Photo by Finding Dan | Dan Grinwis on Unsplash

How To Create A Career In Conversational AI

Here Is A Step-By-Step Guide To Set You On Your Path

Cobus Greyling
7 min readMay 11, 2022



I believe anyone can create a career in Conversational AI for themselves using a few simple tools and methods, all self-taught.

What you do need:

  1. Access to a laptop or some kind of personal computer
  2. Internet connection, it need not be super high-speed
  3. Email address
  4. A credit card is useful in some instances, but more about this later…

What you don’t need:

  1. Don’t spend money on courses, online training and material. The best content, documentation and tutorials are free.
  2. You don’t need a super computer with excessive processing power
  3. You don’t need dedicated hardware, for instance a GPU
  4. You don’t need to buy any specific software or have access to costly development environments
  5. You don’t need a credit card to most software environments, you can have full access to chatbot frameworks like Cognigy,, Nuance Mix, Rasa and others.

It is helpful to have access to a good text editor like Notepad++, Large File Editor for big text files, and Visual Studio Code (VS Code).

AI In General and Conversational AI In Specific

Artificial Intelligence (AI) is a highly sought after and popular field currently.

There are disciplines of AI that demand high-end dedicated hardware (NVIDIA) or large amounts of data and long training times. You want to stay away from these areas. The iterations are too large and it could be more error than trial.

A subset of AI is Cognitive AI. This involve areas of cognition. For instance: seeing, hearing (natural language understanding), speaking, facial expressions etc.

Of these, natural language understanding is the most accessible. The process of understanding what a human is saying, and then responding based on the human utterance. This is Conversational AI in simple terms.

The basic process followed in a human-to-machine conversation

Prototype, Write, Publish

The best place to start is to prototype. That is building a chatbot from scratch on a Conversational AI platform. Choose a simple use-case like a bot that tells you the time or the date.

Once you have built the prototype, and you can demonstrate the bot works, write about your experience “I have built my first bot and this is what I have learnt!”. You will be amazed by the feedback you receive and how many people find it helpful.

Publish your story on LinkedIn, Substack, Medium and the like.

You do not need any coding or software engineering experience.

Where To Start

Your timing is perfect, there are a few cloud based, SaaS Conversational AI platforms currently giving free access to their platform. These include Cognigy, & Nuance Mix. You can access these platforms with no credit card or any commitments. This is quite a generous offer and a sure way to get started in the world of Conversational AI.

I would suggest start with Cognigy, and wrestle through the process of building a simple bot and using the documentation to guide you.

Building a working and demonstrable prototype will frustrate you initially, this is due to your brain being programmed. Once you start grasping the concepts and experiencing that breakthrough of a working prototype and iterating on your prototype, there is an immense sense of satisfaction.

With a platform like Cognigy all the work is done in your browser, no need for any software installs or coding. These platforms are all no-code to low-code environments.

Don’t worry about training data, with limited thought-up training examples you can train your first machine learning NLU model.

Your next step might be to explore other functionality within this environment, or start exploring other environments.

Local Installs, Virtual Machines & NoteBooks

Once you have mastered to some degree the cloud-based, browser-bound environments, you might want to install a chatbot framework on your local machine.

This will open a whole other dimension and allow you to start looking at system configurations and more pro-code approaches.

The open-source environments you might want to consider are Rasa and Botpress. You merely need an average laptop or PC to get going with a locally installed chatbot.

Making use of a virtual machine does make sense when you are experimenting and prototyping, but the Rasa demo will take you through the process step-by-step.

Touching on Notebooks…a notebook is an application in a browser, in which you can install and run software. Quick and easy, often with powerful processors, at no cost.

Initially this concept might seem abstract, but start with a simple example, run the code, and in no-time you will get the hang of it.

Whenever I wanted to create a virtual machine, I have made use of AWS EC2 instances. There are free tiers, setup is quick and easy, and when you do not use your virtual machine, you can stop the instance. This is very effective in saving on costs.

Lastly, at this stage you would be well aware of what training data is, and will realise the need to manage training data. Using a tool like HumanFirst can be insightful in how to approach the management of training data.

Getting started with HumanFirst requires no credit card and you have access to 1 user, 1 workspace and 1,000 utterances.

Join Communities

Join communities on mediums like Slack. Rasa has a very active developer and user community. Notable Slack channels are and Rasa.

VUX World and both have exceptional content on their podcasts.

Using Your Credit Card

Be very vigilant when making use of your credit card. The IBM Cloud demands credit card details, but have quite a large selection of free tiers to make use of, especially for IBM Watson Assistant. AWS is also quite economical, if you minimise the uptime of your EC2 instances.

I have found costs on Azure and Azure Cognitive Services can escalated significantly and lead to a shocking bill. With Azure the cost is just not worth the benefit.

On most platforms you can set a maximum on the spending limit, and billing alerts can be set to avoid bill-shock.


We always overestimate what we can do in a short term. And underestimate what we can do in the long term.

Don’t underestimate the compound effect over time of…

  • taking a few hours a day,
  • building a personal experience while building a prototype, and
  • creating clarity for yourself by writing about it.

Your work will be of value to people and you will be helpful, and good things are bound to come from that.



Cobus Greyling

I explore and write about all things at the intersection of AI & language; LLMs/NLP/NLU, Chat/Voicebots, CCAI.