Enterprise AI knowledge assistants: RAG engineering with access control
Netbase engineers retrieval-augmented generation (RAG) over company knowledge: the ingestion, search, permission checks, cited answers and evaluation that let a language model answer from your own documents and systems. It is a growth capability with access control and evaluation designed in from the first prototype, and Netbase has delivered a RAG knowledge assistant for a client that is not named.
Book a solution review View relevant workReviewed by David (CEO) · Updated 17 Sep 2026
Planning to let a model answer from your company's documents? Book a solution review.
What enterprise knowledge RAG includes
A language model on its own knows nothing about your contracts, procedures or product data, and it will still produce a fluent answer. RAG closes that gap: the system first retrieves the passages that answer a question, then asks the model to answer only from them and to cite each one. The quality of the result depends far more on the retrieval pipeline than on the model. This service is that pipeline. It belongs to our AI & Data family; the packaged employee-facing product built on it is the AI knowledge assistant.
Included
- Source inventory, ownership and freshness rules
- Ingestion and chunking for documents, wikis and system records
- Search indexes combining keyword and semantic retrieval
- Permission-aware retrieval that mirrors source access rights
- Answer generation with citations and refusal rules
- Evaluation sets for retrieval and answer quality
- Monitoring of questions, gaps and feedback
Not included
- Cleaning up years of outdated documents on your behalf
- A public chatbot answering from the open web
- Answers without sources
- Legal or medical advice generated by the model
- Guaranteed accuracy figures
When to use it, and when another route fits
RAG is worth engineering when people lose time finding answers that already exist in writing.
Good fit
- Policies, procedures or product data spread across many tools
- Different people may see different documents
- Answers must point to a source someone can check
- Document owners exist who can fix gaps
Another route fits better
- A few pages that a good search box would cover
- All content is public: a site search may be enough
- The answer needs a calculation over live figures: use reporting
- Nobody maintains the sources
If you want a ready solution for employees rather than an engineering component, start from the AI knowledge assistant. Every route is in the services directory.
Outcomes and buyer jobs
Knowledge, operations and IT leaders look for RAG development to:
Every answer links to the passages it used.
Retrieval applies the same access rights as the source systems.
Retrieval and answer quality are scored on a test set of real questions.
Unanswered questions show document owners what is missing.
The pipeline is built so the model can change without re-indexing everything.
Capability modules and deliverables
The systems, owners, formats, access rules and update frequency of each source
Connectors, parsing, chunking and metadata, with scheduled refresh
Keyword and vector indexes, ranking and permission filters
An API that returns cited answers or declines when sources are missing
Real questions with expected sources, scored on every change
Logs of questions, sources used, feedback and unanswered topics
Delivery process, team and governance
RAG engineering follows our six-step delivery lifecycle:
-
Discovery and strategic alignment
We collect real questions, list the sources that should answer them and agree who owns each source.
-
Team assembly and architecture planning
An architect designs ingestion, the index, the permission model and where data is stored.
-
Agile execution with outcome-based milestones
One source and one user group come first, measured on the evaluation set.
-
Modular and productized components
Connectors and chat interfaces reuse proven modules where they fit.
-
Training, rollout and optimization
Users learn to check citations and give feedback; access widens in stages.
-
Ongoing support and co-building
New sources are added and chunking and ranking are tuned as the question log grows.
Teams of 3 to 30 people typically start one to two weeks after discovery. Delivery is Agile and remote-first, in English, with weekly reviews, KPI dashboards and a named account and project manager; sources connect API-first, so a new system joins without re-indexing everything. Security is central because a RAG system concentrates company knowledge in one place: role-based access control, MFA for admin dashboards, TLS in transit and AES at rest, secure code review and vulnerability scanning; NDAs and DPAs are available on request. Where personal data is indexed, we follow GDPR alignment, HIPAA-aligned methods and CCPA practices.
The OWASP Top 10 for LLM Applications (2025) lists vector and embedding weaknesses as LLM08, alongside prompt injection and sensitive information disclosure. We test for cross-user leakage through the index and for instructions hidden inside documents. NIST AI 600-1, the Generative AI Profile, names confabulation among generative AI risks, which is why answers must cite their sources.
AI in this service
AI changes how people reach knowledge: instead of guessing search terms, they ask a question and check the cited answer. That only works with disciplined engineering. Our practice versions prompts and chunking settings, runs an evaluation set on every change, keeps a person responsible for each source, and monitors the cost of every answer. AI-assisted coding and test generation support the build, and engineers review every change. Netbase applies the ISO/IEC 42001 AI management system framework to its own AI delivery practice.
-
In delivered work
RAG knowledge assistant
Delivered for a client that is not named; the record describes the kind of system and Netbase's role only.
-
Available capability
NLP and generative AI beyond knowledge retrieval
Capability areas Netbase offers; not yet tied to a published knowledge project.
Engagement models and commercial variables
Most Netbase projects are delivered on fixed-price contracts agreed after discovery; for RAG, a fixed-price pilot over one source and one user group, measured on an agreed evaluation set, is the natural start. Milestone-based, monthly team retainer and KPI-linked terms are also offered. We do not publish rate cards.
What moves effort and cost: the number and formats of sources, how complex the access rules are, document volume and update frequency, languages, hosting and data residency requirements, and the expected number of questions.
Technology as an implementation choice
Netbase is model-agnostic: we work with models from OpenAI, Anthropic (Claude), Google (Gemini) and Meta (Llama), among other commercial and open-weight models and embedding models, and choose them per project by retrieval quality on your test set, data residency and cost. We adopt new technology project by project, never as a vendor partnership. The search, vector and model components we compare are on our data and AI stack page, labelled by maturity.
Industry applications
Professional services. Firms run on precedents, templates, proposals and internal methods. RAG helps staff find the right prior document quickly, while client confidentiality is enforced by permission-aware retrieval.
Operations-heavy companies. Procedures, product specifications and supplier terms change often. Scheduled refresh and source ownership keep answers current, and unanswered questions show where documentation is missing.
Professional services: projects, clients and teams on one AI-assisted workspace
Agencies, consultancies and other service firms need projects, proposals, invoices and teams on one workspace instead of five tools, with AI drafting plans and finding answers in past work for people to check. Netbase builds that kind of workspace, and the proof is our own: Cloodo Workspace, the AI-powered work SaaS that Netbase's Cloodo Business Division designed, built and operates.
Learn More
Proof: platforms that organise company knowledge
Evidence maturity: enterprise knowledge RAG is a growth capability. One delivered RAG project is recorded anonymously; the other records are platform proof: products Netbase builds and runs that structure company information and access rights.
Delivered AI: RAG knowledge assistant (anonymised client). Netbase delivered a retrieval-augmented knowledge assistant for a client that is not named, as AI consultant, engineer and integrator. The record publishes no client name, figures or results. See the anonymised record.
Platform proof: Cloodo (Netbase Business Division). An AI-powered digital workplace for company profiles, services, projects and team collaboration that connects internal staff and outsourced specialists in one hybrid workspace, with CRM, HRM, Cloud ERP and AI modules. It shows the role and access model a knowledge system has to respect. See the Cloodo record.
Platform proof: Printcart (Netbase Business Division). A web-to-print and print-on-demand platform that turns online orders into print-ready files and routed fulfilment, with large product and clipart catalogues to organise. See the Printcart record.
More projects are in our work.
Printcart is Netbase's own web-to-print SaaS, and this record shows what we built.
Keep Reading
Cloodo Workspace is the work-management SaaS Netbase built and operates for service businesses.
Keep Reading
Netbase delivered a retrieval-augmented generation (RAG) knowledge assistant for a client, in a role that combined AI consulting, engineering and integration.
Keep ReadingBuyer FAQ
They should not, and we design against it. Retrieval filters results by the user's rights in the source system before the model sees anything.
The system says it cannot find a source and logs the question for the document owner.
With an evaluation set of real questions and their expected sources, scored on every change to data, prompts or models.
Hosting, providers and data residency are agreed in discovery and written into the DPA.
This service is the engineering layer. The assistant is a packaged solution for employees built on it.
Indexing everything at once. Outdated and conflicting documents produce confident wrong answers, so sources are added with owners, one at a time.
Related solutions
AI knowledge assistant: answers from company knowledge, with sources and access control
An AI knowledge assistant is an internal tool that answers employees' questions from your company's own documents and systems, cites the source behind every answer and shows each person only what they are allowed to see. Netbase offers it as a growth capability, designed with retrieval, evaluation and human review from the first pilot, and has delivered one for a client that is not named.
Learn More
Service owner and next step
This service is owned by David, Netbase's Chairman, founder and CEO, and maintained by the Netbase Editorial Team. Bring twenty real questions and the sources that should answer them: book a solution review, or view relevant work first.
Discuss a project
Netbase JSC helps organizations design, build, modernize, and operate digital products and AI-enabled business systems.+84 937 869 689
91 Nguyen Chi Thanh, Dong Da, Hanoi, Vietnam
Get in touch
Tell us what you want to build, modernize, or operate.