Why businesses are watching ChatGPT

Chatbots have been around for a while. But ChatGPT introduced users around the world to an all-new level of performance — and its performance improves as the user list booms.

In fact, SimilarWeb data showed that ChatGPT is the fastest-growing app, ever, to hit 100 million users.

ChatGPT (chat generative pre-trained transformer) is powered by technologies that are part of artificial intelligence (AI). It uses AI natural language processing (NLP) to understand the content that users enter and form the text of its responses. Its responses will improve as its use grows because it uses another type of AI, machine learning (ML), to analyze user entries and further refine its decision-making models, making them more accurate, clear and human-like.

The unprecedented effectiveness of ChatGPT has triggered a race to find new business uses and bring new competitors to market. Other “large language model tools” that are already in the market or rumored to release soon include YouChat, Jasper, Chatsonic, OpenAI Playground, Perplexity AI, Character AI, PALM, BLOOM, Google Bard and LaMDA, Meta Opt and DeepMind Sparrow.

The biggest deal so far

So far, the biggest business deal in this space has been when Microsoft paid $10B for a 9% stake in OpenAI, the creator of ChatGPT. Microsoft followed that up by starting to integrate ChatGPT with Bing, Azure and its Office product suite. The apparent aim is to elevate customer experiences and capabilities across a range of applications.

Beyond end-user experiences, large language model tools can help customer service representatives quickly craft personalized responses or interact in other ways that help boost customer engagement and retention.

However, early integrations of ChatGPT for business purposes have exposed the raw edge of the growing technology, yielding news about bizarre responses and errors. The viability of the new integrations, and ChatGPT itself, is still evolving.

The initial use cases

There’s still a lot of ground to cover before businesses can consider completely outsourcing their higher-level support, marketing, sales or other customer-facing functions to an AI tool.

Customer relationships, and your brand, are simply too important to gamble on unpredictable interactions. However, there are some more internal tasks where the latest generation of AI tools can help accelerate work:

• Technology services — coding and structure
• Healthcare — accelerating research across volumes of information
• Asset management and financial firms — regulatory research legwork, with verification from subject matter experts

The current AI tools provide answers that still require close verification. But, some of their value lies in their power to analyze masses of data. Some businesses might be ready to consider where or how they could tap into that power.

The phases of maturity

If you think you have a use case for an AI large language model tool like ChatGPT, start by identifying your current maturity in AI language model solutions. Think of your maturity in phases: 

• Phase 0: Make sure that you have the data, people, processes and technology in place for your AI solution to succeed. An AI solution that performs poorly can quickly erode the trust of your customers, partners and others across the business spectrum.   
  • Data quality: You need clean and current input data. Make sure that your AI solution will not ingest outdated manual exports or siloed data that has inconsistencies, redundancies and other errors that could contaminate results.
  • Data security and privacy: Your cybersecurity and data privacy plans are designed to keep data secure within your established environment. An AI solution can introduce new factors into that equation, so make sure that you audit any new data risks from the solution and update plans accordingly. As the landscape of cybersecurity and privacy laws continues to change, you need to be sure to understand exactly how and where data is used.
  • IP infringement: Large language models can sometimes ingest or reference content that is proprietary. Make sure to consider whether there are scenarios where your solution might infringe on intellectual property rights.
  • Governance: Consider all of these factors as you develop controls to ensure the ongoing quality, security, privacy and availability of the data that your solution needs in order to deliver the output you require. Identify the people or teams who will shepherd this process, perform research, provide feedback and ensure the solution’s ongoing performance.
• Phase 1: Establish your organization’s basic competency with a small and limited-risk project. For instance, start with an isolated process or scenario to address, identifying your anticipated returns, change management and any revisions to your governance.
  • ROI: Define an identifiable metric and goal for your ROI.
  • Change management: Establish change management for development and post-solution enhancement.
  • Governance: Test the governance framework identified in Phase 0.
• Phase 2: Grow a pilot program for small-scale implementations. This can have several goals, as your organization’s AI capabilities mature.
  • Experience: Build your organization’s experience and confidence for implementing and using AI solutions.
  • Proof points: Adjust models to increase your ROI and prepare to share your results as proof for larger business cases.
  • Integrations: Identify other tools that you could integrate to enhance the solution.
  • Governance: Continue to adjust your governance framework as needed.
  • Productivity: Plan for productivity shifts, as the meaning of productivity in the workforce shifts.
• Phase 3: Move to large-scale complex implementations, integrating multiple machine learning algorithms to generate unique value.
  • New use cases: You can use multiple forms of AI together for greater productivity. For example, consumer or employee profiling applications can help inform your language model with context, for quicker learning and more accurate output. Or consider building smart presentations by combining ChatGPT with Tome.
  • Growing with AI: Look at how your team can develop the skills to teach language models and function as stewards of the content. For example, is there content that you are currently outsourcing which you can generate with a large language model? This can take advantage of your existing project management structure while accelerating the content creation process.
  • Simpler interfaces: Consider whether you can use a large language model like ChatGPT to serve as a simplified interface, using its API to connect to the functionality offered by other solutions.

Your organization might be in different phases of AI maturity across various teams. Discussions about governance can help you align on consistency, and ensure your organization’s overall growth at an even pace across the subsequent phases.