Use Generative AI to talk with your data and have it talk back

Share on Social Media

If data is the new oil, then the key to unlocking its value lies in refining it. Without effective management and analysis, the insights in your data will go unrealized. Data management and analysis looks a lot like the upstream energy supply chain. Just as energy companies collect, transport, and process crude oil to make a refined, useful product, businesses do the same with data to produce actionable information. That involves collecting, organizing, storing, processing, and interpreting data to support decision making and problem solving.

This management and analysis is critical for enterprise workflow solutions (EWS) providers, because most business processes rely on data. Mining that data for insights makes the processes more accurate and efficient.

Managing data has traditionally been a complex and time-consuming task. Working data into something useful required specific knowledge in a programming language like Python or R. Employees who needed answers relied on overworked data scientists to process data into knowledge.

HOW GENERATIVE AI DEMOCRATIZES DATA

Generative AI is changing the game for employees around the world by democratizing data. Now, people can ask questions using natural language, rather than the typical commands used in data science programming. Generative AI can ingest huge data volumes and automate much of the underlying work, creating a data pipeline that produces actionable intelligence at scale.

An example: ChatGPT's Data Analysis Plugin

ChatGPT's new data analysis plugin allows anyone to extract valuable insights from data without needing to code. After you upload a data file, the service quickly analyzes the data, working out its structure. It will quickly identify data headers and use its language capabilities to understand what they mean, understanding Country, Gender, and Population, for example.

This enables you to ask natural-language questions such as 'List the five countries with the highest mean average salary for women' and it will write and run Python code behind the scenes to give you a written response.

You can even use ChatGPT's existing knowledge to ask it questions about information not in your data. For example, the generative AI already has an idea of which countries are considered developing vs developed. This enables you to ask 'are gender pay disparities higher in developing countries than in developed ones?' even if your own data set doesn't delineate countries in this way.

How generative AI can enhance data management and analysis

Generative AI supports data management as a service in various ways, allowing businesses to focus on making decisions:

Data cleaning

Like crude oil, data is rarely useful immediately. It needs careful treatment first, especially when it comes from different documents and sources. Generative AI tools are excellent at correcting mistakes in data, identifying missing fields, and automatically standardizing data formats for accurate analysis. This prevents messy or inconsistent data producing misleading results.

Analysis and interpretation

Generative AI tools are good at identifying trends and correlations in large data sets that might not be easily noticeable to humans. Microsoft's Excel Copilot can even color data elements in a spreadsheet to highlight these trends based on natural commands.

These natural language interfaces allow non-programmers to converse with their data in ways that were not possible before. You can ask questions and drill down on the AI's responses with more detailed queries. This interactive process allows for a deeper understanding of the data, enabling more informed decisions.

Data visualization

Not only can you ask generative AI tools to plot your data in different types of graph, but you can easily make changes and amend the graphs as you work. Want that bar chart to become a line graph? Just ask. Now turn the line green and make it solid instead of using dots. Pull in data from another source and map that on the graph too – all without a line of code.

Data sharing

Telling stories with data is better if you can share them with others. Generative AI tools allow you to download your data conversations as files or share them as links. Take things a step further by using generative AI office productivity tools to automatically produce presentations using those assets.

Anomaly detection

AI can train on large data sets to understand normal patterns in data. It can then quickly check new data to spot any outliers that could indicate business issues, such as fraud.

ANALYZING DIFFERENT DATA TYPES

Generative AI's powers of analysis don't stop at numeric data. It is perfect for ingesting and processing unstructured data including text. For example, a worker dealing with a complex technical support case could use generative AI tools like ChatGPT to mine a knowledge base, analyzing other responses to similar issues. This could cut resolution times and provide a better customer experience.

HOW GEN AI ENABLES UNSTRUCTURED DATA ANALYSIS AT SCALE

Generative AI tools are useful in data analysis tasks even when the data is highly technical and focused on specific domains. For example, SPS is experimenting with generative AI as a tool to extract knowledge from multiple-page documents.

Staff traditionally had to read, understand and index these documents manually, using complex business rules to work out which information they should then retrieve and preserve.

Using Generative AI models, we can analyze and summarize those documents, collecting the salient points from each one, interpret the context behind the data, and present them in an easily-readable form. Knowledge workers with access to these verified documents can then make fast, informed decisions.

Like all generative AI use cases, this task should not be completely automatic. Human operators still check the summaries against the documents to ensure their accuracy, but this process is still far faster and less error-prone than it would be otherwise.

RELATED ARTICLES

In 2017, Google researchers pushed the boundaries of the possible. They published an academic paper called...

The personal computer (PC) was invented in the 1980’s. However, it was not until the late 1990’s, when unit costs had dropped and PCs started being connected to the internet, that this breakthrough started having a noticeable impact on productivity figures.

As explored in the first post in this series, generative artificial intelligence (AI) is a ground-breaking technology that uses large language models (LLM) to create...