Company context
Company context
Company context is written-down, verified knowledge about the company (facts, rules, examples and constraints), selected for a task and used by people and AI. A knowledge base is the place where those sources are stored. The context covers who you are, what you offer, who you work for and what you must never promise. You borrow the language model from a provider, and the context is yours and works in every tool. It starts as a single page and grows with each review.
Illustrative example: A team writes down how an offer is really prepared. The same document then goes to an AI assistant, a salesperson and a new colleague.
See: Build a company knowledge base your team and AI agents can use. ↗Channels and tools
Channels and tools
Channels and tools are the places where context does its work: the website, social media, CRM and automations. All of them draw on the same source, so the message stays consistent without manual copying.
Illustrative example: A change in the offer description reaches the website, draft posts and CRM replies without being fixed three times.
See: We automate business processes and connect the tools you already use. ↗Channels and tools
CRM
The CRM collects conversations with clients, so it is a mine of questions and objections. In return it receives ready answers and the offer description from the context.
Illustrative example: A salesperson sees prepared answers to the most common objections on the client record.
See: Organise the journey from enquiry to customer delivery. ↗Channels and tools
Automations
Automations move information between tools according to agreed rules. The process description in the context says what should happen, and exceptions go to a person.
Illustrative example: A new website enquiry creates a CRM record and notifies the owner, while incomplete data goes to manual verification.
See: We automate business processes and connect the tools you already use. ↗AI assistants and agents
AI assistants and agents
AI assistants and agents receive context before every answer. They read approved knowledge, apply the brand rules and their output goes through human review. Answers then speak in the voice of your company.
Illustrative example: In SyncBooster, the agent first reads the brand context and only then writes a post.
See: We implement AI in your company, starting with one process. ↗AI assistants and agents
Approved knowledge
Approved knowledge is the part of the context a person has checked and cleared for use by AI. The rest waits for review.
Illustrative example: A new service description reaches the assistant only after the context owner accepts it.
See: Company context for AI: how to build and maintain it ↗AI assistants and agents
Instructions and prompts
Instructions and prompts tell the assistant its role, what to use and what not to do. Kept together with the context, they can be fixed in one place.
Illustrative example: The website assistant's instruction says that a question about price is routed to a conversation with a person.
See: An AI chatbot that answers from your company’s own knowledge. ↗AI assistants and agents
Answer review
Answer review checks whether AI sticks to the sources and rules. Errors return to the context as a correction, so the same mistake does not repeat in later answers.
Illustrative example: The responsible person reviews a sample of assistant answers each week and marks those that need a fix in the source.
See: Understand how AI is used in your company and what needs attention. ↗AI assistants and agents
Agent tasks
Agent tasks are clearly described actions: where the agent gets its data, what it may do alone and what needs approval. The scope is narrow and checkable.
Illustrative example: An agent prepares a draft reply to an enquiry, and only a person sends it.
See: We put AI agents to work on specific tasks in your company. ↗Team
Team
The team is the people who use the context and keep it current. One shared resource saves you explaining everything from scratch every time the team changes.
Illustrative example: A new freelancer gets the same brief as the team and writes in the brand voice from day one.
See: How to build a B2B marketing team ↗Team
Onboarding
Onboarding gets shorter when company knowledge is written down and organised. A new person reads the context and finds answers to the most common questions.
Illustrative example: In the first week, a new salesperson works through the offer, objections and brand rules kept in one place.
See: How to build a B2B marketing team ↗Team
Decisions
Decisions are a record of what was agreed and why, with a date and a responsible person. Nobody reconstructs agreements from memory, and AI does not build on rejected ideas.
Illustrative example: The agreement "we drop the humorous tone in offers" enters the context together with its date.
Team
Partners
Partners are the agencies, freelancers and suppliers who work on the company's behalf. They receive the part of the context they need, and the scope of access and rights is set in the agreement.
Illustrative example: A designer receives the brand rules and audience description. Documents containing client data stay inside the company.
See: We run your company’s B2B marketing within an agreed scope. ↗Brand rules
Brand voice
Brand voice describes how you speak to clients, which words you use and which you avoid. It works best when based on a few of your own texts that you consider exemplary.
Illustrative example: Three chosen articles annotated "we write like this" and "we do not write like this" give the model a concrete pattern of your voice.
See: How to publish consistently in B2B social media ↗Brand rules
Confidentiality
Confidentiality sets what must never go into the context or into AI tools: client data, contracts, passwords. The rules are agreed before anyone pastes a document into a chat.
Illustrative example: A "we do not paste this into chat" list sits next to the assistant instructions and applies to new team members too.
See: AI use policy for a small business: what to include and how to roll it out ↗Brand rules
Limits on promises
Limits on promises record what the company neither guarantees nor implies. They protect the client from disappointment and the brand from retracting promises.
Illustrative example: When describing a service, the assistant points out what is agreed individually and leaves out deadlines and results.
See: Understand how AI is used in your company and what needs attention. ↗Products and services
Products and services
Products and services are the agreed description of the offer: what you sell, to whom, in what scope and what sets you apart. A single source of truth for the offer prevents the website, the salesperson and the assistant from telling three different versions.
Illustrative example: After a service scope changes, you fix the description in one place, and the website, the offer and the assistant use the new version.
See: AI implementation and B2B marketing for small and medium-sized businesses. ↗Products and services
Offer
The offer is the full, current list of what the company sells, together with the limits of each scope. We write it in plain language that a model can understand too.
Illustrative example: A service description states what is in scope and what needs a separate quote, so the assistant does not promise more than the company delivers.
See: AI implementation and B2B marketing for small and medium-sized businesses. ↗Products and services
Use cases
Use cases show the situations in which a client turns to a given service. We describe each as a concrete scenario with a start, steps and an outcome.
Illustrative example: The scenario "a company wants to organise its answers to enquiries" describes the input, the people and the document produced at the end.
See: Business process audit: review one workflow. ↗Products and services
Pricing and scope
Pricing and scope are the quoting rules and service boundaries written down so that sales and AI say the same thing. Specific amounts are agreed before work starts.
Illustrative example: Asked about a price, the assistant points to a scoping conversation where a person sets the amount.
See: Start with what you want to improve. ↗Audiences
Buying decisions
Buying decisions describe who on the client side decides, in what order and on what basis. This knowledge shapes content for every stage, from the first question to the signed agreement.
Illustrative example: If the board decides after accounting gives an opinion, we prepare a separate piece of material for each of those people.
Audiences
Segments
Segments split audiences into groups with similar needs, for example by industry or company size. Each segment gets its own message, but all of them rest on the same context.
Illustrative example: A service for manufacturers and for agencies can share one core offer and use different examples.
See: For companies that want their current team to work better. ↗Knowledge and evidence
Knowledge and evidence
Everything that can be verified lives here: documents, completed work, sources and figures. Every fact needs an owner and a review date so that AI does not rely on outdated information.
Illustrative example: A service description enters the context together with the name of the person who last confirmed it and the date of that confirmation.
See: Company context for AI: how to build and maintain it ↗Knowledge and evidence
Documents
Documents are procedures, offers, terms and notes in which the company describes how it really works. Current versions enter the context and old ones leave circulation.
Illustrative example: A sales enquiry procedure moves into the shared context and becomes a source for the assistant.
See: Company context for AI: how to build and maintain it ↗Knowledge and evidence
Completed work
Completed work is a set of described projects the company can refer to: what was done, in what scope and what came of it. We record only what can be confirmed, together with whether the client agreed to publication.
Illustrative example: A short description of an implementation, with its scope and publication consent, feeds the offer and answers to questions about experience.
See: SyncBooster: the product where we validate processes before applying them with clients. ↗Knowledge and evidence
Sources
Sources say how we know something is true: a report, a contract, system data or a conversation with an expert. They let you check an answer and correct it at the source.
Illustrative example: A market claim carries a link to the study and the date it was checked, so the assistant can cite it directly.
See: Company context for AI: how to build and maintain it ↗Knowledge and evidence
Facts and figures
Facts and figures are a short list of claims the company uses in public, each with a date and a source. Figures that cannot be confirmed stay out of the context.
Illustrative example: The slogan "we are the leader" drops out, and the context keeps the founding year and the scope of services, because those can be checked.
See: Company context for AI: how to build and maintain it ↗