In this article, we’re going to explore various ChatGPT integration concepts and delve into the practical aspects of incorporating AI into your business operations.
Before ChatGPT API: Surf’s experience with AI
Surf’s been working with AI even before ChatGPT API was a thing. Back in 2017, Surf established an AI and ML division, sparked by Google’s proclamation at a conference in London that AI is the future. Firmly believing in this premise, our AI team has successfully completed more than 20 machine learning projects for clients spanning the retail, fintech, and insurance sectors.
But our experience has demonstrated that these projects are most efficient when the necessary data is gathered and validated within the client’s company. Consequently, we gradually phased out commercial projects in this area and ultimately decided to close the department.
At Surf, we’ve been working with cutting edge tech since 2011 — from Mobile to Machine LearningBrowse our cases
The emergence of ChatGPT API was like a breath of fresh air, reigniting our interest in AI and ML technologies. Here’s why:
- ChatGPT API made the model 10 times more affordable. The cost per 1000 tokens using gpt-3.5-turbo is now $0.002.
- The API provides seamless access to the ChatGPT functions. Developers can directly utilize the API endpoint or official OpenAI libraries in Python or Node.js, which are available on Github.
- As of the gpt-3.5-turbo model, ChatGPT can generate over 4096 tokens per request, a significant increase from the previous limit. This allows for more complex interactions with the model.
- The new system parameter allows developers to adjust the creativity of the model’s responses with temperature — a parameter that controls the randomness of AI’s response — enabling it to better tailor it to different applications.
With all of the above, integrating AI into real-world projects has become considerably more viable. In fact, a number of companies have already capitalized on this opportunity.
For instance, Notion AI has integrated autocompletion and AI-assisted documentation into their platform, while Shopify employs AI to automatically generate product descriptions. However, these are merely a few examples amidst a sea of potential applications.
At Surf, we are super excited about the opportunities the ChatGPT API presents. We’ve already begun actively exploring potential applications and have even integrated AI into our own workflow.
ChatGPT API integration: the possibilities
Here are a few more ChatGPT integration ideas we believe to be especially promising — along with our experience with some of them.
ChatGPT-website integration. With its expanded token limit — 8192 for the GPT-4 model, and likely to increase even more — ChatGPT can utilize resources like a knowledge base or company wiki to inform its responses. This capability can streamline helpdesk interactions, enhancing customer experience while easing the support team’s workload.
AI-powered shopping assistants. Ecommerce apps can use ChatGPT for personalized recommendations much like an on-demand personal shopping assistant. This enhances customer experience and potentially increases cart value thanks to upselling. Some companies — like Shopify — are already offering this product.
AI-assisted search engines. Content-rich applications or those featuring extensive product catalogs can incorporate OpenAI technologies into their website navigation. By providing contextual search results, these AI-powered search engines can help reduce churn caused by users struggling to locate the needed information or resource.
Custom AI plugins. Businesses can introduce custom plugins into various software to automate routine tasks, minimize human errors, and streamline specific processes. A prime example of this is the smart recap feature launched in the premium version of Microsoft Teams. Leveraging text-to-speech technology and the GPT-3.5 model, it automatically generates meeting notes, ensuring the entire team remains in sync following lengthy calls.
How we use ML and ChatGPT at Surf
Before proposing ChatGPT integration to our clients, we wanted to assess its viability firsthand. As a result, we’ve developed several internal tools that the entire Surf team has been extensively using over the past few months.
In project management. We experimented with various concepts for utilizing LangChain to streamline the creation of applications based on large language models (LLMs) and chat models. We tackled various tasks with ChatGPT using embedding, which allowed for faster and more efficient processing of vast amounts of information — such as a large collection of tasks in Jira. We discovered a way to simplify data collection and processing for machine learning, resulting in improved training quality.
Marketing and analytics. We also used the Text Completion API to examine a wealth of materials and develop an Ideal Customer Profile (ICP) classification for our company, ultimately gaining valuable insights. We also actively use the GPT for Sheet AI tool—a souped-up version of Google Sheets and Google Docs. This tool allows you to directly utilize ChatGPT within documents, assisting with various text-related tasks such as editing, extracting, cleaning, translating, summarizing, presenting, and more.
Looking to integrate ChatGPT into your project?
If you need a reliable and experienced development partner to help you with AI integration, Surf is here to help. Simply reach out to us using the form below, and let’s discuss your specific needs.