Questions
Frequently asked questions
Straight answers about how we work with Canadian organizations on grounded AI, retrieval pipelines, and communication workflows.
Before you write, scan the list below. If your question is not covered, book a discovery call or email [email protected].
Is CogniNarrative a course, a content-writing agency, or does AI replace my team?
No. We are an AI consultancy that helps organizations ground AI in their own knowledge — RAG assistants, documentation and support workflows with humans in the loop. We are not a training academy or a content-for-hire shop, and we are not datanarrative.pro. AI augments your team and needs human review; we do not guarantee savings, ROI or headcount outcomes.
What is the difference between ChatGPT and a grounded assistant you build?
Consumer ChatGPT draws on broad training data and cannot reliably cite your internal policies. A grounded assistant retrieves from sources you designate—wikis, ticket archives, PDF libraries—before the LLM drafts. Answers include traceable references your reviewers can check. We tune chunking, metadata filters, and refusal rules so the model stays inside approved material rather than improvising plausible-sounding policy.
Do you work with Microsoft Copilot?
Yes. Many clients already license Copilot and need help scoping which data sources are safe to index, how to extend Copilot with custom connectors, and where a dedicated RAG stack still makes sense. We are vendor-agnostic and also deploy on Azure OpenAI, Anthropic Claude, and private open-weight endpoints when requirements demand it.
How long does a pilot take?
Discovery sprints typically run two weeks. A first RAG or copilot proof on client documents often lands within three to four weeks after discovery, assuming source access and a named internal reviewer. Production rollout timelines depend on security review, integration complexity, and documentation cleanup—we quote ranges in the statement of work rather than promising a universal deadline.
How do you handle PIPEDA and privacy?
We treat personal information minimization as a design requirement. Contact form data is processed in Canada with explicit consent. Client pilots include data-flow documentation, retention settings, and guidance for your privacy impact assessment. We do not use client content to train public models unless contractually agreed and legally permitted.
What does a discovery call cost?
The initial discovery call is complimentary. Paid discovery sprints begin after mutual fit is confirmed and a scope letter is signed. Implementation fees vary by source volume, integration count, and enablement depth—we provide ranges after understanding your primary use case.
Do you support French or bilingual content?
Yes. Many Canadian deployments require English and French sources or responses. We configure retrieval and evaluation sets for both languages and align tone guides with your brand standards in each locale.
What industries do you serve?
Professional services, regulated retail, SaaS with dense help centres, nonprofits with compliance libraries, and public-adjacent agencies are common. The through-line is organizations where a wrong sentence carries real cost and source material already exists but is hard to search.
Can you integrate with our existing LLM vendor?
In most cases, yes. We design retrieval and evaluation layers that sit alongside Azure OpenAI, Anthropic APIs, or Copilot extensibility rather than forcing a rip-and-replace. The integration approach is documented during discovery so your IT team knows which credentials, webhooks, and logging endpoints are required before sprint one begins.
FAQ answers are general information only—not legal, financial, or medical advice. Specific recommendations require a signed engagement and access to your systems and policies.
Still have questions?
Book a discovery call and we will walk through your sources, tools, and constraints.
Book a discovery call