI was built for.
Last updated · Reviewed by the WiseTREND team