Why AI Gives Generic Answers About Your Business, and How a One-Page Brief Fixes It
AI chat tools don't know your prices, policies or service area. Here's how a small business writes a one-page AI brief and saves it in ChatGPT or Claude.

Short answer: AI chat tools give generic answers about your business because they don't know your prices, policies or the way your shop works. The fix is cheap: write a one-page "AI brief" with the facts the AI keeps getting wrong, save it inside a ChatGPT or Claude project, and date it so someone owns keeping it current.
Key points
Databricks CEO Ali Ghodsi said at his company's Data + AI Summit 2026 that today's AI is already smart enough, and what it lacks is a company's own context.
In Workiva's 2026 Midyear Executive Benchmark Survey, released Aug. 11, 2026, only 11% of executives said their data quality is good enough for AI, yet 84% were at least somewhat confident in AI-written material going into an annual report with no human review.
Both ChatGPT and Claude have "projects" that keep your files and standing instructions in one place, so you don't retype the background in every chat.
Keep the brief current. An AI working from an old policy document can repeat it with confidence.
What to do this week: write a one-page AI brief for your business, date it, and load it into one AI project your team uses.
Why does AI give generic answers about my business?
Because it starts every chat knowing nothing about you. A chat tool has read a lot of the public web, but it has never seen your price sheet, your return policy, your service area or the unwritten rules your longest-serving employee carries around.
That gap is now a talking point at the top of the industry. Speaking at Databricks' Data + AI Summit 2026, CEO Ali Ghodsi said, "We don't need more intelligence for the AI, we need more context." His company sells tools that feed company data to AI, so it has a stake in that view. A small business doesn't need those tools to act on the point, though.
What does the research say about AI and messy business information?
Confidence is running ahead of the groundwork. Workiva's survey, from a company that sells reporting software, asked 2,272 people working in finance, risk and sustainability (847 of them in the C-suite). Only 11% of executives said their data quality is good enough for AI, and 26% said internal audits had caught AI errors that reached outside audiences or board members. Thomson Reuters' Checkpoint News reported that 84% were at least somewhat confident letting AI-written content go into an annual report unchecked by a person. It also noted the survey didn't say what kinds of errors got through or name any companies.
In an Oct. 19 column for TechTarget, longtime enterprise technology consultant David Linthicum warned that when policy documents are stale, AI can quote the outdated policy with confidence, and that AI tends to expose undocumented "tribal knowledge" rather than fix it. His advice to large companies applies on Main Street too: decide which decisions AI can make, which it can only suggest, and which stay with a person.
Those surveys and columns are about larger organizations. Our view is that the lesson scales down well: a five-person shop has far less information to sort out, so its first brief can be short.
What goes in a one-page AI brief?
Only facts you'd want every answer to get right. Keep it to one page so it's easy to keep current.
Who you are: business name, what you sell, towns you serve and your hours.
Current prices or price ranges you're comfortable having AI use in drafts, with the date you last checked them.
Policies: returns, deposits, cancellations, warranty, and how fast you reply to customers.
Your voice: three to five words on tone, plus words you never use.
Hard limits: for example, never quote a price that isn't on the list, never promise a date, and flag anything about refunds or legal matters for a person.
Who owns it: the person who updates the brief, and a "last updated" date at the top.
Leave out anything you wouldn't want stored by an AI company: customer details, passwords, account numbers and employee records. Our guide to AI privacy settings covers which plans may use what you type for training.
Where do I put the brief so the AI uses it?
In a project, so you set it up once. OpenAI's help center says a ChatGPT project holds a set of chats, files and instructions in one place, and that instructions set for a project are used only in that project and take priority over your account-wide custom instructions. As of Oct. 31, 2026, OpenAI lists 5 files per project on the Free plan, 25 on Go and Plus, and 40 on Pro, Business and Enterprise.
Anthropic's help center says Claude projects are available on every plan, with free accounts limited to five projects. Anything added to a project's knowledge base is used across all chats in that project, and you can set project instructions the same way. Anthropic also notes that one chat doesn't see another unless the information is added to the project knowledge.
A simple setup: paste the brief into the project instructions, upload your current price sheet and policy page as files, and start every customer-facing draft from inside that project.
How do I keep the brief from going stale?
Treat it like your price list. Put a review date on your calendar once a month and after any price or policy change, and update the copy in the AI project the same day. If two documents disagree, decide which one wins and delete or archive the other.
What does this look like for a small business in Southwest Michigan?
Take a hypothetical Allegan County landscaping company that uses ChatGPT to draft replies to quote requests. Without a brief, the drafts guess at seasonal rates and promise spring start dates the owner can't keep. With a one-page brief in a project (service towns, this season's price ranges, "never promise a start date," and a friendly, plain tone), the drafts start closer to right, and the owner still reads each one before it goes out.
What should you tell your IT provider or developer?
If a website chatbot or automation answers customers for you, ask where it gets its business facts, who updates them, and how fast a price or policy change reaches it. Ask for a test after each update so you can see the new answer before customers do.
What should you do this week?
Write the one-page brief, put a date and an owner at the top, and load it into one ChatGPT or Claude project. Then ask the AI three questions a customer really asks and check the answers against your own records.
For more, see our guides to checking what AI tools keep from your chats, rolling out AI one task at a time, and choosing between one AI agent and many.
FAQ
Do I need special software to give AI my business context? No. A one-page document saved in a ChatGPT or Claude project covers most small-business needs. Larger data tools are built for companies with many systems.
Is a project the same as custom instructions? Not quite. OpenAI says project instructions apply only inside that project and take priority over your general custom instructions, so you can keep business and personal settings apart.
Will the AI still make mistakes with a brief? Yes. A brief cuts down on guessing, but a person should still check anything that goes to a customer, especially prices, dates and policies.
How long should the brief be? One page is a good target. Longer briefs are harder to keep current, and a stale brief can spread old information.
Disclosure: No company named in this article paid for it, reviewed it or saw it before publication. Plan limits and features are as published by each company and were checked on Oct. 31, 2026; they can change. Not sponsored.
Sources
- PRIMARY: Databricks, "What it takes to scale agents in the enterprise: context, control and choice" (Ali Ghodsi, Data + AI Summit 2026 recap), accessed Oct. 31, 2026
- PRIMARY: Workiva, "One in Four Executives Say AI Errors Have Reached External Audiences or Boards," Aug. 11, 2026
- Thomson Reuters Checkpoint News (Denise Lugo), "AI errors reach boardrooms and investors, midyear report finds," Aug. 13, 2026
- TechTarget (David Linthicum), "The AI happy path is easy, but enterprise reality isn't," Oct. 19, 2026
- PRIMARY: OpenAI Help Center, "Projects in ChatGPT," accessed Oct. 31, 2026
- PRIMARY: Anthropic, Claude Help Center, "How can I create and manage projects?," accessed Oct. 31, 2026