Give an AI assistant approved, organized business information—not unrestricted access to everything
A useful AI assistant needs a focused source of truth covering what your business offers, how it works, what customers commonly ask, and which questions require a person.
Start by collecting the business information you would trust a trained employee to use: public company facts, product and service details, current policies, common questions, and approved procedures. Organize it by topic, remove contradictions, label time-sensitive material, and specify when the assistant should refer someone to a person. OceSha AI provides a self-service platform for bringing in sources such as documents, websites, text, audio, and video, without requiring you to begin with technical configuration.
- Give the assistant a curated source of truth rather than every file your business owns.
- Prioritize accurate answers to real customer questions before adding large amounts of background material.
- Separate public information from private, internal, and account-specific information.
- Test ambiguous, outdated, and unsupported questions—not just easy questions with obvious answers.
- Assign an owner and update the knowledge whenever products, services, policies, or procedures change.
Start with the questions the assistant must answer
OceSha AI is the self-service creation platform of OceSha Ventures, and Lumi is its AI Concierge. But the first step is not choosing software or configuring an assistant. It is deciding what people need to know about your business. Begin with the questions customers, prospects, learners, or other visitors repeatedly ask before they understand an offering, make a decision, or take a next step.
Write those questions in the language people actually use. Group them into practical categories such as company information, products, services, eligibility, policies, purchasing, delivery, support, and next steps. Then write the answer you would want a capable team member to provide. If you are still deciding whether AI is approachable for your team, start with using AI without being a technical person rather than treating technical expertise as the entry requirement.
A business knowledge source is the approved collection of facts, explanations, policies, and procedures an AI assistant is expected to rely on when answering questions. It should define both what the assistant knows and where its knowledge ends.
This approach is different from using a general-purpose chatbot for open-ended work. A general tool can help draft or explore, while a business assistant needs authoritative information about your particular organization. If you are deciding between those approaches, examine ChatGPT versus a custom business assistant before assembling the content. The right standard is not whether the assistant sounds intelligent; it is whether its answers are grounded in information your business approves.
Build a small, authoritative source of truth first
Do not begin by uploading an entire shared drive. Volume does not solve uncertainty. A smaller collection of current, approved material is more useful than a large archive containing obsolete pricing, duplicate policies, unfinished drafts, and contradictory explanations. Start with the minimum body of knowledge required to answer your highest-value questions correctly.
- List the recurring questions people ask before they understand or choose your offering.
- Record one approved answer for each question, using direct language and defining any necessary business terms.
- Add the supporting facts: product or service descriptions, current policies, procedures, audience information, and clear next steps.
- Identify material that changes frequently, including availability, plans, promotions, policies, and operational details.
- Mark questions that require judgment, private account information, or human approval so they can be referred appropriately.
- Name a person responsible for approving the initial material and keeping it current.
For many businesses, this focused knowledge project is the simplest way to start using AI. It has a defined audience, visible inputs, and answers that can be reviewed. It also produces a reusable business asset: the same organized knowledge can support customer education, internal guidance, content planning, and future assistant improvements.
Do not combine current policies with old versions, publish draft answers as final guidance, or expect the assistant to resolve disagreements between source files. Settle conflicts before the information becomes part of the assistant’s working knowledge.
Choose information for authority, clarity, and scope
Each source should earn its place. Ask three questions: Is it authoritative? Is it current? Is it written clearly enough to answer a real question? A signed-off policy is stronger than an informal note. A current service description is stronger than an old campaign page. A direct answer is stronger than a document that merely hints at the answer across several paragraphs.
A business with expertise, products, or services could begin by organizing the questions people ask about those offerings. It could provide approved descriptions, relevant educational material, current procedures, and clear routes for further help. This is a stronger first project than attempting to automate every conversation at once. For more help choosing a contained use case, consider a realistic first AI project for a small business.
An assistant should not be expected to answer beyond the material it has been given. When a question depends on an individual situation, private records, a recent change, or judgment that has not been documented, route it to the appropriate person instead of filling the gap with a plausible-sounding answer.
Test the difficult questions before customers use it
Testing should prove more than whether the assistant can repeat an obvious fact. Ask the same question in several ways. Use incomplete questions, misspellings, vague wording, and terms your customers use instead of your internal terminology. Check whether the answer stays within the source material, reflects the current policy, and gives an appropriate next step.
- Direct question
- Confirms that a clearly documented fact can be found and explained.
- Rephrased question
- Checks whether different wording still leads to the correct information.
- Ambiguous question
- Reveals whether the assistant asks for clarification rather than choosing an unsupported interpretation.
- Outdated premise
- Tests whether an old product name, policy, or assumption produces a current response.
- Private-account question
- Confirms that general guidance is separated from information requiring authenticated access.
- Unsupported request
- Checks whether the assistant recognizes the boundary of its supplied knowledge.
Create a test sheet containing the question, expected answer, source, observed answer, and correction needed. That process turns subjective reactions into repeatable quality control. A dedicated guide to testing an AI assistant before customers see it can help you plan this stage in greater detail.
Test the experience where people will encounter it, not only in an isolated preview. Page context, visitor intent, and the route to human help all affect usefulness. If your next concern is deployment, review whether you can add an AI assistant to Wix, Squarespace, or WordPress after its knowledge and boundaries are ready.
How OceSha AI fits into the process
Once the knowledge has been selected and approved, the OceSha AI self-service creation platform provides a place to bring it into a broader creation workflow. Users can supply documents, websites, text, audio, video, and other existing material. OceSha AI can use that information to help create courses and content, publish and distribute content, educate learners, build a professional presence, engage an audience, and measure results.
The platform is designed for individuals and organizations with knowledge, expertise, products, services, or educational information they want to turn into useful content and learning experiences. Users create an account, configure their profile and professional information, and provide information that helps OceSha AI understand their knowledge and expertise. Recommendations can then draw on the supplied information so suggested content is relevant to the user’s business, interests, expertise, and existing material.
Lumi is OceSha’s AI Concierge. On public OceSha pages, it is trained on OceSha’s public information and answers from available public material about OceSha, including its products, services, capabilities, plans, and use cases. In an authenticated context, Lumi also understands the application’s pages, navigation, features, workflows, course capabilities, content tools, publishing processes, integrations, analytics, settings, and subscriptions. It provides informational and how-to guidance; for example, it can guide users through feature navigation, course creation, publishing, finding leads, viewing analytics, and choosing the appropriate platform area.
OceSha AI is the self-service platform of OceSha Ventures and its AI-first solutions. If you want to discuss whether your material and intended use fit the platform, contact the OceSha team with the questions you need the assistant to answer and the types of source material you already have.
Uploaded knowledge and other user-specific information can be used for personalization or generation without becoming public. Even so, classify information before providing it: keep public guidance, internal operating material, and private account data separate, and give each only the access appropriate to its purpose.
Treat maintenance and measurement as part of the knowledge system
A correct launch is temporary unless someone maintains it. Products change, services evolve, policies are revised, and old explanations remain in circulation. Keep a source register that records each topic, its authoritative source, its owner, its approval date, and the event that should trigger review. When something changes, replace the old source rather than simply adding another version beside it.
Use a regular review rhythm for stable material and immediate updates for consequential changes. The precise schedule matters less than clear ownership. The practical discipline is explained further in keeping an AI assistant current as the business changes. Updating the knowledge should be part of the same operational process that approves the underlying business change, not an afterthought discovered when someone receives an old answer.
Measurement should connect to the assistant’s assigned job. Review whether people receive accurate answers, whether unclear questions expose missing information, and whether interactions reveal recurring subjects that deserve better documentation or education. OceSha AI’s Growth Overview provides high-level information about audience and business activity, while knowledge and performance information can feed the next cycle of recommendations, content, education, and audience engagement.
Do not use activity alone as proof of value. A busy assistant can still be unhelpful if answers are incomplete or visitors cannot reach the right next step. Decide what success means before launch, then use measuring whether an AI assistant helps the business to build an evaluation around that objective.
Bring your priority questions and existing source types to a conversation about creating an organized, useful AI knowledge foundation.
Talk to OceShaFrequently asked questions
Should I rewrite all my business information specifically for AI?
Not necessarily. Start with existing authoritative material, then improve sections that are unclear, contradictory, outdated, or difficult to use as direct answers. Documents, websites, text, audio, video, and other existing material can be brought into OceSha AI.
What should I do when two source documents disagree?
Resolve the disagreement before using them as authoritative sources. Select the current answer, remove or archive the obsolete version, and record who owns future updates. An assistant should not be expected to decide which business policy is valid.
Should public and internal information be stored together?
Separate them by purpose and access. Public facts can support public answers, while internal procedures and private account information require tighter boundaries. Using private information for personalization or generation does not itself make that information public.
Can product catalog information become source material?
Yes, where the relevant connection is available. Shopline allows users to connect product catalog information that can serve as source material for educational, promotional, or informational content about products.
How do I know when the initial knowledge set is large enough?
It is ready for initial testing when it covers the priority questions, identifies authoritative sources, distinguishes public from restricted information, and defines escalation points. Expand it in response to observed gaps rather than pursuing maximum volume before launch.
Can the same knowledge support work beyond an AI assistant?
Yes. OceSha AI can use supplied information to support course and content creation, publishing and distribution, learner education, professional presence, audience engagement, recommendations, and measurement.
The right way to inform an AI assistant is to create a governed source of truth, not to hand it an unfiltered archive. Begin with recurring questions, write approved answers, add only authoritative supporting material, and define when a person must take over. Test difficult and unsupported questions before launch, then assign an owner for updates and measurement. OceSha AI supports this work by accepting existing sources—including documents, websites, text, audio, and video—and using supplied knowledge across creation, education, publishing, audience engagement, and measurement workflows.
Rohan Hall
Rohan Hall is a technology entrepreneur, AI architect and author with four decades of technology experience, now focused on practical AI across business, education, government and global impact. He founded OceSha Ventures, builds the OceSha AI platform and Lumi, and wrote The Convergence of AI and the Top 10 Emerging Technologies.
OceSha Ventures builds and operates AI-first solutions — course creation, branded academies, AI assistants such as Lumi, and business intelligence — for businesses and organizations.