Generic AI gives generic answers. Your business deserves better. RAG Training lets you upload your company's documents, processes, and knowledge — so every AI agent you build becomes a specialist in how YOUR business actually operates.
Upload files through a simple web form. Or use the CLI to ingest entire folders with one command. Either way, your agents go from "generally helpful" to "genuinely transformative" in minutes.
You ask an AI tool a question about your business, and it gives you… a generic answer. A textbook answer. An answer that's technically correct but completely useless because it doesn't know:
It doesn't know your pricing, your features, your competitive differentiators, or what you actually sell.
It doesn't know your onboarding flow, your support escalation path, your sales qualification criteria, or how your team actually gets work done.
It doesn't know your ICP, your verticals, your use cases, your common objections, or what your customers actually care about.
It doesn't know your infrastructure patterns, your API conventions, your database schemas, or your deployment preferences.
It doesn't know how your brand communicates, your tone guidelines, your terminology, or the specific language your industry uses.
So you spend the first 5 minutes of every conversation re-explaining context. And the output still doesn't quite sound like it came from someone who works at your company.
That's the gap RAG Training closes.
RAG stands for Retrieval-Augmented Generation. Here's what that means without the jargon:
Documents, PDFs, process docs, playbooks, product specs, code files, API documentation — whatever your agents need to know.
Your files are parsed, chunked, and indexed so agents can search through them intelligently — understanding the meaning and context of your content, not just keyword matching.
Before generating any response, the agent searches its RAG memory for relevant information from YOUR documents. Then it uses that specific, retrieved knowledge to inform its output.
An agent that doesn't just sound smart in general — it sounds smart about YOUR business. It references your actual processes. It uses your actual data. It gives answers your team would give, because it's learned from the same knowledge your team uses.
| Question | WITHOUT RAG Training | WITH RAG Training |
|---|---|---|
| "How do we handle enterprise refunds?" | Generic refund process advice from the internet | Your exact refund policy, pulled from your customer service SOP, with the specific approval workflow and escalation path your team actually uses |
| "Write a project spec for our API" | Generic API spec template | A spec that follows your coding conventions, references your actual infrastructure patterns, uses your naming conventions, and aligns with your tech stack |
| "Qualify this inbound lead" | Generic qualification framework | Qualification based on YOUR ICP criteria, YOUR deal stages, YOUR pricing tiers, and YOUR competitive positioning |
| "Draft a response to this customer complaint" | Polite but generic customer service response | A response using your brand voice, referencing the specific product, following your escalation guidelines, and offering the remedies your policy allows |
No technical skills needed
Open the CEO.ai app and navigate to your agent's profile.
Click "Add Knowledge" and you'll see a simple upload form.
Select your files and upload. That's it.
Your agent's RAG memory updates immediately. The next time the agent works on a task, it uses its new knowledge.
.txt, .md, .pdf, .json, .yaml, .csv, and moreScreenshot: Web form interface
with drag-and-drop zone & agent selector
This is the path most CEOs and non-technical team members use. No terminal. No commands. No code.
For developers and power users
For technical team members who want speed, automation, and the ability to handle large knowledge bases efficiently.
# Add a single file to an agent's knowledge
ceo addRag ./docs/pricing-guide-2025.pdf
# Add an entire folder recursively
ceo addRagDirectory ./documentation/
# Works with any text-based file format
ceo addRag ./knowledge-base/product-specs.md
ceo addRag ./data/api-patterns.json
ceo addRagDirectory ./src/.txt, .md, .pdf, .json, .py, .js, .ts, and morenpm install -g @ceo-ai/cli
This is the path developers use when they have large document libraries, codebases, or knowledge bases to ingest quickly.
The short answer: anything your agents need to know to do their job well. Here's a practical guide by use case:
RAG memory is per-agent, not per-account. This is an important design decision, and here's why it matters:
Your sales agent doesn't need to know your infrastructure patterns. Your architect agent doesn't need your customer service SOPs. Your support bot doesn't need your competitive battlecards.
Agents only search through knowledge relevant to their role. No noise. No confusion.
Smaller, targeted knowledge bases mean more precise answers than one bloated knowledge dump.
More targeted knowledge means faster retrieval and more relevant results.
Sensitive information stays with the agents that need it. Your HR agent knows compensation data. Your customer-facing agent doesn't.
Your Agent Roster:
├── sales-qualifier
│ └── RAG Knowledge: ICP docs, pricing, battlecards, objection handling
├── support-specialist
│ └── RAG Knowledge: Product docs, FAQs, SOPs, troubleshooting guides
├── terraform-architect
│ └── RAG Knowledge: AWS patterns, Terraform modules, infra conventions
├── content-writer
│ └── RAG Knowledge: Brand voice guide, style guide, past blog posts
└── onboarding-assistant
└── RAG Knowledge: Employee handbook, onboarding checklist, tool guidesEach agent is a specialist with exactly the knowledge it needs. Nothing more, nothing less.
Documents get updated. Processes evolve. New products launch. Pricing changes. Your AI agents need to reflect current reality, not last quarter's reality.
# Add updated knowledge
ceo addRag ./docs/product-guide-v4.pdf
# Or re-upload an entire folder
ceo addRagDirectory ./docs/current/Best practice
Set a quarterly calendar reminder to review each agent's knowledge base. Are the docs current? Has anything changed? A 15-minute review every 90 days keeps your agents sharp.
On SMB and Enterprise plans, your monthly check-ins with our team include RAG knowledge review. We'll help you identify what needs updating and make sure your agents always reflect your current processes.
Here's where RAG Training gets really powerful — when you combine it with the CEO Agent's rating system.
The CEO Agent assigns a project to an architect and sub-agents
The agents produce output using their current RAG knowledge
You review and rate the results — total project, architect performance, and each sub-agent's work
You identify knowledge gaps — "the architect didn't know about our API rate limiting patterns"
You update the RAG knowledge — upload the missing documentation
Next time, the output is better — because the agent now has the knowledge it was missing
"The CEO Agent's first pass was about 90% of the way there. When we reviewed the architect's output, we identified specific gaps in its knowledge about our API integration patterns. We updated the architect's RAG memory with more detailed documentation. The second pass — using the same architect with updated knowledge — came out nearly perfect."
"Total time: ~2 hours. And the architect is now permanently better for future projects."
Every RAG update doesn't just improve the current project — it improves every future project that agent works on. Your agents accumulate institutional knowledge, just like your best employees do. Except they never forget, never leave, and never need to be retrained from scratch.
We know what you're thinking: "If I upload my company's documents, who else can see them?" The answer: nobody.
No other CEO.ai customer can see, access, or search through your agents' knowledge bases.
If you opt an agent into the marketplace, other customers can use the agent — but they cannot see the underlying RAG data. The source files are never exposed.
Even within your own account, each agent's knowledge base is separate. An agent can only access its own RAG memory.
You decide what to upload, which agents get which knowledge, and when to remove it. Full control, always.
On Enterprise plans, we can discuss specific compliance requirements, data residency, encryption standards, and access controls tailored to your organization.
This is where the guided setup earns its keep. On every plan, our team helps you identify which agents to create, what knowledge each needs, and how to organize it for maximum impact.
We help you set up RAG memory for your first few agents as part of your guided onboarding.
We help you build comprehensive knowledge bases for agents across up to 4 use cases — and review them during monthly check-ins.
Full knowledge audit. Build and maintain knowledge bases across your entire agent roster. Continuously optimize based on performance data.
You don't need to be an AI expert to train your agents well. You just need to know your business — and we'll help you turn that knowledge into agent intelligence.
The difference between AI that's "kind of helpful" and AI that transforms your operations is one thing: whether it knows your business. RAG Training makes sure it does.
We'll help you identify what to upload, which agents need it, and how to organize it for maximum impact — on every plan.
Guided RAG training setup included on every plan. Most agents are trained and working within the first week.