Over the past few weeks, the world has been buzzing with chatter about ChatGPT. It is becoming a media sensation and it gained over 1 million new users in its first pilot week.
For those of us at MeBebot who have immersed ourselves in the advancements of Artificial Intelligence (AI) for years, specifically chatbots and conversational AI, this is an exciting time for the world to realize both the variety of applications of AI and the implications of this type of technology for humanity.
For business users within companies, it also creates curiosity about the possible ways to extend using AI within organizations. However, it can also surface more questions and potential concerns about the privacy, security and accuracy of information that is generated, both internally and externally.
What is ChatGPT?
ChatGPT (Chat Generative Pre-Trained Transformer) is an AI language model that can carry out natural language generation tasks (such as answering basic questions, composing emails, essays, and software code). It was launched by OpenAI and trained on massive amounts of data scraped from the internet up until 2021. The model is continually being monitored and fine-tuned for specific tasks, labeled, and tagged by humans and leverages generative AI techniques.
ChatGPT can create “echos” of what it has ingested, but it does not understand fact vs. fiction or truth vs. reality. However, it can learn and become “smarter” over time. And since its language model is tagged and trained by humans, ChatGPT has inherited biases, implicitly encoding common racial and gender stereotypes.
What is MeBeBot?
MeBeBot is an AI enterprise software-as-a-service (SaaS) solution, built specifically for business applications, to digitally transform workplaces to become more operationally efficient using conversational AI.
MeBeBot’s AI Intelligent Assistant is a Chatbot that installs as an app in Slack, Microsoft Teams, or as a web chat, designed to be the employees’ “digital front door,” (or the end user interface) to gain access to valuable information within organizations.
Employees can ask MeBeBot common questions for a variety of internal business questions like:
- General: What are the company holidays? When is the next company all hands meeting? Where can I find an employee directory? Do we have a remote work policy?
- Payroll: How do I locate my paycheck receipt? Can I enroll in direct deposit? When do we get paid? How can I access my W-2 form for taxes?
- Benefits: When is the annual open enrollment for benefits? How do I get a new insurance card? How do I request parental leave? Are annual exams covered by our medical insurance?
- IT: How do I use the VPN (Virtual Private Network)? Do I need to have an app on my phone to get my email? How can I order a new computer monitor? What is the guest Wi-Fi password?
- Facilities: How do I get a badge for the office? Do I have to reserve a desk? Are there COVID procedures I must follow? What’s the policy for bringing in a visitor to the office?
While these types of questions are common to all employees (no matter what company you work for), what is unique to each company are their answers to these questions.
How is ChatGPT different from MeBeBot (and vice versa)?
- AI Language Models and Training of Data Sets
Since ChatGPT does not ingest or scrape content “behind the firewall” of companies (from company Intranets or shared drives for documents like employee handbooks, benefits documents, IT reference guides, etc.), ChatGPT has not been trained nor does it have the content in its language model to answer these types of questions.
Examples of ChatGPT’s errors when attempting to answer employee specific questions:
MeBeBot’s AI language model is trained by employees asking questions against a contained knowledge base, curated to be specific to business users and employees.
Examples of MeBeBot’s Answers to Employee Questions:
*Using Microsoft Teams:
Instead of “scraping” company Intranets or outdated content on company shared drives (from policies, to handbooks, to instructions), MeBeBot provides each customer with an instance of our web-based Customer Admin Portal.
MeBeBot’s Customer Admin Portal contains over 350 default answers to the frequent questions employees ask within business (from HR, Payroll, Benefits, IT, Facilities and general business questions). Each customer’s authorized business user simply completes Placeholders (or a text field of information) that makes their business unique. The answer is automatically updated to include this company specific detail in its answer. All answers can be further edited and updated in real time (by the authorized business users), so that employees receive answers to their questions that are consistent, accurate, and adhere to company policies and procedures, to keep up with the rapid changes at companies today.
MeBeBot’s solution uses natural language processing and applies supervised machine learning, where humans actively analyze and guide MeBeBot’s AI language model training. This methodology ensures that authenticated users are asking the types of questions that are relevant, from around the globe, and these questions asked improve the solution for all MeBeBot users.
2. Compliance of employee privacy, security, and accuracy of answers.
ChatGPT, in its current state, is gaining accuracy in producing content to users based on a general query or statement. It can provide an outline or a rough draft that users can manually update and improve on for accuracy of responses or to get to the desired content state. ChatGPT does not have safeguards for personal information or privacy and currently, it is not compliant with enterprise industry standards.
The best applications of ChatGPT for business are for writing blogs, content for emails, websites, or other marketing materials. However, it is often producing this content out of context. An example is while software code can be written using ChatGPT, it lacks a full understanding of the actual use case for the code, and it requires technical review and editing.
With MeBeBot, our customers have complete control of the answers to their employees’ questions. Using MeBeBot’s Customer Admin Portal, non-technical business users (with permissions) can view the answers employees will receive and they can enhance or improve the answers at any time to ensure accuracy and that the information is compliant to their business.
3. Costs to the end users or to the business
ChatGPT’s parent company, OpenAI, states that it is in its research and feedback-collection phase. The big draw to ChatGPT right now is that it is currently open to the public and free of charge during the pilot stage of this solution. However, last week, OpenAI announced it is moving to a paid subscription model, starting at $42/month per user for their “pro” version and it is estimated that it costs $100,000/day to run ChatGPT in Microsoft Azure.
MeBeBot is affordable to any size organization, at a fraction of the cost of ChatGPT and other types of enterprise chatbot solutions for employees. MeBeBot is a software-as-a-solution and each customer’s subscription is based on a per employee per month basis, that is less than a cup of coffee per month.
While ChatGPT and other types of generative AI (or algorithms that are used to create all types of content) are still “learning,” most consumers have been using AI in their day to day life for over a decade. Conversational AI, both in speech/audio messaging, has been advancing rapidly and devices like Alexa can not only answer your general questions, but Alexa can also place a grocery order for you.
The team at MeBeBot is passionate about the employee experience, as technology for the workplace should enable employees to be efficient and remove frustrations so they can focus on the work that matters to the business. Companies can save countless hours of unproductive time each week when employees use MeBeBot. And, best of all, MeBeBot also helps drive the ROI value of other software and tools purchased by the company, driving adoption to other programs and benefits offered to create a “digital front door” for your employees.
Click here to schedule a one-to-one meeting.