Use Cases for ChatGPT - How to use LLMs For Your Business
Software ate the world, turned it into data, and now AI is not just digesting it but translating it into unprecedented value. It's been doing this for years now. Except for the first time, AI can be used by just about anyone, accelerating digestion. This monumental shift is partly attributable to Large Language Models (LLMs) like ChatGPT, which have democratized AI, making it accessible even to those without a technical background.
For the first time, every person—regardless of their tech expertise—wields the transformative power of AI simply by talking to it. In this new world, LLMs have shifted the balance of power from those who can execute to those who can ideate. But with great power comes great responsibility. Think of LLMs as chainsaws – incredibly potent in skilled hands, yet potentially hazardous if misused. This is why 55% say they are more worried than excited about the future of AI, according to a recent Dataiku + Databricks Survey of senior AI professionals.
You can do amazing things with this new superpower if you can control your chainsaw to build things without hurting yourself and others. The big question on every business leader's mind should be how do we implement an enterprise generative AI strategy to amplify our teams' human potential and productivity in a way that's governed, secure, and adopted as quickly as possible, or risk being left behind. 64% of organizations will “Likely” or “Very Likely” use Generative AI for their business over the next year. ChatGPT is too useful to ignore. Delaying a company strategy could easily result in your team using it at home on their own terms because it makes their lives easier. Time is of the essence. You need a roadmap. Here's how to seamlessly integrate LLMs into your business strategy in 7 concrete steps:
1. Identify Use Cases for ChatGPT and other LLMs
Implementing large language models in an enterprise setting can significantly boost productivity across various lines of business (LOBs). The extent of the benefit largely depends on the specific tasks, data types, and the current inefficiencies present in each LOB. Let's explore the potential impact of each.
Customer Support
Potential: Very high. Boston Consulting Group (BCG) estimates LLMs like ChatGPT will increase the productivity of customer service operations by 30% to 50% — or more.
Reasoning: This LOB often deals with a massive volume of repetitive questions. LLMs can handle many of these inquiries, providing instant responses to customers and freeing up human agents for more complex issues. They can also assist in drafting responses, guiding newer agents, or analyzing feedback for continuous improvement.
HR
Potential: High. In a recent Gartner survey of HR professionals, 84% believe that generative AI will make existing HR activities more productive, while two-thirds think generative AI will eliminate redundant activities within the function.
Reasoning: HR receives numerous queries about policies, benefits, procedures, etc. An LLM can automate responses. It can also help in screening resumes, crafting job descriptions, or providing training materials.
Software Development
Potential: High. A recent McKinsey study shows that software developers can complete coding tasks up to twice as fast with generative AI.
Reasoning: LLMs can assist developers in code documentation, understanding error messages, suggesting code snippets, and even automating repetitive coding tasks. They can also be used for bug report analyses.
Legal
Potential: Medium to High. According to the results of a survey released by LexisNexis in March 2023, 84% of respondents in the legal field believe generative AI tools will increase their efficiency, while a majority believe it could advance and revolutionize the entire practice of law.
Reasoning: Legal teams deal with vast amounts of text, including contracts, regulations, and case law. LLMs can assist in drafting, reviewing, and summarizing these documents, though human oversight is essential due to the critical nature of legal work.
Marketing
Potential: Medium to High. According to a Salesforce.com survey, 60% of marketers say generative AI will transform their role.
Reasoning: LLMs can assist in content creation, analyzing customer sentiment, and responding to social media inquiries. They can also help brainstorm sessions, draft marketing messages, and understand industry trends from unstructured data.
Sales
Potential: Medium. According to another Salesforce.com survey, 84% of salespeople using generative AI say it helps increase sales at their organization by enhancing and speeding up customer interactions.
Reasoning: LLMs can help generate reports, respond to customer emails, and create sales pitches or proposals. They can also analyze customer feedback to inform sales strategies.
R&D
Potential: Medium.
Reasoning: While the core research often requires human intuition and expertise, LLMs can assist in literature reviews, summarizing findings, and drafting research documentation.
Operations
Potential: Medium.
Reasoning: LLMs can optimize processes, assist in employee training, and answer operational queries. They can also help analyze feedback from the ground level to make improvements.
Engineering
Potential: Medium.
Reasoning: Similar to software development but in a broader context, LLMs can help in documentation, error analysis, and even in some design brainstorming sessions.
Finance
Potential: Low to Medium. According to a KPMG Generative AI Survey, 59% of executives said their organizations are using emerging AI technology in their finance or tax departments.
Reasoning: While a lot of financial data is structured, LLMs can assist in generating reports, forecasts, and explanations in human-readable formats.
Procurement
Potential: Low to Medium. According to Deloitte, Generative AI’s greatest potential in source-to-pay is likely proactive risk management, process automation, and decision-making, and three out of four Walmart suppliers prefer negotiating with AI over a human.
Reasoning: LLMs can help in vendor communication, contract reviews, and demand forecasting from unstructured data sources.
Manufacturing
Potential: Low.
Reasoning: While there are many automation opportunities in manufacturing, they often lean towards other types of AI and robotics rather than LLMs. However, LLMs can assist in training, process documentation, and safety guidelines.
Generally, LOBs with high engagement with textual data, repetitive queries, and manual documentation will see the most immediate and significant productivity gains from LLMs. However, it's essential to remember that each company is unique, and what works best for one enterprise might differ for another. A careful analysis of each LOB's specific tasks, pain points, and data types is crucial before implementing LLMs.
Prioritize and focus on the lines of business (LOBs) with
High-volume manual tasks
Repetitive questions or processes
Large sets of unstructured data
Why should you focus on unstructured data use cases? Structured and unstructured data serve different purposes and are processed differently by LLMs and other AI models.
Unstructured Data
Nature: Typically consists of text, images, audio, video, etc. This can include emails, documents, social media posts, etc.
Utility for LLMs: LLMs, like ChatGPT, are primarily designed to handle and generate human-like text. Thus, they excel at parsing and understanding unstructured textual data.
Applications: Content creation, sentiment analysis, customer support chatbots, and more.
Structured Data
Nature: Organized in defined fields, typically found in databases, spreadsheets, etc. Examples include customer names, purchase histories, product prices, and more.
Utility for LLMs: LLMs can use structured data combined with context or generate human-like interpretations or summaries. However, traditional data analytics, machine learning models, and database query systems often better handle purely structured data.
Applications: Predictive analytics, data visualization, database queries, and more.
To increase employee productivity and extract value from business data using LLMs, unstructured data offers more direct opportunities because LLMs can help automate, summarize, generate, or interpret this type of data in ways that humans naturally communicate.
That said, structured data is immensely valuable. It's crucial for many enterprise operations and analytics. The distinction is that LLMs naturally shine when dealing with the complexities of unstructured data, whereas other AI or traditional data-processing tools are often more apt for structured data tasks. However, a holistic strategy could involve LLMs working in tandem with other systems, extracting insights from structured data, and communicating them in a human-like manner. For a deeper dive into structured vs unstructured data and the leading storage technologies for each, see Snowflake vs Databricks: Where Should You Put Your Data?
2. Select LLMs for Your Business
Proprietary Models (like ChatGPT)
Pros: High performance, continuous updates, managed services.
Cons: Cost, potential data leak concerns, lack of customization.
Open-Source LLMs
Pros: Customization, no external data handling, potentially lower costs.
Cons: Requires internal expertise, infrastructure setup, and maintenance.
3. Data Security & Compliance
Data Handling
Use models that can operate in a standalone manner for sensitive data.
Avoid sending sensitive data over the network.
Ensure strong encryption for data in transit and at rest.
Fine-tune With Own Data (For open-source LLMs)
Anonymize and sanitize data.
Use differential privacy techniques.
Compliance & Audit
Ensure that GDPR, CCPA, or other relevant data privacy laws are followed.
Establish routine audits for data handling and model responses.
4. Implement LLMs For Your Business
Infrastructure Setup
Cloud (public/private/hybrid) or on-premises based on data sensitivity and compliance needs.
Scale to handle enterprise-level requests.
Integration
Integrate LLMs into existing enterprise tools (CRM, ERP, CMS, etc.)
Use APIs for seamless communication and real-time responses.
Feedback Loop
Incorporate mechanisms for employees to provide feedback on model suggestions.
Iteratively improve and fine-tune the models based on feedback.
5. Training and Change Management
Employee Training
Organize training sessions explaining the capabilities and limitations of LLMs.
Encourage an understanding of when to use LLMs versus human judgment.
Change Management
Keep stakeholders informed about the changes and benefits.
Foster a culture of acceptance and adaptation.
6. Evaluation & ROI Analysis
KPIs
Track performance indicators such as time saved, task efficiency, error rates, and employee satisfaction.
Continuous Improvement
Use analytics to identify areas where LLMs are underperforming or overused and adjust accordingly.
7. Scaling & Expansion
Expand to More LOBs
After successful deployment in prioritized areas, extend to other LOBs.
Keep Abreast of Advancements
Technology in the AI and LLM domains evolves rapidly. Ensure you leverage the latest advancements and subscribe to the Datagrom weekly email newsletter.
By following these steps, you can strategically implement LLMs across the enterprise, ensuring productivity gains while prioritizing data security.
Subscribe to our Weekly Data & AI Newsletter
Stay ahead of the curve with data science & machine learning Insights, resources, and tips
Subscribe to our weekly Data Science & Machine Learning Technology Newsletter
