Over the past years, integrating AI in the workplace has become increasingly common, especially among enterprises. We’ve seen chatbots used in customer service, AI-based product recommendations, smart filters and categorization assisting users with email correspondence, and numerous other examples of AI in the workplace.
Most recently, the surge of generative AI has encouraged businesses to automate more complex tasks, for example—managing enterprise knowledge or streamlining document processing. To understand how this becomes possible, we took the large language model (LLM) that runs under the hood of ChatGPT and tested its capabilities applied to enterprise data management.
Further in this blog post, we share our thoughts on automation and AI in the workplace, provide a comprehensive understanding of LLMs and their capabilities, and reveal the potential of ChatGPT for document management.
What is the potential of generative AI in the workplace?
According to Goldman Sachs, artificial intelligence is likely to replace about 300 million jobs globally. In the US solely, generative AI can automate up to two-thirds of all occupations.
In their most recent report, McKinsey analyzed the economic potential of this technology and revealed several impressive insights.
- Generative AI could augment the global economy by up to $4.4 trillion annually—that’s more than the United Kingdom’s entire GDP for 2021.
- Combining generative AI with other technologies, work automation could enable x3 productivity growth.
- About 75% of the value that generative AI use cases could deliver falls across four areas: customer operations, marketing and sales, software engineering, and R&D.
- Generative AI’s ability to understand natural language dramatically accelerates the potential for technical automation—work activities that account for 25% of total work time.
But disconnected from the promising future, what is AI currently capable of in a digital workplace for enterprise and is it enough to withstand market competition?
The current state of automation and AI in the workplace
As of 2022, 70% of organizations already implemented automation technologies in one or more business units or functions.
According to Zapier, the largest share of automation and AI in the workplace falls on tasks related to data entry, digital file management, sales and customer service, and inventory management.
Once automating manual data processing and administrative tasks, organizations report improving employee productivity, reducing operational costs, and accelerating innovation. But is this scale of automation enough to gain a competitive edge in the enterprise market today, when the speed of transformation has reached its peak?
Where is AI in the workplace still lagging behind?
Together with tech companies and market researchers, we analyzed key tasks and processes that make up part of enterprise daily routines and revealed the following statistics.
- 90% of global workers are still burdened with tedious, recurring tasks that take 21 days of their working time annually.
- 47% of digital workers struggle to find the information needed to effectively perform their jobs.
- Lack of quick and easy access to documents is reported as a key issue of modern digital workplaces by 58% of employees.
At Trinetix, we are extensively partnering with enterprises to tackle operational inefficiencies by introducing automation and AI in the workplace. Our experience shows that 9 in 10 large-scale businesses’ transformation plans risk smashing against poor internal data management that is usually caused by:
- Data storage inconsistencies
- Multiple data formats
- Poor integrity of the sharing processes
- Numerous sources of information and knowledge holders
- Legacy documentation
The lack of a value-driven approach to automation on the one side and multiple inconsistencies of knowledge and document management on the other eat up a lion’s share of enterprise productivity and make them find new ways to introduce automation. And often, these considerations come down to implementing sensational solutions like ChatGPT. Let’s have a closer look at those.
The potential of ChatGPT in the workplace
In November 2022, the world was stirred up by OpenAI releasing ChatGPT, an artificial intelligence chatbot capable of understanding natural human language and generating human-like replies. In just a few months, the tool has reached 100 million monthly active users, and in less than a year was reported to be used by companies like Coca-Cola, Snap.inc, Slack, and Salesforce.
Becoming the most successful AI project in the world so far, ChatGPT has proudly broken into digital workplaces and become a reliable ally to customer support teams, marketers, content creators, software engineers, sales managers, and recruiters.
A recent marketing report on the practical use of ChatGPT and AI in the workplace has revealed that the tool brings an impressive 74% boost to enterprise productivity. Still, the question remains—how to make ChatGPT practical if we go beyond using the web app for casual data search and email text generation?
To understand the potential of ChatGPT for enterprise document management, and evaluate its capabilities in relation to specific organizations and business challenges that exist within them, let’s zoom in on the technology that works under the hood of the tool.
Things you didn't know about generative AI!
How does ChatGPT work?
Tools like ChatGPT represent large language models (LLMs), computer algorithms that process inputs in natural language and generate outputs based on their knowledge. Trained on huge amounts of data available on the internet, the models are capable of providing diverse information once they are given a prompt or a question.
In detail, the process of user interaction with ChatGPT looks as follows:
- The user provides ChatGPT with a prompt (input).
- A deep learning algorithm that runs under the hood analyzes the input to detect keywords and phrases that provide a context for the question.
- ChatGPT uses natural language to generate a response that would be grammatically correct and relevant to the context.
- ChatGPT generates the response (output) to be displayed to the user.
How can enterprises benefit from ChatGPT for data management?
If an LLM is capable of learning from data available on the internet, it can potentially learn from any data, including unique business- and industry-specific knowledge and documentation that creates major challenges in enterprise workplaces.
- Using deep learning technology, data extraction tools can also pull out text from different file formats, including PDF and JPEG. This allows enterprises to analyze the diversity of enterprise documents, from digital tax reports to paper invoices.
- Moving further, LLM-based tools allow users to interact with multiple enterprise documents at a time by transforming bits of data into embeddings, specific word representations used to categorize text bits based on their semantic meaning.
Integrating ChatGPT in the workplace is associated with vector stores, dedicated databases for storing documents and their embeddings. This means that any data uploaded is broken into chunks and stored in a vector format, and the same applies to user queries. Using semantic search, the system calculates the distance between vectors to determine the ones that are located close to the query—this information is considered the most relevant in the given context.
Using vector-based document content search as a basis, enterprises can proceed with developing “chat with your document” capabilities and create a ChatGPT-like virtual assistant trained on the company information.
How does that work in practice?
Let’s say an Account Executive needs to find information about a specific client the company has been working with for the past 30 years. Apparently, part of that documentation is still located in offline archives, while the other part is stored locally in the form of Word documents, Excel spreadsheets, and PDF files. In addition, pieces of information about the client may be scattered across more generic sources, like annual corporate reports or historic databases.
Without an AI-based solution, analyzing such data may take months or even years. Here is what it looks like when the information is processed and analyzed with an LLM-based tool.
Step 1. Uploading files that would make an internal database.
Step 2. Providing input. For example, the client’s name.
Step 3. Using prompt engineering to “сhat with the documents” allows the worker to fully manage the way LLM analyzes data and get all the necessary information in a blink of an eye.
Step 4. Discovering a few more ways to work with documents. Once given a prompt, LLMs can perform more useful actions. For example, total amount calculation and data summarization.
These steps are just a few examples of how LLMs like ChatGPT apply to document management within a specific client-related enterprise role.
Acknowledging the new scale of automation
Solutions like this can help to streamline the work of employees from different departments in many industries.
- Accountants can dramatically decrease the time they spend processing financial documentation, invoices, and payrolls.
- HR department and talent acquisition teams can benefit from automating employees’ and candidates’ data screening.
- Data analysts can get insights into a business's customers faster and automate rebuilding historical data.
- Logistics and transportation workers can streamline supply chain tracking and identify the reasons for supply chain delays.
- Healthcare professionals can streamline the analysis of patient data by summarizing medical histories.
- In construction, civil engineers can get precise information about cost estimates and view historical data about the structures and infrastructures they manage.
In fact, by introducing AI-enabled document search and “chat with document” capabilities, businesses get a document management studio with an LLM running under the hood. Using the power of prompt engineering companies can further adapt it to solve a number of tasks. Some open source LLMs can also be trained on custom data, which makes the expected scale of automation even more massive.
What’s next for ChatGPT in the workplace?
We’ve already mentioned some roles that are going to benefit from implementing AI, and LLMs in particular. However, the real potential of generative AI is yet to be discovered. Therefore, limiting the capabilities of tools like ChatGPT to specific industries and user roles would be a mistake.
What’s clear so far is that humans will still play a key role in managing emerging technology in the workplace, and the results companies achieve with AI will depend on how fast they tame the innovation.
Apart from that, organizations should already start thinking about the other side of automation—managing the risks, determining the skills and capabilities to acquire, and rethinking core business processes in line with each new level of AI enablement.
Get maximum value from generative AI with a tailored solution
Implementing AІ in the workplace requires a deep understanding of business-specific processes and workflows. At Trinetix, we leverage a strategic, discovery-first approach that helps enterprises not only embrace technology to elevate their business outcomes but become future-ready. If you are interested to learn how cutting-edge technology can help you achieve measurable productivity results, let’s chat and make it real!
