AI Prompting for Kenyan Professionals: Turning Baby Steps into Giant Leaps

Whether Kenyan professionals like it or not AI is here to stay in the workplace. And cards on the table, I’m here to convince you to like it. If the sight of an AI tab open on you colleague’s or intern’s screen makes you sweat, you may be thinking that it will take something from you. Instead, consider what could you gain from it. Like all tools, AI is for human use — not the other way around. And so, the next logical question is: how can you use it? Welcome to AI prompting for professionals.

The Basics

Here are a few terms you may have heard alongside prompt: input, conversion, tokens, output. Although seemingly random, they are key steps in how your AI tool works, in exactly that order. Let’s dive in.

Prompt/ Input:

Essentially the same, these are the way in which you interact with your AI tool. Your input can come in many forms, i.e. natural language in the form of text, images, documents, or videos fed into the tool. Your prompt is the specific instruction that is part of, or in addition to the input such as asking the AI to answer a question, summarize a piece of text, or generate/ review an image, audio or video.

Conversion (aka Tokenization):

Your AI tool does not ‘read’ words or ‘see’ pictures like you and I, and so whatever you give it has to be converted into something it can process—a token. Additionally, it breaks down the data you give it into smaller chunks that are more easily and efficiently processed. Let’s keep going to see why this is important. 

Tokens: 

These are what your AI tool processes in relation to your prompt and mediates between its inputs and outputs. Once your prompt is converted into tokens the AI tool can process your request. A practical example for text-based prompts, the word ‘spaceship’ may be treated as a single token assigned the number 196 or split into two smaller tokens as ‘space’ and ‘ship’ and assigned the numbers 247 and 382. Meaningless to you and I, but just what AI tool needs to work with. I know geeky stuff, but here’s why all this is important to know.

Despite your AI tool’s best efforts at breaking down the data into manageable tokens; too many tokens resulting from lengthy prompts, or tokens that are unhelpful in processing from unclear requests, impact the resources required by the AI tool. This may lead to longer response times, expenses for higher computer processing needs, and at the extreme a waste of efforts in a “garbage input = garbage output” scenario. So, your prompts need to be short and clear. Sit tight I’ll guide you through it in just a minute.

Output: 

After processing the tokens, the AI tool generates response tokens and converts them into a ‘human-readable’ format supported by the AI tool. ‘Human-readable’ here means text, image, video or another format that you or I can understand. This is your output.

All the AI tools out there generally follow the same steps described above. And it’s okay to admit — it’s a lot less scary once you peek behind the mask.

Prompting like a Boss (or, For Your Boss)

The Bible, for all its potential divisiveness, does offer universal wisdoms. 

“What has been will be so again,
 and what has been done will be done again;
 there is nothing new under the sun.”
– Ecclesiastes 1:9

The challenge of AI prompting is really a historical problem. During the rise of search engines like Google in the early 2000’s, many struggled with how to properly search the internet. Even in 2025 the concept of ‘Boolean Searches’ — critical to effective search engine queries — may be new to some readers. Long before computers any kind of research required asking the right questions, often in the right way.

Today you could break down AI prompting into the same two blocks:

• prompt engineering as asking the right questions, and
• prompt optimization as asking the right way.

Prompt Engineering the Right Questions

As explained earlier, lengthy or unclear prompts may cause longer response times, increased costs from higher computer processing needs, and wasted efforts from “garbage inputs = garbage outputs”. 

What you really want to do, is reign the chaos in your workflow and provide your AI tool with clear guardrails to facilitate your work. Here’s how you can make that happen. 

1. Be Clear & Specific
Keep things simple. Narrow down on what you want the AI tool to help you with into a brief prompt. If you need multiple things, break them down into multiple prompts. If the multiple prompts are related, use them one after the other to build up a structured response. As for specifics, if you want a document summarized consider; into what length, and whether it is a general summary, or only a summary of specific information among other things. The clearer your input, the clearer your output.

2. Add Context 
Giving your AI tool context of what you are looking for is a great way to improve its output. A small addition like ‘in Kenya’ at the end of your prompt can guide the AI tool to generate responses that assume a Kenyan context and provide greater accuracy and relevance to your output. Stating ‘for a high school student’ could lower the complexity in the information generated and how it is explained.

3. Instruct your AI tool
Guide your AI tool along the right path by telling it what to do and what not to do. For even deeper analysis or specialized insights, provide it with a persona for it to act as. An example of both would be: ‘Advise me only on the drawbacks, as a marketer/ lawyer/ accountant’.

4. Balance between open-ended and directive
While specificity is important, be sure not to shut out potentially helpful information and analysis from your AI tool. Craft your prompts widely for general outlooks, and narrowly towards specific outcomes according to your ongoing needs.

5. Set Your Presentation & Tone
Finally, tell your AI tool how to display the output in a suitable format. This can range from bullet points, to a report, or a slide presentation. For image generation, terms like ‘playful’, ‘professional’ or the now popular ‘Studio Ghibli’ may help. The tone within the presentation format is another important consideration. An ‘easy-going’ tone for ‘working professionals’ may suit a team building speech, while ‘formal’ for ‘undergraduates’ better suits lecture notes.

Prompt Optimization the Right Way

While asking the right questions certainly yields results, asking them the right way can make the results more useful, or better suited to a professional firm’s specific needs. Standardizing the right way within a firm creates a framework for long-term best practices.

1. Practice with Tests & Refinement
This is not a long or complex process. You may have previously used a prompt that didn’t get all the answers you wanted at first — so you adjusted it in tone or context for better results. That’s testing and refinement in action. In a professional setting with specific and repetitive workflows, your refined prompt becomes valuable for future use.

2. Use Examples 
Examples are specified input and output instructions that form part of a prompt. They are also useful for subsequent prompts to reduce ambiguity and improve consistency in future tasks. If you ask your AI tool to translate sentences to follow into Swahili, it will do so for any individually prompted sentences that follow outside of those you included with the instructions. If the translation was to Kenyan Swahili, consistency is maintained by avoiding a further prompt such as ‘translate into Swahili again’ which may lead your AI tool to start afresh but with Tanzanian Swahili.

3. Set Up Scaffolding
Scaffolding is a structure for your AI tool to follow and assists in generating more reliable outputs. ‘Produce a report on this data’ could still leave much to be desired in terms of format. Specifying sections to the generated report, steps to be followed, and the reasoning process for analysis of information presented gives your output better structure. In professional firms, memos, reports, and other documentation may have unique in-house formats that AI responses will need to fit into.

4. Create Best AI Practices with Prompt Templates
Finally, prompt templates can be created and placed in a repository of a professional firm’s ‘best AI practices’. Templates are especially useful for standardizing use of AI tools across teams or workflows — improving both efficiency and consistency. Essentially, a prompt is pre-prepared using knowledge from the 3 optimization steps above, with a few blanks for information input. A summary of the day’s finance news for circulation within a professional firm could be generated in a desired format by simply adding the web address or RSS feed of several news agencies into a pre-set prompt. 

Conclusion

Apprehension about AI in the workplace has a lot to do with a fear of the unknown. Having a better understanding of how AI tools work not only eases these tensions but also provides the baby steps into harnessing their potential. Professional firms can make giant leaps in the quantity and quality of their work by managing the efficiency, accuracy and consistency of workflows using AI tools. However, a fair amount of AI upskilling is required before professional firms can mature from baby steps to giant leaps — and that’s where we at KRL come in.

Kiende Research Labs LLP is at the forefront of developing tools for law firms that are at the bleeding edge of innovation. Contact us at hello@krltech.com to find out more.

This article is part of a monthly series of commentaries on the state of legal technology and AI development in Kenya.