Managed AI
Your AI platform in your own AWS or Microsoft cloud, kept safe and in budget. We run the cost controls, policies, security, guardrails and identity every month.
Your AI platform in your own AWS or Microsoft cloud, kept safe and in budget. We run the cost controls, policies, security, guardrails and identity every month.
Problem
You have pilots on Bedrock or Foundry. They work. Then the pilot team moves on, the invoice grows, and nobody is sure whether customer data reaches the model or who still has access. Managed AI builds the platform your AI needs and runs it every month, so the pilots can go to production and stay safe there.
Approach
A short project builds an AI landing zone in your own cloud and deploys AWS AgentCore or Microsoft Foundry on top. Then Managed AI looks after it every month. The agents that run on it are covered by Managed Agents.
Managed service
Your governance rules become technical controls, built with our tested Terraform modules in your own code repository. Then we keep them in place month after month.
Managed service
Managed AI
Run
Managed AI
Flat monthly fee, set in your proposal
Outcomes
The monthly fee is set in your proposal. Model usage and cloud costs are billed at cost; you can buy them through us under AWS and Microsoft CSP, so they show in the same report. Agents that run on the platform are priced per agent in Managed Agents.
We run the platform configuration, cost reports, policies, guardrail changes, access reviews and security incidents. You approve changes and own each use case. AWS and Microsoft keep the models available.
We don't replace it. Platform alerts go to your SOC or SIEM, and we handle the work that is specific to AI.
A data processing agreement, listed sub-processors and processing in the EU. Your data never trains models. We tell you where AI agents work in our delivery.
How It Goes
The platform takes two days. Then we run it every month, and your agents move onto it one by one.
An AI landing zone in your own cloud, built from ready-made parts, with AWS AgentCore or Microsoft Foundry deployed on top and your rules switched on. Without a governance model, our standard policies apply.
Identity, network, guardrails, policies and budgets in Terraform, in your own repository.
AgentCore or Foundry deployed and secured, ready for your first agents.
Access to your cloud account, your existing code repositories, a connection point to your SOC and a named owner for each use case.
AI Landing ZoneWe keep the controls in place and the costs visible.
A monthly report per use case, anomalies explained, and proposals such as a cheaper model or caching.
New tools and agents checked against your rules, and shadow AI the platform detects reported back to you.
Configuration drift fixed, access reviewed, security incidents handled with your SOC.
Hits reviewed, filters tuned, and every change tested before it goes live.
Your agents move onto the platform one by one, each with its own test set. Managed Agents watches them and manages every version.
Observability, version management and lifecycle for every agent, priced per agent.
Managed AgentsTwo Clouds
| Area | AWS Bedrock + AgentCore | Microsoft Foundry |
|---|---|---|
| Access and data | ||
| Identity | IAM roles per application, SCPs restricting models and regions, AgentCore Identity for agents | Entra ID and managed identities, no API keys, Entra Agent ID for agents |
| Network and data | PrivateLink endpoints, EU inference profiles, KMS keys | Private endpoints, no public access, EU Data Zone deployments, customer-managed keys |
| Controls | ||
| Guardrails | Bedrock Guardrails: content filters, denied topics, personal data, grounding checks | Content Safety, Prompt Shields, groundedness detection |
| Policies | SCPs, AWS Config rules, AgentCore Policy for agent tool calls | Azure Policy for AI, Defender for Cloud, Purview DSPM for AI |
| Gateway | AgentCore Gateway for MCP tools | API Management AI Gateway: token limits, caching, token metrics |
| Running it | ||
| Observability | Model invocation logging, CloudTrail, AgentCore Observability | Application Insights, Foundry tracing and evaluations, Log Analytics |
| Costs | Inference profiles and cost allocation tags, FOCUS exports, budgets, anomaly detection | Deployments per use case, API Management quotas, FOCUS exports, budgets |
| Security monitoring | GuardDuty and Security Hub, logs to your SIEM | Defender for AI alerts to Sentinel or your SOC |
| Agent platform | AgentCore Runtime, Memory and Gateway, deployed in the project | Foundry Agent Service, deployed in the project |
You get the same service and the same reports on both clouds.
FAQ
Field Notes
What we learn running AI platforms in production.
Data
6 min read
Building a data platform on Microsoft Fabric should be straightforward. So why does it still take months? Because the hard part was never the technology.
AI
2 min read
2NS conducted a security assessment of the twoAI platform developed by Cloud2. The goal of the testing was to verify the software’s level of security and to obtain external expert validation for Cl...
Platform Engineering
6 min read
This blog is part of our Platform Engineering series. For a deeper understanding of our perspective on platform engineering, check out our other posts:
Two days to a secured AI platform in your own cloud. Then we keep it safe and in budget, month after month.
