A recent AI training and workflow session with a Kenyan safari company provided a practical example of what this can look like.
The goal was not simply to teach the team how to write better prompts. It was to explore how AI could help transform a time-consuming process, turning safari inquiries into realistic itineraries and proposals, while keeping humans responsible for important decisions such as pricing, accuracy and client data.
The real AI opportunity is bigger than writing content
Safari businesses receive inquiries that can vary enormously.
A client may want a short luxury safari, a family holiday combining wildlife and the beach, a budget-conscious trip, or a complex multi-country itinerary.
Someone then has to interpret the request, determine what the client is actually looking for, select suitable destinations and accommodation, consider travel times and seasonality, construct a logical route, calculate costs and eventually turn everything into a professional proposal.
AI can assist with many of these tasks.
But simply telling an AI tool, “Create a 7-day Kenya safari itinerary” is unlikely to produce a proposal that accurately reflects how a particular safari company operates.
The AI needs context.
It needs to understand the company’s preferred lodges, hotel categories, routes, policies, itineraries, FAQs, pricing logic and the type of customer being served.
This is where the concept of an AI-powered business workflow becomes much more interesting.
From a blank chatbot to a company-specific AI assistant
One of the key ideas explored during the training was the creation of a dedicated project environment containing the company’s approved information.
Instead of repeatedly explaining the business to an AI tool, the team can create a central workspace containing materials such as:
- Sample safari itineraries
- Lodge and hotel information
- Frequently asked questions
- Destination information
- Route and drive-time information
- Hotel and lodge classifications
- Company-specific guidelines
- Approved operational rules
The result is potentially much more useful than a generic AI conversation.
Rather than asking AI to invent an itinerary from general internet knowledge, the system can be instructed to work from the company’s own approved information.
That creates an important distinction:
Generic AI answers questions. A properly configured AI workflow can help execute a business process.
Why the AI needs a defined role
Another important lesson from the training was the importance of giving AI a clearly defined role.
For example, instead of simply telling an AI tool to “create an itinerary”, the system can be configured to behave more like a senior safari consultant or itinerary designer.
Its instructions can specify that it should:
- Interpret the client’s underlying travel intent.
- Identify the type of safari that best fits the request.
- Consider the season and destination conditions.
- Design the route before writing the itinerary.
- Use approved accommodation and destination information.
- Respect the company’s operational rules.
- Flag information that requires human verification.
- Produce a proposal in the company’s preferred format and tone.
This is an example of prompt engineering becoming workflow design.
The objective is no longer simply to find the perfect prompt.
It is to define how the AI should think through a recurring business task.
AI should not replace human judgment
One of the most important points from the session was also one of the simplest:
AI output still needs human verification.
Large language models are extremely capable at producing fluent text, organising information and generating possible solutions. But fluent output does not automatically mean accurate output.
For a safari business, this distinction matters.
An itinerary may look excellent while containing an incorrect drive time, an unsuitable lodge, an unrealistic route or an incorrect price.
Financial calculations are particularly important.
AI can help structure a quotation, identify the components that need to be included and present the information clearly. But actual costing should be checked against the company’s current rates and commercial rules.
A useful way to think about the workflow is:
AI drafts → human verifies → AI assists with refinement → human approves.
This is not an AI failure. It is a sensible division of responsibilities.
AI is powerful when it knows your business
Imagine asking a generic AI tool:
“Create a luxury 8-day Kenya safari for an American couple.”
You may receive a perfectly readable itinerary.
But the AI does not necessarily know:
- Which lodges your company actually sells
- Which properties are preferred by your clients
- Which suppliers you trust
- Which routes your team considers practical
- How your company defines “luxury”
- Which accommodation belongs in different price categories
- Which destinations work best at different times of year
- Which services are included or excluded from your quotations
Once that information becomes part of the AI workflow, the output can become much more aligned with the business.
This is why companies should start thinking about AI knowledge architecture, not just AI tools.
Your company’s knowledge may be one of its most valuable AI assets
Most businesses already possess large amounts of useful knowledge.
It may be sitting in:
- Google Drive folders
- Word documents
- Spreadsheets
- Email conversations
- FAQs
- Previous proposals
- Training documents
- Standard operating procedures
- Website content
- Internal notes
Historically, employees have had to search through these resources manually.
AI creates the possibility of making that knowledge much easier to access and use—provided it is organised, accurate and governed properly.
This means one of the first steps in an AI project should not necessarily be buying another AI tool.
It may be cleaning up and organising the information the business already owns.
Data privacy cannot be an afterthought
There is another issue that becomes increasingly important as businesses begin feeding their information into AI systems: data protection.
Client inquiries can contain names, email addresses, telephone numbers, passport information, travel details, payment information and other personal data.
Kenya’s Data Protection Act, 2019 establishes requirements around the processing of personal data, including principles such as lawfulness, fairness and transparency, purpose limitation, data minimisation, accuracy, storage limitation and confidentiality.
The Office of the Data Protection Commissioner also identifies information such as names, phone numbers, email addresses, location information, photographs and audio or video recordings as examples of personal data.
That means businesses should not treat client information as harmless text that can simply be copied into any AI tool.
A safer workflow is to ask:
- Does the AI actually need this information?
- Can identifying details be removed?
- What information is necessary for the task?
- What happens to the information after it is processed?
- Who has access to the connected AI system?
- What permissions have been granted to connected applications?
- Does the organisation have appropriate policies and safeguards?
AI adoption without data governance can create a new business risk while trying to solve an old business problem.
For businesses handling customer information, data protection should therefore be part of the AI workflow from the beginning.
Connecting AI to other business tools
The training also explored the possibility of connecting AI systems with other tools used by a business.
Depending on the platform and permissions available, AI assistants can potentially work alongside tools such as email, cloud storage, calendars and design applications.
This opens up a much larger possibility.
Instead of AI being a separate website employees visit when they need help, it can become part of the existing workflow.
For example, a future workflow might look something like:
- A client inquiry arrives.
- The inquiry is classified according to the client’s needs.
- The AI identifies missing information.
- The system references approved company knowledge.
- A proposed itinerary and route are generated.
- Pricing information is added or flagged for verification.
- A human reviews the proposal.
- The final quotation is sent to the client.
The important point is that AI is supporting a process rather than operating as an isolated tool.
Automation should come after the workflow is understood
There is a temptation to automate everything as soon as a business discovers AI.
That can be a mistake.
If the underlying process is unclear, automating it simply makes the confusion happen faster.
A better approach is:
Map the process → improve the process → add AI assistance → verify the results → automate appropriate steps.
For the safari business discussed during the training, this means first establishing clear definitions, approved data and itinerary logic before trying to automate large parts of the inquiry process.
For example, defining exactly what the company means by mid-range, premium and luxury accommodation can help both employees and AI systems make more consistent recommendations.
AI can also become a business intelligence assistant
The opportunity does not stop at itinerary creation.
Once a company has a structured AI workspace and reliable business information, the same system can potentially support other activities.
For example, AI could help teams:
- Summarise industry developments
- Research new accommodation options
- Analyse recurring customer questions
- Identify gaps in FAQs
- Compare itinerary options
- Prepare internal briefs
- Organise company knowledge
- Generate first drafts of marketing content
- Surface information from internal documents
Scheduled AI tasks can also be useful for recurring information needs—for example, receiving a regular summary of relevant tourism developments or new accommodation information.
The important question is not “What can AI do?”
It is:
“Which repetitive information or decision-support tasks are consuming valuable human time in our business?”
The bigger lesson for Kenyan SMEs
The safari industry is only one example.
A Kenyan real estate company could build an AI workflow around property listings, customer inquiries and viewing requests.
A hotel could use AI to assist with guest communication, FAQs and internal knowledge.
An accounting firm could use AI to organise client documents and draft routine communications.
A professional services business could build a knowledge assistant around its procedures, proposals and frequently asked questions.
A small retailer could use AI to analyse customer questions and identify recurring product or service issues.
In each case, the biggest opportunity may not be “using ChatGPT better.”
It may be turning the company’s existing knowledge and processes into an AI-assisted operating system.
What businesses should do before implementing AI
If your business is still at the beginning of its AI journey, you do not necessarily need to start by subscribing to ten different tools.
Start with the workflow.
1. Identify repetitive work
Look for tasks that happen frequently and consume significant employee time.
2. Document how the task is currently performed
Write down the steps, decisions, information sources and approval points.
3. Identify the business knowledge involved
Determine which documents, spreadsheets, policies, FAQs and databases employees rely on.
4. Separate AI-friendly tasks from human decisions
AI may be excellent at drafting, summarising, classifying and organising information. Humans should remain responsible for decisions requiring judgment, accountability or verification.
5. Establish data protection rules
Decide what information can be entered into AI systems and what must be removed, anonymised or handled differently.
6. Build a small pilot
Do not attempt to transform the entire company at once. Choose one workflow and measure whether AI actually saves time or improves quality.
7. Refine the system
Good AI workflows are rarely perfect on the first attempt. Instructions, knowledge sources and approval rules should be tested and improved based on real outputs.
The future is not simply AI replacing employees
The more practical future for many SMEs is likely to be employees working with AI systems that understand the company’s processes and knowledge.
The employee brings judgment, experience, relationships and accountability.
The AI brings speed, organisation, pattern recognition and the ability to work through large amounts of information quickly.
When those strengths are combined properly, the result can be considerably more powerful than either one working alone.
The lesson from the safari workflow is therefore bigger than Claude, ChatGPT or any particular AI model.
The competitive advantage may come from how well a business designs the relationship between its people, its data, its processes and its AI tools.
Frequently Asked Questions
Can a small business benefit from an AI workflow?
Yes. In fact, smaller businesses may benefit significantly because they often have limited staff and many employees perform multiple roles. Automating or accelerating repetitive knowledge work can free employees to focus on customers and higher-value decisions.
Should businesses allow AI to send quotations automatically?
Not necessarily. For high-value or complex transactions, a human approval step can be important. AI can prepare the draft, but pricing, availability, contractual terms and other important details should be verified before a quotation is sent.
Is it safe to paste customer information into an AI tool?
Businesses should not assume that it is automatically safe. Personal data should be handled according to applicable data protection requirements, the AI provider’s terms and the organisation’s own security policies. Data minimisation and appropriate safeguards are important considerations.
Do I need an expensive AI system to get started?
No. The first step should be identifying a useful workflow. A business can start with a small pilot using tools it already has, then invest in more sophisticated integrations only when there is evidence that they provide value.
Final Thought
AI adoption is moving beyond the question of “Which AI tool should I use?”
The more important question for a business is:
“Where can AI become part of the way we work?”
For a safari company, that might begin with turning inquiries into better itinerary drafts.
For another business, it could be customer service, marketing, reporting, research, administration or sales.
The technology will continue to change. The businesses that gain the most from it are likely to be the ones that learn how to connect AI to their people, processes, data and real business objectives.
AI is not just another tool to add to your business toolkit. Used properly, it can become part of how the business operates.