Implementing bots in an organization

Different ways to train chatbots

Before implementing chatbots in the organization, it is worth taking a look at the technical infrastructure behind the bots. Bots can be programmed and trained in two different ways: rule-based and self-learning.


Implementation of Chatbots

There are several ways to develop and implement chatbots within an organization. They can be roughly broken down into the following three approaches.

1. Programming a chatbot in-house

In this approach, an organization programs the bot mainly from scratch and only uses a few small building blocks (“libraries”) contained in the programming language. The bot can be tailored to the company’s existing technical infrastructure.  


  • The company is developing its own system and can build up a lot of expertise and fully customize the bot. 
  • No internal information is shared with any outside parties. 


  • Chatbot can only be trained on internal data, that development is much more time-consuming than an external solution.
  • Lack of experience when implementing the chatbot for the first time. 


2. Using external partners to develop a chatbot system

More often companies collaborate with an external specialist in chatbot systems. Often these are start-ups dedicated to the specific field of conversational automation. Even large companies work on the basis of these external services as they neither have the capacity nor see the need to build their own systems. 


  • Faster implementation as external contractors contribute existing tools for the company to work with
  • Being able to rely on the contractor’s experience from previous projects.
  • Fewer internal resources are tied up as most of the work is done by external contractors.


3. Using external large-scale bot-building frameworks

The most practical way to implement a chatbot is to rely on existing frameworks from large tech companies. Several such frameworks exist for different purposes. The best-known examples are Microsoft Azure, Google Dialogflow, and IBM Watson.

When implementing these systems, companies can also work with start-ups. They often offer specialized solutions built on these large-scale frameworks and have prior experience of implementation.


  • Most efficient and sustainable way to set up a chatbot as these companies have acquired a wealth of expertise in the field of conversational technology


  • Potential drawbacks in terms of data security because the data is often stored or processed on servers in the US
  • Difficulty to compete with leading tech companies, that have more expertise

Exemplary frameworks for building chatbots


  • The research project was headed by Prof. Stefan Stieglitz and Florian Brachten  (University of Duisburg–Essen) from 2018 to 2020.
  • It is one of the first studies in Germany to provide insights into bots for communication experts. 
  • For the first time, the researchers analyzed millions of social media posts to find out whether social bots try to influence the social media comunication of the top 30 German corporations (DAX-30 stock index). 
  • Secondly, in-depth interviews with representatives of companies and consultancies were conducted to find out about the scenarios in which chatbots are already used.


Downloads und further readings



  • Reeves & Nass (1996): The Media Equation: How People Treat Computers, Television, and New Media like Real People and Places.

  • Gentsch, P. (2019): AI in Marketing, Sales and Service. How Marketers without a Data Science Degree can use AI, Big Data and Bots.


  • Ross, B., Pilz, L., Cabrera, B., Brachten, F., Neubaum, G. & Stieglitz, S. (2019). Are social bots a real threat? An agent-based model of the spiral of silence to analyse the impact of manipulative actors in social networks. European Journal of Information System (EJIS), 28(4), pp. 394–412.
  • Bot or not? The facts about platform manipulation on Twitter.