Many people are using artificial intelligence as a better version of Google.
They open a chatbot, ask a question, copy the answer and close the conversation. The next day, they open a new chat and start explaining everything again.
Who they are. What the business does. Who the customer is. What tone they prefer. What they worked on yesterday.
AI can still be helpful this way, but it is not where the real value lies.
The bigger opportunity is to create an organised AI workspace that understands the context of a specific business project. Instead of starting from scratch every morning, you give the AI access to the relevant instructions, reference documents and previous work.
This idea was explored in a recent episode of the Leveraging AI podcast featuring AI strategist Loren Bartley. The episode argues that poor AI results are not always caused by bad prompts. Often, the real problem is the workflow surrounding the prompts.
This is an important shift.
The question is no longer only:
“What should I ask AI?”
It becomes:
“How should I organise my business information so AI can help me repeatedly?”
Why one-off chats create inconsistent results
Imagine that you ask a new employee to write a customer email.
You would probably explain:
- What your business does.
- Who the customer is.
- What has already happened.
- What tone to use.
- What outcome you need.
- What the employee should avoid saying.
Now imagine that the employee forgets all of this every evening.
The following morning, you would have to explain everything again.
That is how many people currently work with AI.
They repeatedly paste the same background information into new conversations. Important decisions become scattered across different chats. Brand guidelines sit in one document, customer notes in another, and previous drafts somewhere else entirely.
This causes three common problems:
Lost context: The AI does not have enough background to understand the task properly.
Inconsistent output: Different chats produce different tones, formats and recommendations.
Repeated work: You spend time rebuilding information that already exists.
An AI system can only work with the context it has been given. It does not automatically understand your full business simply because you mentioned it in an unrelated conversation several weeks ago.
This is why organisation matters as much as prompting.
What does it mean for an AI project to “remember”?
When we say that an AI project remembers your business, we do not mean that it develops a human memory or independently learns everything about your company.
It means you create a dedicated workspace containing the information needed for a particular area of work.
For example, you could create separate AI projects for:
- Your monthly marketing.
- A specific client account.
- A product launch.
- Staff policies and procedures.
- Proposal development.
- Training material.
- Business planning.
- Customer support.
Inside the project, you can add the documents and instructions that should guide the AI.
A marketing project might contain:
- Your company profile.
- Brand voice guidelines.
- Customer personas.
- Product and service information.
- Previous social posts.
- Approved terminology.
- Content objectives.
- A list of claims or phrases to avoid.
You can then use different conversations inside that workspace for campaign ideas, blog drafts, newsletters, social posts and performance reviews.
ChatGPT Projects are designed to keep related chats, files and project-specific instructions together. They can also use the conversations and files within the project as context, helping you continue recurring work without rebuilding the background each time.
Claude Projects work in a similar way. Users can create a dedicated project, upload documents to its knowledge base and provide project instructions that apply to conversations within that workspace. Claude’s help documentation does note an important distinction: information is not automatically shared across every project conversation unless it has been added to the project knowledge base.
The tools may work differently, but the business principle is the same:
Put the right information in the right workspace before asking the AI to do the work.
How to build your first business AI project
You do not need to automate your entire company.
Start with one repeated area of work where you regularly have to explain the same background information.
Step 1: Choose one clear project
Avoid creating a project called “My Entire Business”.
That will quickly become cluttered.
Choose something specific, such as:
- Weekly social media.
- Client proposal writing.
- Customer enquiry responses.
- Monthly management reports.
- Course development.
- Recruitment and onboarding.
A clear project makes it easier to decide what information belongs inside it.
Step 2: Write simple project instructions
Tell the AI what role it should play and how you want it to work.
For example:
You are helping me manage the monthly marketing for my business. Write in plain English. Keep the tone warm, confident and practical. Do not make exaggerated claims. Ask for missing campaign information before producing final copy. Use South African spelling and examples where appropriate.
These instructions become the project’s operating rules.
You can also specify:
- Your preferred formats.
- The intended audience.
- Words you use or avoid.
- Approval requirements.
- Legal or compliance boundaries.
- The steps the AI should follow.
- The information it must never invent.
Step 3: Add useful reference information
Upload only information that is relevant to the project.
For a sales proposal project, this might include:
- Your company profile.
- Service descriptions.
- Pricing structures.
- Case studies.
- Proposal templates.
- Frequently asked questions.
- Standard terms and conditions.
For a customer support project, it might include:
- Product guides.
- Approved troubleshooting steps.
- Returns policies.
- Escalation procedures.
- Common customer questions.
More documents do not automatically create better results.
Good project knowledge should be current, clearly named and easy to understand. Remove duplicate, outdated or conflicting information wherever possible.
Step 4: Separate the work into conversations
Do not put every task into one enormous chat.
Create different conversations inside the project for different workstreams.
A marketing project could include conversations called:
- August Campaign Plan.
- LinkedIn Drafts.
- Newsletter Ideas.
- Website Updates.
- Monthly Performance Review.
This makes the project easier to navigate while keeping the shared business context close to the work.
Step 5: Create a simple update routine
Your business will change.
Products are updated. Prices change. New decisions are made. Campaigns are approved. Client requirements develop.
The AI project must be updated too.
At the end of an important piece of work, ask:
- What was decided?
- Which document is now the current version?
- What should be added to the project knowledge?
- What information is outdated?
- What should the next conversation know?
You could maintain a simple document called Project Status containing:
- The current objective.
- Recent decisions.
- Completed work.
- Outstanding actions.
- Important dates.
- The next priority.
This becomes a basic command centre for the project.
A practical South African business example
Imagine that you run a small training business.
Every month, you develop workshops, write proposals, prepare facilitator notes, create social posts and respond to potential clients.
Without a project, you may explain the same information repeatedly:
We provide practical training. Our audience is non-technical. We do not use complicated jargon. Workshops must include exercises. Keep the examples locally relevant. Do not promise that AI will replace staff.
Instead, you could create an AI Training Development project.
Add:
- Your list of courses.
- Your teaching method.
- Previous workshop decks.
- Brand guidelines.
- Exercise templates.
- Proposal examples.
- Feedback from earlier sessions.
- Your privacy and responsible AI guidelines.
You could then ask:
Using our existing workshop structure, create a beginner-friendly exercise for a group of operations managers. The activity should take 20 minutes and use a realistic South African business process.
The AI is not producing the answer from one clever prompt alone.
It is using the structured context you have already prepared.
That is the beginning of workflow orchestration: combining instructions, business information, repeatable steps and separate AI conversations into a more organised way of working.
Remembering is useful, but human oversight still matters
An organised AI project can improve consistency, but it should not become an unquestioned source of truth.
AI can still misunderstand instructions, miss outdated information or produce an answer that sounds convincing but is incorrect.
You still need a person to:
- Check important facts.
- Approve customer-facing communication.
- Review legal or financial content.
- Protect confidential information.
- Update outdated documents.
- Decide which recommendations make business sense.
Be careful about uploading personal information, confidential client material, passwords, banking details or sensitive employee records. Check the privacy settings, data controls and organisational policies that apply to the AI platform and account you are using.
The goal is not to hand your business over to a chatbot.
The goal is to build a better-organised system in which AI can support your people without forcing everyone to start from zero each time.
Start with one project, not a complete transformation
You do not need an advanced AI agent or a complicated automation platform to begin.
- Choose one repeated workflow.
- Create one dedicated project.
- Add a small set of trusted documents.
- Write clear instructions.
- Organise the conversations.
- Then use it consistently for a few weeks.
Pay attention to where the system saves time and where it still needs better information. Improve the project as you learn.
The real power of AI is not found in a single perfect prompt. It comes from building a useful working environment around the technology.
When your AI has the right context, clear boundaries and an organised workflow, it becomes more than a place to ask occasional questions.
It becomes part of how your business plans, creates and improves its work.
Keep exploring, keep learning, and don’t be afraid to experiment with these powerful tools. The future of your business is in your hands, and AI is here to help you unlock its full potential.
This blog was created with the assistance of AI, but the content and focus were generated by me.