By now, chances are that we are all thinking of Generative Artificial Intelligence (Gen AI) and how we can leverage it in our personal and professional lives. It is commonly accepted that Gen AI will disrupt any. The potential enterprise use cases are limitless—from Innovation to Customer Service to Product Design, and the list goes on. Customers do not want to be left behind and start POCs and projects around Gen AI. At the same time, they are a bit unsure—they do not fully understand how Gen AI works, how they would train the Large Language Model (LLM) with their data, are worried about security, loss of data privacy, and how they start their Gen AI journey.
This blog seeks to educate the reader about Gen AI, the potential Use Cases, some of the infrastructure that can be leveraged, Costs, and how to get started on a POC. The ideas below have come after talking to Unvired’s customers, our CTO, and doing research that includes articles from the Harvard Business Review and Bain Capital Ventures. The blog has not been created by Gen AI.
Traditional AI versus Gen AI
Many of us already benefit from deploying traditional AI projects or what can also be described as Predictive AI. This type of AI is good at algorithms that can examine time series data and predict outcomes or perform classifications. While traditional AI has been in use for quite some time, it is only in the last year that Gen AI is beginning to make its mark.
What is Gen AI?
Gen AI includes algorithms that can be used to create content, including text, images, audio, code, and videos. It falls within the broader area of Machine Learning, and ChatGPT is one of the most popular examples of Gen AI. Some popular Gen AI systems include:
Major software vendors have also introduced Gen AI in their applications as below:
Potential Use Cases
There are several Use Cases:
- Document Search—Get answers to questions, get a summary, extract information
- Chatbot/Virtual Assistant for Customer Service, Field Service, CRM, or Asset Maintenance: Mine the Knowledge Base of the company to get answers and information instantly
- Sales: Generate emails for customers and prospects based on context and history
- Product Design: Create text and images for new product designs
- Innovation: Generate and evaluate new ideas
- Marketing: Write Marketing blogs
- Code: Generate code and increase developer productivity
- Learning/Training: Generate quizzes from training material, summarize documents, answer questions
Vendors for Gen AI
There are a lot of technology providers, including OpenAI, Microsoft, Google, Amazon, Salesforce, SAP, Oracle, and ServiceNow, as well as many smaller vendors and start-ups. We provide below a few links as references:
Costs of Gen AI
The cost of Gen AI depends on the model used, and there are various price points. For example, Open AI charges per 1,000 tokens, where a token can be taken as pieces of words, with 1,000 tokens being about 750 words. For the GPT-4 model with 8k context, the price is $.03 per 1,000 tokens for Input and $.06 per 1,000 tokens for Output. Below is a link to what OpenAI charges:
The costs for a Gen AI POC should be affordable for most enterprises.
A Quick Start POC
We suggest the following steps to get started on the Gen AI journey:
1. Select the Use Case: Service-related and Analytics business scenarios seem apt for this.
Customer Service, Field Service, CRM, or Asset Maintenance service. Document repository search to answer questions, retrieve information, or summarize information using a Virtual Assistant/Chatbot interface has the power to Simplify user interaction.
Gen AI can save hours in creating reports and dashboards. Microsoft Copilot embedded in Power BI can be used to create content like dashboards and also be used to ask queries via a conversational chat.
Embedded Gen AI in Business Applications: Salesforce, SAP, Microsoft, Oracle, and others have embedded Gen AI in their core business apps. This is a natural starting point to explore Gen AI.
2. Get Ready: It is important that you dedicate some time to this and have the talent to sustain these Gen AI initiatives. Do you have the datasets that you can leverage? What about a budget?
3. Choose a Vendor: One may elect to build an LLM from scratch, but that would need time, investment, and talent. Data privacy, security, and other considerations may create the need to build your LLM in some cases. However, we recommend starting with an existing Foundation LLM (open source or closed), fine-tuning it, and leveraging prompt engineering for specific use cases. The vendor options range from OpenAI, Microsoft, Google, and Amazon to application vendors like Salesforce, SAP, Oracle, and Microsoft Dynamics/Power Platform.
4. Start Small: It is best to start with a low-risk business scenario and start small. As you learn, you can then grow the POC to a pilot.
Schedule a call with us if you would like to learn more about generative AI in the Enterprise.