7 min read
Discover how to build your Copilot and AI context to improve results
Building an "AI context" is a concept that isn't widely known, but it's a really important one to grasp because getting it right will greatly improve the output you get from AI. It will also save your team significant time by no longer needing to repeat context for every prompt you write.
- Why poor context causes poor AI results
- What is business context for an AI?
- What is personal context for an AI?
- Where to build context
- How to create your business AI context
- How to create your personal AI context
What is context in terms of AI?
Context is everything that you already know that an AI doesn't.
Think about it this way, when you ask an AI to do something, you write a prompt. A prompt tells the AI what to do—it's the instruction (if you want to learn how to write great prompts, take a look at our blog about using prompt frameworks). The context, however, tells the AI the background information and constraints that shape how the task should be completed.
Without proper context, even the most precisely worded prompt will produce generic, inconsistent outputs that may not align with your business standards. Context is what transforms AI into a tailored assistant that understands your business environment, communicates in your voice, and operates within your guidelines. Microsoft and Anthropic have both published documentation that backs up this approach; Comprehensive context creates outputs that require less revision and better reflect their business requirements. Even Gartner has been getting in on the action.
Context for AI really has two levels that you need to think about: Business Context and Personal Context. Both need to be fleshed out to make an AI useful. Let's explore what the difference is.
Why poor context causes poor AI results
The drawbacks of working without a shared source of business context quickly become obvious when they're pointed out, but it's something many business owners haven't thought of yet. Every employee ends up repeatedly explaining the same information to AI tools, wasting time and effort across the organisation.
It also leads to inconsistency. Different people will provide different levels of detail, use different terminology, or describe the same process in different ways. The result is AI-generated content that varies from person to person, whether that's customer emails, marketing materials, proposals or internal documents.
As the business grows, information will undoubtedly become fragmented, teams may create their own versions of FAQs, service descriptions or standard responses. This contradictory and disparate landscape of data will confuse an AI (not to mention new staff).
Whilst much discussion around AI adoption focuses on productivity gains, context serves an equally important function as a governance and risk management tool. For businesses operating in regulated sectors particularly, this aspect of AI context deserves careful attention.
Documented context establishes approved parameters that reduce the likelihood of AI producing inappropriate or non-compliant outputs. This includes embedding approved terminology that reflects your brand positioning and avoids language that could create legal exposure. Regulatory requirements relevant to your industry can be incorporated, helping AI-generated content be to compliance standards, reducing the amount of rework required.
Legal disclaimers, internal policies, and operational procedures all benefit from contextual embedding. The more precisely these parameters are documented and made available to AI systems, the lower your risk exposure. This is particularly relevant for small businesses where a single compliance failure or brand misalignment can have disproportionate consequences.
The above said, ALWAYS check the output of AI. Context improves the quality and consistency of AI outputs, but it should sit alongside human review, data governance, access controls and clear AI usage policies. It is not infallible, far from it.
One final thought on the consequences of not building out context. When someone leaves your business, the knowledge they have in their heads goes with them leaving the next person to make up their own. A shared business context means a continuity of data in your organisation. That in itself makes this process worthwhile.
What is business context for an AI?
Business context forms the foundation of how AI understands and represents your organisation. It's the baseline that's applied by AI regardless of who is using it in your organisation. Here we are talking about your broad organisational parameters such as your tone of voice, brand guidelines, policies, regulatory requirements, and operational guardrails.
A good analogy would be thinking about a really skilled contractor coming into your business. Before you set them off working, you need to give them an induction, tell them all about your company and how you operate. The same goes for AI.
What is personal context for an AI?
Whilst business context addresses organisational standards, personal context tailors AI output to individual people. It helps the AI respond as you would, and it will help it keep responses to queries and requests aligned with your specific requirements for your role.
Personal context includes information about an individual's role, specific responsibilities, communication preferences, and working style.
For example if a sales manager and a customer account manager give an AI the same prompt:
"draft a follow-up email after the meeting with Customer Ltd about adopting Microsoft Copilot"
The sales manager's personal context tells the AI that they are measured on generating new opportunities and growing revenue.
The customer success manager's personal context tells the AI that they are measured on customer retention and adoption of platforms.
With these two different contexts the AI might reply to the sales manager's query with an email that focuses on business value, booking a follow-up discovery session, including a call-to-action, and highlighting potential ROI and productivity gains. The customer success manager'ail might instead focus on agreed actions, adoption milestones, training recommendations, and next steps to help users get value from Copilot.
Where to build context
An important consideration that many won't have thought about is where to store and maintain AI context. You have two options.
The first approach involves building context directly within AI tools using built-in memory and remember functions. Microsoft Copilot and other AI platforms offer native capabilities to store preferences and contextual information within the application itself. This approach provides immediate integration and requires no external document management.
However, the second approach—storing context in structured documents within shared locations that AI tools can access—offers significant advantages. This method creates a centralised repository of organisational knowledge. When your business context exists as accessible documents, you can instruct any AI tool to reference these materials, eliminating platform lock-in. If you transition from one AI service to another, or use multiple AI tools simultaneously, each can access the same authoritative context source.
In addition, document-based context storage simplifies maintenance. When your business changes its guidelines or policies, you only need to update a single document rather than reconfiguring multiple AI platforms or individual user settings. This becomes particularly valuable as your team grows—new employees can be granted access to established context documents, ensuring immediate consistency without requiring individual setup for each team member and each AI tool they use.
So our recommendation is build out a library of word documents and store them in a controlled SharePoint or Teams location, make sure the right people have access, and encourage users to reference the relevant context documents when asking AI to complete business tasks. It's worth having users tell the AI to 'Remember' to always reference them but don't rely on it alone. Memory features vary between tools and are not a foolproof way to ensure the AI always uses your context.
How to create your business context
Establishing clear business context ensures that AI tools understand the professional standards you've set for your employees. This includes defining your preferred terminology—whether you use specific industry terms or avoid certain language—your communication style, and any compliance requirements relevant to your sector.
Business context also establishes the boundaries within which AI should operate. This includes your quality standards, customer service approaches, and any legal or regulatory constraints that must be consistently applied. By documenting these parameters once and making them available to your AI tools, you ensure that every interaction maintains professional consistency and meets your organisational standards.
We recommend creating the following suite of documents with appropriate headings and subheadings to help the AI understand how everything fits together. They can be written in plain English and even just use bullet points to simplify the writing for you.
- Company summary
Why it matters: Helps the AI understand who you are, what you do, and what matters to the business.- the company name, your industry, your company mission and purpose, your size, number of departments and what they are, what areas you operate in and the location of any offices, who your key competitors are, a brief company history, your strategic priorities for the next 1-3 years, reporting structures within the organisation, any common acronyms you use internally, a summary of the business systems you use, and lastly any KPIs that matter to the business.
- Product catalogue
Why it matters: This helps prevent the AI inventing service features that do not exist.- Detailed descriptions of each service, what is included, what is excluded, pricing approaches, target audience, common use cases, benefits and outcomes, related services, frequently asked questions, service delivery process
- Brand guidelines
Why it matters: Keeps AI-generated content consistent with your tone, language, visuals and style.
- tone of voice/formality and acceptable/unacceptable language (e.g. we refer to customers as clients. Also include examples of writing style - "write like this", "don't write like this"), image styles (e.g. always use miniature plastic model figurines on a plain white background), formatting preferences for reports, emails and presentations, preferred document structures, logo usage guidance, colour palette and brand colours, accessibility requirements, standard email sign-offs and greetings, preferred calls to action, and you may even want to include an instruction to always use UK spelling for words.
- Target customer profile
Why it matters: The AI can then tailor content to the audience rather than producing generic responses.- Ideal customer profiles, typical job roles of decision makers, common customer challenges, customer goals and objectives, frequently asked questions, customer misconceptions, customer buying triggers, reasons customers choose your organisation, reasons prospects typically hesitate, key industries served, examples of successful customer outcomes.
- Business processes and workflows
Why it matters: The AI can then suggest actions that align with how the business actually operates
- How work flows through the business, your sales process, your customer onboarding process, project delivery process, escalation procedures, approval processes, customer complaint handling, procurement process, renewal process
- Regulatory framework
Why it matters: Helps AI stay aware of the rules and constraints that apply to your sector.
- If you operate in a regulated industry make sure the AI knows which regulations apply to you and ideally provide it links to find those documents on official pages.
- How we make decisions
Why it matters: This is arguably the most important document of all
- What the organisation values, how priorities are chosen, factors considered when making recommendations, risk appetite, customer service philosophy, quality standards, what success looks like (this latter should include examples of great emails, reports, proposals, and presentations.)
For example:
We prioritise long-term customer relationships over short-term revenue.
We prefer practical solutions over complex solutions.
We communicate honestly and avoid exaggerated claims.
- What the organisation values, how priorities are chosen, factors considered when making recommendations, risk appetite, customer service philosophy, quality standards, what success looks like (this latter should include examples of great emails, reports, proposals, and presentations.)
Remember to review your business context regularly, and update it. Stale information is unhelpful to staff and AIs alike.
How to create your personal AI context
Creating your personal context should be significantly easier. Again you can create this directly in an AI using the Remember command but we'd suggest storing your personal context in a Word document stored in OneDrive and just point an AI at it.
Personal context will vary for every person so there's no easy checklist that we can provide you but some of the basics you might want to consider are:
-
Your title and role summary (maybe put your full job description in if you have it documented)
-
your decision-making authority
-
your preferred communication format (formal reports versus concise summaries)
-
stylistic preferences such as whether you use emojis in internal communications or prefer strictly professional language.
-
Your working hours and patterns (typical meeting schedule etc)
-
details of recurring tasks that you do, or specific workflows you follow.
When AI understands these personal parameters, it can provide more relevant suggestions, prioritise information more effectively, and generate outputs that require minimal adjustment before use.
Next steps...
If you're looking for assistance with implementing, or improving your use of AI in your business then get in touch with us. Our Edinburgh-based team provide AI consultancy across Scotland and northern England.
