ShareOur client, Blackfish had already proven that AI could add significant value to its media intelligence platform. The prototype they had allowed users to ask natural-language questions about media coverage, campaigns and sentiment. Early results demonstrated strong potential, but like many successful proofs of concept, it raised a much bigger question whether it was actually ready for production? Although the AI prototype worked, the organisation lacked confidence in several critical areas before making a larger investment.
In this article, we will share a case study where we had two jobs: make an AI Readiness Assessment to see whether that’s what’s needed for the AI PoC to become a trusted system. Afterwards, we created the enterprise AI assistant, which became the product.
Phase one: AI Readiness Assessment
Before writing additional production code, Peruzzi conducted a comprehensive AI Readiness Assessment. Rather than evaluating the AI model alone, the review focused on everything required to operate the solution in production. The prototype had demonstrated technical feasibility, but the business still needed evidence that it could become a secure, scalable and supportable Azure service.
Questions remained around:
- scalability under production workloads
- architecture and workload separation
- grounding and source traceability
- deployment and operational ownership
- monitoring and recovery
- identity and access controls
- governance and long-term support
Our AI Readiness Assessment examined the application and Azure architecture, data flows, retrieval patterns, workload scaling, deployment process, observability, failure and risk management and operational ownership. If you’d like to learn more about our AI Readiness Assessment, click here.
The goal wasn’t to redesign the prototype, but to identify which components could remain unchanged, which required production engineering, and which governance controls were still missing. The assessment also established a phased implementation roadmap based on architecture, trust controls and production operations.
Key findings of the AI Readiness Assessment
The readiness assessment identified several important production considerations.
- Architecture
Background processing needed to be separated from user-facing services to improve scalability and resilience.
- Operational Readiness
Production deployment required monitoring, telemetry, repeatable deployments and failed-job recovery before enterprise rollout. Rather than presenting a list of technical issues, the assessment produced a prioritised roadmap that clearly distinguished immediate production blockers from improvements that could follow later.
Phase two: building the enterprise AI assistant
With a clear roadmap in place, Peruzzi implemented a ***Secure Enterprise AI Assistant ***designed specifically for Blackfish’s media intelligence platform. The objective was to give users secure, conversational access to trusted media intelligence while ensuring every answer remained grounded in verifiable data.
The assistant enables users to ask natural-language questions about:
- campaign performance
- media coverage
- sentiment
- individual articles
- client activity
Instead of manually searching thousands of records, users receive contextual answers backed by identifiable articles and analytics from the platform itself. The solution combines approved organisational content, Azure AI services and existing identity controls to deliver grounded responses with citations and secure access.
The solution: an AI assistant
Peruzzi delivered an Azure-native Enterprise AI Assistant using Retrieval-Augmented Generation (RAG).
The solution combines:
- Microsoft Azure
- Azure OpenAI
- Azure AI Search
- Hybrid keyword and vector search
- Platform analytics APIs
- Microsoft identity and access controls
This doesn’t mean it’s an OpenAI/ChatGPT wrapper. Instead of relying solely on the language model, the assistant retrieves relevant evidence from Blackfish’s media platform before generating responses. Every answer remains linked to the underlying articles or filtered datasets, allowing users to verify the information immediately. Hybrid search and analytics APIs ensure answers remain tied to available evidence while reducing unsupported responses.
The business impact of the AI assistant
The project delivered much more than an AI assistant: it provided Blackfish with a clear route from experimentation to production.
- Reduced implementation risk
The AI Readiness Assessment identified production gaps before significant implementation investment, allowing Blackfish to prioritise the changes that mattered most.
- Faster access to information
Users can now retrieve answers through natural-language questions rather than manually searching across articles, reports and campaign data.
- Trusted AI responses
Every answer is grounded in approved organisational content and linked back to identifiable source material, increasing confidence and transparency.
- Secure enterprise deployment
Identity-aware access controls ensure users only receive information they are already authorised to access.
- A scalable foundation
The resulting architecture provides a production-ready platform capable of supporting future AI capabilities rather than remaining a standalone prototype. The engagement delivered a documented readiness baseline, prioritised production actions, and a defined implementation route, while the knowledge assistant enabled natural-language exploration, grounded retrieval and traceable answers.
If you’d like to read another case study about how an AI prototype was deployed as a product, click here: From AI prototype to production in two weeks.
Key takeaways
Blackfish demonstrates that production success isn’t just about building an AI assistant, but about understanding whether the surrounding architecture, governance, operations and security are ready for enterprise use. By combining an AI Readiness Assessment with the implementation of a Secure Enterprise AI Assistant, Blackfish reduced delivery risk while creating a trusted, scalable AI capability built on Microsoft Azure.
If your organisation has already built an AI proof of concept, the next challenge isn’t creating another demo, you probably already have a demo: it’s understanding whether it’s ready for production. Peruzzi helps organisations assess AI readiness, identify production risks, and build secure Enterprise AI Assistants that deliver measurable business value on Microsoft Azure.