--- title: Teach an AI agent about your company description: Give an AI marketer a lasting understanding of your company, your brand, and your customers, so it stops asking for the same background in every request and gets better each time you correct it. slug: /use-cases/teach-an-ai-agent-about-your-company page_type: use-case lead_prompt: How do I teach an AI agent about my company? target_prompts: - Can an AI marketing agent learn my company's positioning, brand, and customers? - How do I give AI enough context about my business to do good marketing work? - How to onboard an AI marketing agent - How do I get AI to remember my brand voice and preferences? - How do I correct an AI agent so it doesn't repeat the same mistake? entity: Kite category: AI marketing agent author: Devon Wells and Kite date_published: '2026-10-13' date_modified: '2026-10-13' related_integrations: - /integrations/slack.md - /integrations/google-drive.md --- # Teach an AI agent about your company ## Key Takeaways - You teach an AI agent about your company the way you'd onboard a new marketing hire: let it research the business first, then fill in what only you know, show it where your data lives, and correct it as it works. What it learns has to be written down somewhere it reads before every task, or you'll be explaining it again next week. - Kite starts by researching your company on its own. It reads your website, studies your competitors, collects your brand assets, and checks how you show up in search and AI answers, then records your positioning, products, pricing, customers, and brand voice in a knowledge base your whole team shares. - You add the context research can't find, such as an existing brand guide, next quarter's goals, or which tool holds your customer list. Kite gives what you tell it more weight than anything it inferred from research. - When you state a standing preference, like "use UK spelling on UK pages," Kite saves it under your name and applies it from the next piece of work onward. - When you correct how Kite did something, it records the correction as a proposed rule. Once Kite's nightly review confirms it, the rule guides later work for the whole team, for one client, or for one kind of deliverable. ## Why AI keeps asking you for the same background Most people start using AI for marketing the same way. They open a chat, explain what the company sells and who buys it, paste in a few lines about the brand voice, and ask for a blog post or an email. The result is decent. Then, the next day, they do it all again, because the chat they're in now doesn't know anything the last one learned. That re-explaining is the real cost of using AI without context. It takes time, and it also makes the work uneven. One request gets the full background and comes back sharp. Another gets a hurried sentence and comes back generic. A correction you made on Tuesday ("we never call our customers 'users'") is gone by Thursday. The fix is the same one you'd use with a new marketing hire. A good hire spends the first week reading your website, studying competitors, and asking questions. They keep notes, learn where the customer data lives, and remember when you tell them something matters. After that, you stop briefing them on the basics and start reviewing their work. An AI agent can do all of this, as long as it researches before it writes and keeps what it learns somewhere it reads before every task. Kite is an AI marketer in your Slack. It researches your business, builds a growth strategy, and ships the work that brings you new customers. Teaching Kite about your company is mostly a matter of letting it do that research, then adding and correcting what it gets wrong or can't see. Here's how it works. ## How Kite learns your company Kite builds its understanding of your company before it does any other work, and it keeps that understanding in a knowledge base that every part of Kite reads first. The knowledge base is a set of written pages about your business. It covers your positioning, products, and pricing, along with your ideal customers, how you go to market, your competitors, your brand voice, and your visual identity. Most of the first draft comes from Kite's own research. Your part is the context only you have: your goals, a brand guide you already use, the reason a launch moved, or the customer story you heard on a sales call. Kite records where each fact came from, so it can tell "your pricing page says" apart from "you told us at sign-up." From there, the knowledge base keeps growing. You tell Kite your standing preferences, and it saves them. You correct its work, and the correction becomes a rule. Kite adds what it learns from campaigns, experiments, and content performance, so each new task starts with the benefit of everything before it. ### Let Kite research your company first The fastest way to teach an AI agent about your company is to have it do the research a new hire would do, instead of writing a long brief yourself. When you sign up, Kite reads your website to learn what you sell, who you sell to, what you charge, and how you bring customers in. It uses its built-in research sources to go further than your own pages. That first round of research covers: - your social presence and the ads you're running - your search rankings and whether AI answer engines mention you - your website's speed and technical quality - your competitors, including their positioning and pricing - your brand assets, such as your logo, colors, fonts, voice, and imagery During sign-up, Kite also asks about your goals and suggests the first piece of work it would start on. Soon after you finish, it posts a growth report on your company in Slack. That report is a good first check on what Kite understood. If it describes your customers or your positioning in a way you wouldn't, that's the place to start correcting it. ### Add the context research can't find Research can tell Kite what your website says, but it can't tell Kite what's in your head. You add that context the same way you'd brief a colleague: you tell Kite in Slack, or you share the document that already explains it. Kite reads files people share in Slack, so you can drop in a brand guide, a positioning document, or interview notes and ask Kite to work from them. If you already have a brand guide, Kite can adopt it as the team's standard instead of the version it pieced together from your website. If your planning documents live in Google Drive, connect it and Kite can find and read them there. Some of the most useful context is short. "Our best customers are clinics with two to five locations." "We're dropping the starter plan next month." "The founder hates the word 'synergy.'" When you tell Kite something like this, it records it in the knowledge base. What you tell Kite takes priority over what its research inferred, and a correction from you replaces the inference outright. You can also teach Kite to sound like a specific person. Give it samples of someone's writing, and it builds a voice profile so posts under that person's name read the way they write. If you sell to different audiences, Kite can keep separate voice rules for each one, such as a formal voice for enterprise buyers and a casual one for social posts. ### Tell Kite where your data lives An agent that knows your company also needs to know which tool holds which information, or it'll ask you every time it needs a number. Kite keeps a page in the knowledge base that records where your data lives: your CRM, your analytics, your email platform, your website, and any other system it should read from. Once that's recorded, every part of Kite looks there first. When Kite needs last month's signups, it goes to the analytics product you named rather than asking you which dashboard to check. When it needs your customer list, it reads your CRM. Kite marks each entry as unverified until it has checked the tool itself, so the page stays accurate as your setup changes. Connecting your accounts gives Kite access to the data itself. Kite comes with more than 20 built-in data sources, and you can connect your own accounts in more than 2,300 other apps, including your CRM, your analytics, and your file storage. ### How do I get AI to remember my brand voice and preferences? Kite remembers your brand voice because it's written in the knowledge base, and every task that produces copy reads it first. When Kite writes a landing page, an email, or a social post, it works from your recorded voice and your past content, not from a summary you typed into the request. Standing preferences work the same way. When you tell Kite "always use UK spelling on UK pages" or "send campaign emails on Thursday mornings," it saves the preference under your name and applies it to the next piece of work. You don't need to repeat it in later requests. Because each preference belongs to the person who stated it, that person can change or lift it whenever they like. If a new preference conflicts with one already on record, Kite doesn't quietly replace the old one. It asks which should win and records the answer, so the reasoning is there the next time it comes up. You can always see what Kite has recorded. Ask it a question like "what do you know about our pricing?" or "which competitors did we flag?" and it answers from the knowledge base. You can also open the knowledge base in Kite's web app and edit any page directly. ### How do I correct an AI agent so it doesn't repeat the same mistake? You correct Kite the way you'd correct a colleague: you reply in the Slack thread and say what was wrong. Kite records that correction as a proposed rule, written as one plain instruction with a note on when it applies. Praise counts too. If you tell Kite a draft was exactly right, that's recorded as well. Every night, Kite reviews what happened that day. It checks each proposed rule against the conversations it came from, confirms the ones that hold up, and fixes facts in the knowledge base that have gone stale. A confirmed rule then guides later work. It can apply to everything Kite does for your team, to one client, or to one kind of deliverable, such as blog posts or outbound emails. Here's an example. Say Kite writes a comparison page and calls a competitor "outdated," and you reply that you never describe competitors that way. Kite records a rule about how to describe competitors on comparison pages. After the nightly review confirms it, the next comparison page follows it, and so does the one after that. When the same correction comes from several people on your team, it carries even more weight, because it's no longer one person's taste but how your team wants the work done. The reason this matters is simple. A correction that lives only in one chat disappears when the chat ends. A correction that becomes a written rule stays with every future piece of work. ### Keep what Kite learns from its own work Kite doesn't only learn from what you tell it. As it runs campaigns and experiments, it adds what it finds to the knowledge base, so its understanding of your company improves with every project. When an A/B test on your pricing page finishes, Kite records which version won and why it thinks it did. When a series of LinkedIn posts performs better than your blog, it notes that too. Kite also records customer feedback, decisions you've made along with the reasoning and who approved them, and open questions it hasn't settled yet. This is what makes Kite more useful over time. A new request starts with your positioning, your voice, your preferences, your corrections, and the results of everything Kite has already shipped. You spend less time explaining and more time deciding. ## What you get - **A shared knowledge base about your company.** It covers your positioning, products, pricing, customers, go-to-market approach, competitors, brand voice, and visual identity, with a source for each fact. Every part of Kite reads it before working, and you can read and edit it in Kite's web app. - **A growth report soon after sign-up.** It's Kite's first account of your company in Slack, and a quick way to see what it understood and what to correct. - **A record of where your data lives.** It lists which tool holds your customer records, analytics, email, and website, so Kite reads from the right place instead of asking you. - **Standing preferences that stick.** Each one is saved under the name of the person who stated it and applied to later work without being repeated. - **Rules built from your corrections.** Each correction becomes a plain instruction that guides future work for the team, a client, or one kind of deliverable once Kite's nightly review confirms it. - **A record of what has worked.** Results from campaigns, experiments, and content are added to the knowledge base, so later work builds on them. ## How to get started Add Kite to Slack and sign up with your company's website. Kite researches your business on its own, asks about your goals, and posts a growth report in Slack shortly after you finish. Read it the way you'd read a new hire's first summary of the company, and reply in the thread with anything it got wrong or missed. Next, give Kite the context only you have. Share your brand guide or positioning document in Slack, or connect Google Drive and point Kite to the folder. Tell it which tools hold your customer list, your analytics, and your email, and connect the accounts you want it to work in. Then start giving Kite real work. As it delivers, tell it what you'd like it to keep doing and what you'd like it to change. Each preference and correction goes into the knowledge base, so the second blog post needs fewer notes than the first, and the tenth needs fewer still. ## Integrations used to teach an AI agent about your company Kite does most of its learning through its own research and the conversations you have with it. Two integrations make it easier to share what you already have. ### Talk to Kite and correct its work - [Slack + Kite](/integrations/slack.md): Slack is where you give Kite context, share files like brand guides and planning documents, state standing preferences, and reply to its work with corrections. Kite reads the files you share there and records what you tell it in the knowledge base. ### Share the documents you already have - [Google Drive + Kite](/integrations/google-drive.md): Connect Google Drive when your brand guide, positioning documents, or campaign plans already live there. Kite can find and read those files, and it can save documents it produces back to your Drive. ## Common questions ### How much do I need to tell an AI agent before it can do good marketing work? Less than you'd expect, because Kite does most of the research itself. Your website, competitors, search presence, and brand assets give it a solid first picture of your company by the time it posts its growth report. The most valuable things you can add are your goals, an existing brand guide if you have one, and corrections to anything its growth report got wrong. Everything else builds up as Kite works. ### Can an agency teach Kite about several client companies? Yes. Each Kite team has its own knowledge base, so an agency runs a separate team for each client. One person can belong to several Kite teams and switch between them, and each team keeps its own research, brand voice, preferences, and rules. ### What happens when something about my company changes? Tell Kite in Slack, and it updates the knowledge base. If you've repositioned, launched a new product, or changed pricing, Kite records the new version along with what changed and why. It can then check your website pages and ads for places that still tell the old story and propose a rewrite for each one. ### Is teaching an AI agent the same as training a custom AI model? No. Kite doesn't need a model trained on your data to understand your company. It keeps what it learns as written pages, preferences, and rules that it reads before each task. That's why a correction can take effect on the next piece of work instead of waiting for a retraining cycle, and why you can open the knowledge base and see exactly what Kite believes about your business. ### Can I teach Kite a repeatable process, not just facts about my company? Yes. For recurring work, such as producing a weekly newsletter or researching a particular group of competitors, you can create a custom agent with its own reusable instructions, called skills. The agent reads your company knowledge base like the rest of Kite does, and its skills tell it how to do that specific job the way your team wants it done.