On RAG evolving into Agentic RAG. Q&A with Eudald Camprubí
Q1. How does the shift toward no-code RAG platforms change the accessibility landscape for organizations implementing generative AI, and what are the potential trade-offs between ease of use and customization flexibility that enterprises should consider?
The shift toward no-code Retrieval Augmented Generation platforms makes generative AI far more accessible by removing much of the technical complexity traditionally required to build and operate these systems. This means that organizations no longer need deep AI or engineering expertise to get started and can quickly move from basic use cases to more advanced applications through intuitive interfaces. In practice, however, the primary barrier to adoption is not ease of use, but trust: organizations need to be confident that AI outputs are accurate, grounded in reliable data, and suitable for production environments.
In our case, Progress Agentic RAG is designed with this reality in mind. It combines the accessibility of a no-code experience with agentic orchestration, validation, and governance, so organizations can scale safely as their needs grow. Teams can begin with simple configurations and progressively unlock advanced capabilities, such as multi-agent workflows, context management, and traceable retrieval, without sacrificing control or reliability. This approach balances usability and flexibility, enabling organizations to move from experimentation to production-grade AI while maintaining transparency, quality, and trust.
Q2. With RAG platforms becoming more democratized through marketplaces like AWS, what emerging use cases or industries do you see benefiting most from RAG-as-a-Service solutions, and what technical or organizational challenges might they still face in production deployments?
As RAG platforms become more democratized through marketplaces like AWS, the biggest beneficiaries of RAG-as-a-Service are industries and teams that rely heavily on large volumes of unstructured knowledge and need fast, reliable access to it. Customer support is a natural starting point and remains one of the most common entry use cases, enabling organizations to ground AI responses in product documentation, policies, and historical cases. Beyond that, internal-facing use cases are expanding quickly: legal, marketing, and HR teams increasingly use RAG systems to support internal workflows, answer domain-specific questions, and provide consistent information to employees across the organization. We also see strong adoption in use cases such as automated document classification, summarization, and multilingual search, where RAG pipelines help transform fragmented information into usable, context-aware outputs.
Despite this growing accessibility, production deployments still face meaningful technical and organizational challenges, especially as agentic capabilities begin to mature. The next phase of RAG adoption will require combining retrieval with AI agents at multiple stages of the pipeline: during ingestion, where agents enrich and structure incoming data, and during retrieval, where agents understand user intent and dynamically pull context from multiple internal and external sources. This added intelligence introduces complexity around orchestration, governance, and quality control. Progress Agentic RAG is designed to address these challenges by unifying RAG and agentic workflows within a single, governed platform. It enables organizations to start simply, while ensuring they are prepared for more advanced, agent-driven architectures—supporting data enrichment, intent-aware retrieval, and cross-source reasoning without sacrificing traceability, reliability, or production readiness.
Q3. The press release emphasizes “trustworthy” and “verifiable” AI insights from unstructured data. What architectural or methodological approaches are most effective for ensuring accuracy and reducing hallucinations in RAG systems, particularly when dealing with multilingual datasets?
Ensuring trustworthy and verifiable AI insights in RAG systems requires making quality and validation a core part of the architecture rather than an afterthought. In Progress Agentic RAG, this is achieved through Retrieval Evaluation Metrics Intelligence, an LLM-as-a-judge approach that evaluates every response across three dimensions: context relevance (how well retrieved content matches the query), answer relevance (how accurately the response reflects that context), and groundedness (how strictly the answer is supported by the sources). These metrics directly address the main causes of hallucinations, including semantic drift and multilingual ambiguity, while giving customers full transparency into output quality. By tracking these signals over time and flagging unanswered or low-quality responses, Progress Agentic RAG enables continuous optimization of retrieval strategies and data coverage, helping organizations maintain reliable, production-grade AI across languages and use cases.
Q4. As more vendors move RAG capabilities into SaaS offerings, how do you see the competitive landscape evolving between platform providers, and what differentiation factors beyond ease-of-use will matter most to enterprise customers making long-term technology decisions?
As RAG capabilities increasingly move into Software-as-a-Service offerings, the competitive landscape is shifting quickly from basic retrieval features toward agentic and reasoning-driven architectures. Traditional RAG is already evolving into agentic RAG, where AI agents can decompose complex user queries, retrieve context from multiple sources, and reason across them to produce coherent answers. These sources may include internal knowledge bases, enterprise systems like SharePoint or Salesforce, structured databases such as SQL, and even public information. Platforms that can orchestrate this multi-source retrieval with reasoning agents, rather than treating RAG as a single-pass lookup, will define the next phase of differentiation.
In this environment, ease of use remains important, but it is no longer sufficient for enterprise decision-making. What matters most are quality, traceability, security, and scalability. Enterprises need guarantees that every agent action is auditable, every answer is grounded in verifiable context, and strong guardrails exist around data access and generation. Progress Agentic RAG combines reasoning retrieval agents with built-in governance, quality controls, and full traceability, while allowing companies to start small and scale confidently. In a fast-moving and uncertain AI landscape, long-term winners will be platforms that balance advanced agentic capabilities with enterprise-grade trust, privacy, and operational reliability.
Q5. Looking at the broader AI infrastructure ecosystem, how should organizations balance between adopting integrated platforms like RAG-as-a-Service versus building custom solutions, and what criteria should guide this decision based on factors like data sovereignty, vendor lock-in, and specific business requirements?
When organizations consider whether to build custom RAG solutions or adopt integrated platforms like RAG-as-a-Service, the key recommendation is to avoid betting on building from scratch. The underlying technology is evolving too quickly, and building is only the first hurdle, maintaining, securing, and continuously evolving a RAG stack requires deep expertise, scarce talent, and sustained investment over many years. Unless an organization is prepared to fund and staff a dedicated platform team long term, the practical path is to adopt an end-to-end platform that delivers immediate value and evolves alongside the market.
Beyond build versus buy, enterprises should evaluate platforms on flexibility and long-term control. Avoiding lock-in to a single language model is critical: a modern RAG platform must allow teams to test and adopt new LLMs and embedding models as they emerge, optimizing for both quality and cost. Progress Agentic RAG supports this experimentation while enabling teams to start small, iterate quickly, and scale with confidence. Equally important are deployment options and data sovereignty, whether SaaS, hybrid, private cloud, on-prem, or region-specific hosting to meet GDPR and regulatory requirements. The right choice is a platform that balances innovation speed with governance, portability, and trust, rather than a custom system that risks falling behind the pace of the ecosystem.
Resources
…………………………………………
Eudald Camprubí, Software Fellow, Progress Software
Eudald Camprubí is a Software Fellow at Progress, where he leads innovation efforts around Agentic Retrieval-Augmented Generation (Agentic RAG) and next-generation intelligent application architectures. His work focuses on designing and advancing AI-driven systems that combine reasoning, autonomy, and enterprise-grade reliability.
With deep expertise in distributed systems, data engineering, and applied AI, Eudald has played a central role in shaping Progress’ approach to agentic workflows and RAG-based intelligence patterns. He works closely with engineering teams and customers to develop practical, scalable solutions that bring real-world automation and intelligence to enterprise applications.
Eudald is a frequent contributor to technical discussions on modern AI patterns, system design, and advanced RAG architectures, helping developers and organizations understand the tools and strategies required to build secure, production-ready agentic systems.
Sponsored by Progress Software