Most businesses are using ChatGPT or Claude the way you would hire a new consultant on day one and hand them a task without any briefing. No context about the company, the customers, or the industry. The result is output that could have come from a Google search: technically accurate, practically generic. Meanwhile, one in five business messages in ChatGPT already runs through a Custom GPT that does have company context, according to OpenAI’s State of Enterprise AI report. The difference starts with a document about one page long.
Why does your AI tool keep giving generic answers?
AI models know a great deal about the world and nothing about your company. They do not know your industry focus, your customer profile, your pricing approach, or the way your team communicates. Every time you start a new conversation without context, the model begins from zero.
That is why so many teams hit a wall after the first few weeks of AI adoption. The outputs are technically defensible but off-brand, off-tone, and in need of rewriting before you would send them to a client. A proposal that reads like a template. A customer email your sales team would never actually send. A blog post that sounds like your competitor wrote it.
For context: McKinsey’s 2026 State of Organizations report found that while 72% of organizations now use generative AI, only about 6% qualify as AI high performers, reporting measurable business impact above 5% of EBIT attributable to AI. The gap between tool adoption and real productivity is not a model problem. It is a context problem.
According to Deloitte’s annual AI research, only 34% of organizations globally reach what they classify as deep AI transformation. The difference rarely comes down to which model a team uses. It comes down to how much context the model has to work with.
What goes into a good business profile?
A business profile for AI is a structured document of 500 to 800 words containing the core facts about your organization. Think of it as the day-one briefing you would give a new team member: who you are, what you do, who your customers are, and how you communicate.
The profile covers five sections:
- Company description. Industry, size, location, core activity. Two to three sentences is enough.
- Products or services. What you sell, which problem it solves, and your price range. Be specific: not “installation services” but “heat pumps, solar panels, and EV chargers for residential and small commercial clients in the South of England.”
- Target audience. Who your customers are, what they care about, and what language they use. A housing association communicates very differently from a venture-backed startup.
- Tone of voice. How your company communicates. Three to five adjectives and one real example sentence from a recent email or proposal. This section drives the most visible difference in output quality.
- Constraints. What your AI tool should not claim. Think: certifications you do not hold, prices that are not current, guarantees you cannot stand behind.
Sequence matters. AI models weight the beginning of a context prompt most heavily. Put your company description and target audience first. Put constraints last.
How to write it in an hour
Most businesses make this harder than it needs to be by starting with a blank document. Instead, answer the five questions above as if you were explaining your company to a new colleague on their first day. Write how you would actually speak. Edit later.
Here is a concrete example built around a fictional UK renewable energy installer. Copy the structure and replace the details with your own.
Step 1: Describe the company in three sentences.
“Hartley Renewable Energy is a Bristol-based installer with 22 employees. We install and maintain heat pumps, solar panels, and EV chargers for homeowners and small businesses across the South West. Clients choose us for straight advice and fast response times, including when we recommend against a project.”
Step 2: List your services with key specifics.
“Heat pumps (air-source and ground-source), solar panels (residential up to 20 panels, commercial up to 200), EV chargers (home and commercial). Project range: GBP 3,500 to GBP 45,000. Typical lead time: 3 to 8 weeks.”
Step 3: Describe your customer.
“Homeowners between 35 and 65 pursuing energy cost savings or sustainability goals. Small businesses looking to decarbonize premises or vehicle fleets. Technical knowledge varies widely. All communication must be jargon-free unless the client specifically works in the field.”
Step 4: Define your tone of voice.
“Direct, honest, helpful, never pushy. Plain English throughout. No marketing superlatives. If something is expensive, we say so. Example: ‘An air-source heat pump is the most common choice for a mid-terrace house. Budget GBP 6,000 to GBP 9,000 installed.’”
Step 5: Set the constraints.
“Do not cite specific government grant amounts; these change frequently. Do not guarantee payback timelines. Refer legal questions to a solicitor. Never use ‘cheapest’ or ‘best’ without a specific supporting reason.”
Notice how specific that is compared to “we are an installation company that helps customers.” The more concrete the profile, the less post-editing the output needs. With this profile in place, proposal emails read as if a team member wrote them, including the right tone, the right product detail, and the right price bracket. That is what the difference looks like in practice.
Where to load it in each tool
Each major AI platform handles company context differently. Current pricing in USD:
| Tool | Where to add it | Context space | Price per month |
|---|---|---|---|
| ChatGPT | Custom Instructions (via Settings) | ~1,500 characters | Free or Plus: $20 |
| ChatGPT Team | Custom GPT with Knowledge files | Up to 20 files per GPT | $30/user |
| Claude | Projects (project context) | Up to 200,000 tokens | Pro: $20, Team: $25 |
| Copilot | Microsoft 365 Business Chat | Varies by license | $30 (M365 Copilot) |
Custom GPT usage grew 19x this year, OpenAI reported. One in five business messages in ChatGPT now runs through a custom-context version. This is no longer an early-adopter pattern.
Worth noting: Claude’s Projects feature offers the widest context window of any mainstream AI tool. You can add complete documents as project context, including product catalogs, pricing sheets, and style guides. Everything in the project persists across every conversation within that project, which makes it well-suited to teams running repeatable workflows like client communication or internal reporting.
ChatGPT’s standard Custom Instructions cap at around 1,500 characters, which is tight for a full profile. The practical approach: put the core elements (company description, audience, tone of voice) in Custom Instructions, and save the extended version as a Custom GPT your whole team can open.
What should you never include?
A business profile should make your AI tool more useful, not create a data exposure. Research on shadow AI usage consistently shows that a significant share of employees at small and mid-sized businesses share internal company data with consumer AI tools, often without knowing how that data is retained or processed afterward.
Three rules keep your profile safe:
- No passwords, API keys, or financial figures. Your AI tool does not need to know your revenue to write a good proposal. Keep operationally sensitive numbers out of the profile.
- No personal data about clients. Under GDPR, processing identifiable customer data through AI tools requires a data processing agreement with the vendor. Describe your audience in general terms (“homeowners in the South West”), not with names or contact details.
- No confidential business strategy. Tone of voice and product range are safe to share. Margin strategy, acquisition plans, and internal revenue targets are not.
Paid tiers like ChatGPT Plus and Claude Pro include the option to opt out of using your data for model training. Free tiers often do not. Check your AI tool’s data settings before loading a business profile, and keep a copy of the data processing agreement on file.
How much of a difference does it actually make?
OpenAI reported earlier this year that employees using AI effectively save 40 to 60 minutes per working day. The word “effectively” is doing the heavy lifting there. Effective use starts with context.
The practical difference is visible immediately. Ask ChatGPT to write a proposal email without context, and you get a generic template you will spend 20 minutes reworking. Add a business profile, and you get a draft that reads as if your company wrote it, in the right tone, with the right product detail and price framing. That is the difference between copy-pasting and actually delegating.
The compounding effect is where it gets interesting. Every prompt builds on the same foundation. After a week, you do not have one better email; you have dozens of better outputs across the whole team. According to research from Boston Consulting Group, businesses deploying AI structurally achieved 1.7 times higher revenue growth than comparable companies that did not. Context is not the whole explanation, but it is the fastest single improvement most organizations can make. Most are still skipping it.
Three mistakes that make a business profile worthless
Not every business profile works. These three errors come up most often.
Too vague. “We are a service company that helps clients with digital solutions.” This tells an AI tool nothing it could not have invented. Name your sector, your products, and your customers specifically. Vague is useless; specific is actionable.
Too long. A 3,000-word profile gets buried in the context window. AI models become less precise as instructions grow longer. Stick to 500 to 800 words. Cut everything the model does not need to produce the right output.
Never updated. Your company from a year ago is not your company today. New products, a shifted customer base, a revised tone. A quarterly update takes 15 minutes and stops your AI tool from using outdated positioning in client-facing communication.
What should you do this week?
Put an hour on your calendar. Not next month. This week. Write the profile using the five steps above and paste it into the Custom Instructions of your most-used AI tool. Test it on three tasks you gave AI last week and compare the output to what you got before.
Once you have a version that works, share it with your team via a shared document, or build it into a Custom GPT that everyone can open. A business profile works best when the whole team is working from the same version, not each person running their own variation.
Twenty percent of business ChatGPT users are already doing this. The rest are still typing cold. The gap between those two groups is one hour of work.
