radianit.com

What It Is

Radian IT positions itself as a “Production AI engineering” provider. Its core offering is not a ready-to-sign-up AI SaaS product, but helping enterprises move AI workflows from prototypes, demos, or AI-assisted manual processes into controlled, routine production use. It emphasizes starting from workflow boundaries: first clarifying what the business is doing, which systems need to be integrated, and what evidence release owners need to see—then deciding on models and tools.

Core Capabilities

Its key capabilities include production AI workflow development, evaluation harness design, runtime evidence schemas, release gates, rollback, and runbook delivery. For Agent/RAG scenarios, Radian IT provides memory and retrieval architecture reviews, focusing on inconsistent behavior, debugging difficulty, over-reliance on context, and scaling risks. It also offers managed deployment of OpenClaw and Hermes, as well as agentic governance, observability, and evaluation audits. The website does not describe proprietary models or specific model capabilities; instead, it explicitly states that model and tool choices are subordinate to architecture, integration, and releasable evidence.

Pricing and Delivery

Pricing is project-based with fixed scope and fixed timelines. Core workflow build projects run for 6 weeks and start at GBP 25,000; memory and retrieval architecture reviews run for 2 weeks and start at GBP 8,000; OpenClaw/Hermes deployment takes 3 weeks and starts at GBP 18,000; governance and evaluation audits take 2 weeks and start at GBP 10,000. No free trial or free tier is shown. Deliverables are engineering-oriented, including services deployed in the customer environment, boundary maps, evaluation suites, release rules, handover documentation, and runbooks.

Pros and Cons

The main advantage is its pragmatic methodology: it pays close attention to integration, approvals, logs, evidence, release accountability, and operability, making it suitable for teams moving from “the demo works” to “production is controlled.” It also tends to place deliverables inside the customer’s codebase, cloud environment, and daily operations system, making handover easier. The downsides are that it is not a low-barrier tool and pricing is relatively high. Public information also lacks specific customers, quantified outcomes, supported model lists, cloud platform lists, and Chinese-language service details.

Who It’s For and Access from China

It is better suited to teams in government public services, telecom and media, consumer platforms, and similar sectors that require approvals, audits, stability, and highly visible operations. It is less suitable for individual developers, early-stage experimentation teams, or users simply looking for a general-purpose chatbot. Access from mainland China, payment methods, and Chinese-language support are not disclosed, so china_access can only be rated as unknown. If local procurement or network compliance requirements are important, alternatives may include AI professional services from cloud vendors, domestic LLMOps/RAG implementation partners, or internal platform engineering teams.