This article shares some of my observations over the years as a software architect with my own Chatbot projects.
The number one reason I have seen past Chatbot projects fail:
- Expecting end users and customers to test a Chatbot.
This tactic sets up a company for embarrassing failures like the infamous racist Chatbots failures. It also encourages curious competitors or technology haters to present your Chatbot as a failure with examples of their own personal testing with criteria that you did not prepare for nor think about in the big picture.For a brand new Chatbot software project, I typically schedule at least a month of weekly milestones / tasks for the testing phase (about one week for older long established software projects). During the software testing phase I create my team of testers (searching for specific skills & interests, qualifying testers, inviting testers, managing testers resources, defining test requirements, defining test failures, defining test successes, gathering test results, analyzing test results, scheduling updates based on test results, iterating the process and much more).Chatbot (and any software) projects that do not include this managed testing phase usually fail. Software projects that attempt to force testers in the middle of the development phase also fail because the development work never gets completed with all the unscheduled & unqualified interruptions.
Solution to Avoid Chatbot project failures:
An experienced Chatbot Architect leading the Chatbot Project with Supervised Learning and incentives to ensure the project’s success with weekly accountable documented milestones.
Weekly Milestones
I typically use this iterative process similar to my Mobile App Projects with weekly documented milestones to track, hold myself & the team accountable and it helps manage the entire project with remote teams.
Chatbot Project Plan
I include clearly defined project and business goals, equipment, resources and very specific definitions of success, positive roi and how that success is analyzed and measured. This becomes part of the Software Design Documentation for my team to refer to as they are working and making technical decisions. Each technical decision affects the entire team and every end user, so it’s important the whole team has access to this vital information and the entire team has incentives for the project to be successful.
Know Your Audience
A clear definition of your target market: who, what, where, when, why and how. This is marketing 101 and your Chatbot project will probably fail without this info in your internet business plan. Hire help with this if you are not familiar with sales, marketing and human nature (how and why will humans give your team money to use the amazing Chatbot solutions you are dedicating your time & resources into providing?). These definitions influence the Software Design Documentation to help me and my team remember how the Supervised Learning responses and features should behave. It also influences the testing phase and ensures that testers are not allowed to steer the entire project into an entirely new direction that is not hyper focused on catering to our relevant target audience (Say No to Scope Creep!).
Chatbot Software Design Documentation
As the Chatbot Architect I refer to the business & marketing plans ( target market definition, user stories, business logic) to develop the Chatbot Software Design Documentation. The Chatbot Software Design Documentation is created in the first week of the project. It helps with project management, work schedules, costs and more with weekly accountable measurable milestones that include software design, development, internet security, webhosting, knowledge bases, setting up software teams for alpha testing, beta testing and finally public access publishing, maintenance, maintenance and sales.
Self Hosted Open Source or Proprietary : Higher Costs. More Control.
Self hosting custom open source and proprietary Chatbot solutions require much more resources, time and money to setup, install, maintain, update and debug with the typical annual security issues for any self hosted solution.My personal favorite chatbot projects include AIML opensource solutions such as AliceBot because I am a software architect, curious researcher and entrepreneur that really enjoys coding & experimenting. I was able to convince a tech competition to allow my Mohawk language Alicebot to compete and won the Bronze medal. The first chatbot to earn the Loebner prize is also an AIML AliceBot; ‘Mitsuku’ created by Steve Worswick. Worswick’s Mitsuku bot has earned the Loebner prize twice now.The AliceBot solution is not great for every situation but it is a great solution for a smart team that is able & willing to invest the requirements to be competitive. The downside for this option is your company will be paying much more in time, resources and money to setup, maintain and deal with the security issues.
A Few Success Requirements:
- Clearly defined Chatbot project goals, requirements and definitions of success and a positive roi.
- A dedicated Chatbot Architect with personal incentives for it’s success.
- Daily Supervised Learning Tasks.
- Chatbot Software Design Documentation.
- Daily Chatlog & Security Review.
- Daily Knowledge Updates (then weekly, monthly, annually as Chatbot gets older). This schedule coincides with specific marketing events that bring specific new target markets into the conversations.
- Team of Specialists: PR, Sales, Marketing, Finance, Legal, Hosting, DevOps, Internet Security, Business Admin…
Cloud Chatbot Solutions : Lower Costs. Less Control. Less Incentives to be Unique
Many companies provide templates with standard Chatbot personalities that can be setup and published live within minutes. The setup is usually a free trial with low monthly hosting fees. I find these solutions are also fun and allow me to focus much more time & resources into the daily maintenance tasks. Daily maintenance with supervised learning, reviewing chat logs, responding to any issues immediately creates a winning project.I also appreciate that these solutions do not require me to focus on devops, back end issues, internet security updates and all the issues that the self hosted solutions require from me as an entrepreneur and as a software architect.I can focus on reviewing the chatlogs daily and updating the Chatbot knowledge base as needed. The downside for this option is it removes incentives for me as a developer & entrepreneur: all my work developing a unique Chatbot personality with relevant responses to my own target market is now accessible to a third party for their own company’s benefit while I pay the costs for building & maintaining a new unique Chatbot personality. I tend to build unique and award winning Chatbot knowledge bases with self hosted solutions while building standard run of the mill sales / support Chatbots with the cloud hosted solutions.
Are you planning to add Chatbots to your internet business?
If yes, do schedule a consultation with me to find out if I can help you and your team as a Chatbot Architect. I build and integrate Chatbots with websites, mobile apps, retail animatronics, smart home tech, iOT, games and much more. My Portfolio, Resume and LinkedIn.