Mega Myth - To use AI you need a lot of Data
Midjourney / Antony Slumbers
It is generally assumed that to leverage AI you need to have a lot of data at your disposal. No data = no play! This is such a prevalent belief it’s almost assumed to be a Law of Nature. I’m sure you have heard and read it a million times.
But it is not true.
At least it is not when you are dealing with Generative AI.
With Predictive AI it most certainly is, as that is all about predicting, clustering and classifying. You are applying these actions to specific datasets. So obviously without data you are facing a brick wall.
But with Generative AI the data comes built in. Take ChatGPT - the GPT stands for Generative Pre-trained Transformer, where the training has already taken place on the near entirety of data that exists on the open Internet. And from that vast corpus of data has been developed a statistical model that allows for the creation (the generative bit) of new text, code, images, video, speech or actions.
So when you are using generative AI you have at your disposal if not ‘all the worlds information’ then pretty close to it. You have vast amounts of data at your beck and call. You can, if you have it available, augment this with proprietary data, but to a large degree that is not necessary, or at least does not bring as much as you think to the party. Your data is pretty small compared to what the ‘Large Language Model’ already has intrinsically.
The bottom line is that there is a huge amount you can do with Generative AI without the need for any other data. This is what is not generally appreciated and is why this technology is also known by a different interpretation of the acronym GPT - a General Purpose Technology. Which signifies, like electricity, the internal combustion engine, the Internet itself, a technology that is not a point solution but one that is or will become pervasive throughout society. Generative AI will seep in, often invisibly, to everything. There is little you will be doing within a few years that is not, in one way or another, mediated through AI.
Indeed, there is no need to wait. Below are examples of use cases, by business department, you can implement today. No data required.
McKinsey reckon 75% of potential productivity value and gains will come from the first four categories, but the others are included to show just how much is possible ‘out of the box’. Much of this can be achieved by an individual using public tools like ChatGPT, Claude, Google Gemini and Midjourney. Whilst other areas might require customised products or coding. But either way, almost all of this is available to anyone in your company. And again, with no data required.
Sales & Marketing
Content Creation: Generate engaging marketing copy, blog posts, and social media content.
Email Campaigns: Craft personalised email messages for different customer segments.
Market Analysis: Summarise market trends and news from publicly available sources.
Customer Segmentation: Predict customer preferences using open-source demographic data.
Product Recommendations: Suggest products based on general market trends.
Interactive Content: Create dynamic web content to enhance user engagement.
Predictive Analytics: Analyse customer behaviour for better targeting and segmentation.
Product and R&D
Idea Generation: Brainstorm product ideas based on market analysis.
Prototype Testing: Simulate user feedback on prototypes with AI-generated personas.
Research Summarisation: Compile relevant research to support R&D.
Competitive Analysis: Analyse competitors' product strategies.
Design Optimisation: Propose product design improvements using generative models.
Material Research: Summarise findings on new materials from public databases.
Customer Operations
Chatbots and Virtual Assistants: Implement AI-driven chatbots for customer support.
Feedback Analysis: Analyse customer feedback from public reviews.
FAQ Generation: Automatically generate FAQ content.
Operational Efficiency: Optimise workflows to manage high-volume periods.
Personalisation: Personalise interactions based on behaviour trends.
Software Engineering (Product Development and Corporate IT)
Code Generation: Generate boilerplate code and documentation.
Bug Fixing: Identify potential bugs using publicly available datasets.
Automated Testing: Adjust tests automatically to application changes.
Architecture Design: Suggest improvements based on public best practices.
Security Vulnerability Identification: Identify vulnerabilities from public databases.
Strategy
Trend Analysis: Identify emerging industry trends.
Scenario Planning: Generate business scenarios for planning.
Benchmarking: Benchmark against industry standards.
Innovation Tracking: Track industry innovation trends.
Strategic Diversification: Analyse potential diversification areas.
Legal
Contract Generation: Generate standard legal documents.
Legal Research: Summarise legal precedents from public databases.
Compliance Monitoring: Track changes in laws and regulations.
Dispute Resolution: Suggest resolutions based on similar public cases.
Policy Development: Analyse public compliance standards for internal policies.
Risk and Compliance
Regulatory Compliance Tracking: Monitor regulatory changes.
Risk Assessment: Conduct assessments based on public threat data.
Fraud Detection: Detect fraudulent activity patterns.
Ethical Compliance Monitoring: Monitor public sentiment for ethical issues.
Cyber Risk Analysis: Analyse public data on cyber threats.
Talent and HR
Resume Screening: Automate initial resume screening.
Employee Engagement: Analyse engagement trends to inform strategies.
Training Programs: Develop AI-driven training programs.
Workforce Planning: Use labour market trends for planning.
Diversity and Inclusion: Inform policies based on diversity data analysis.
So, as you can see, it’s time to bury the ‘you need data’ myth. You absolutely do to get the most out of a lot of AI, but with Generative AI the biggest constraint is not data, but your own curiosity, vision and willingness to just get stuck in.