ChatGPT: Simple chatbot or revolution for operations?
24 February, 2023
In November 2022, OpenAI opened the beta version of ChatGPT to all registered users. As a result, the chatbot based on the GPT 3.5 language model became a talking point worldwide. At the same time, the tool's wide range of possible uses increasingly poses the question: how can companies use AI's potential to create added value?
While some use cases are obvious, others are not so immediately apparent. Now that ChatGPT has overcome the initial barrier to experimenting with generative AI like this, this is a question that's definitely worth tackling. This is why we are showing here the opportunities that technology offers for the operative corporate division.
Most important in 30 seconds
- ChatGPT enables organisations to quickly develop scalable AI use cases for operations.
- Promising use cases can be found in areas such as procurement, research and development (R&D) and smart manufacturing.
- Developing promising use cases requires the right strategy for brainstorming, prototyping, testing and implementation.
One technology, many facets
OpenAI is a US artificial intelligence research laboratory founded at the end of 2015 by entrepreneurs such as Elon Musk, Peter Thiel and Sam Altman. Its aim was to research and develop AI to benefit all of humanity. In 2019, OpenAI changed its charter from “Non-Profit” to “For-Profit”/“Capped-Profit” – meaning that each investment in the organisation has a profit cap of 100x.
ChatGPT is just the latest in a series of powerful machine learning products from OpenAI. The organisation has significantly improved the chatbot by training it on a larger database (which is largely based on many billions of web pages and therefore also includes digitised books, articles and Wikipedia entries) as well as advanced internal parameters. The system is based on the GPT models and has been specially trained as a chatbot. It should therefore be able to process several rounds of user feedback and reciprocal interactions.
OpenAI’s portfolio includes the following solutions:
- OpenAI Gym, a platform for the development and comparison of reinforcement learning algorithm launched in 2016/2017, which provides a standard API for communication between learning algorithms and environments.
- DALL-E 2, a text-image synthesiser launched in the second half of 2022. The system creates images from a text-based input. The name of the software is a portmanteau of the names of the animated Pixar robot figure WALL-E and the Spanish artist Salvador Dalí. DALL-E 2 is already being monetised.
- Whisper is a state-of-the-art machine learning model for voice recognition (currently mainly in English). It enables transcription in multiple languages and translation from those languages into English. According to its own information, it is expected to outperform other models by up to 50% in certain tasks.
Bringing ChatGPT to life for operational challenges
Before companies simplify their day-to-day business with ChatGPT and inspire employees, partners and customers with a new user experience, there are a few things to consider.
First things first: Companies must define unique use cases in order to evaluate the required data records based on existing ERP systems such as SAP S/4HANA or Oracle. This ensures that ChatGPT is not only trained on company and function-specific formulations and languages, but also generates proper dialogues. For example, if operational support is planned for the procurement function, ChatGPT must have access to the correct department abbreviations and underlying order data.
Companies should take full advantage of the opportunities offered. This is the first time they have had the opportunity to easily establish generative AI use cases with ChatGP – so the focus must be on the opportunities. This is the only way to make the step towards the workplace of the future. Starting with a proper pilot project builds trust and helps to develop further company-specific use cases.
In principle, companies should never rely entirely on artificial intelligence. Traditional IT projects require proper testing of the results provided to avoid misleading interactions between the system and the user. However, for AI applications, testing must not be limited to initial setup of ChatGPT or any other AI application. As the AI engine evolves continuously and open source, continuous monitoring of application behaviour is critical to long-term user adoption, output quality and success.
Wide range of use cases for operations
With a view to the possibilities for company-wide use, we see a wide range of use cases, especially for operations, in order to support user interaction not only transactionally but also strategically. As food for thought, we have already identified the first use cases.
Use cases for purchasing
- P2P – internal helpdesk: With ChatGPT as a helpdesk, buyers and customers can quickly and easily access the information they need – in their native language and without any clear language norms. Users simply ask how a particular purchase requirement can be met and immediately receive information on the right catalogues or a free text formula. Open purchase requisitions or orders can be tracked effortlessly and without wasting time. This benefits both process compliance and the purchasing experience.
- Interaction with suppliers: Regardless of how well prepared the tender documents are – this includes queries from suppliers. Therefore, companies typically grant a timeframe for submitting the “Request for Proposal”-related (RfP) questions, in which answers are expected from the RfP initiator for clarification. This is where ChatGPT comes in. Questions no longer need to be collected and consolidated answers no longer need to be provided. The tender participants act directly with the chatbot. Queries can be clarified directly in this way, so that the bidder can start preparing the offer immediately. The strategic buyer bidding for the tender no longer has to orchestrate questions and answers in a time-consuming manner.
R&D use cases
- Automatic creation and documentation of software: ChatGPT is an excellent tool for streamlining and accelerating software development processes. It can generate and document code in programming languages such as Python, Java and C++. This significantly reduces development time and error rates, allowing organisations to quickly implement simple use cases and create valuable add-ons. ChatGPT speeds up the process and also provides easy-to-understand documentation in natural language – eliminating the need for time-consuming manual documentation. Whether companies are working on a small project or a large-scale software development initiative, ChatGPT can help generate efficient and accurate code in no time with its natural language processing or template-based code generation and automatic code refactoring.
- Automating design drafts and documentation: When design departments couple GPT with a well-trained AI model and computer-aided design (CAD) software, the system can be trained to draft new design proposals, analyse existing ones, suggest changes and optimisations, and propose documentation in natural language. This enables developers to design faster and smarter. The end result: fewer errors, less time spent and lower costs, as well as more power and engineers who can focus on value-adding tasks again. GPT can help to predict potential error modes and create descriptions of performance and simulation results in natural language. This makes it easier for development teams to understand and improve designs.
Smart manufacturing use cases
- Optimise quality control: ChatGPT enables manufacturers to improve product reliability and customer satisfaction by automatically detecting patterns and generating insights from text-based data such as inspection reports and processed images. The goal? Detect product anomalies. To do this, ChatGPT needs to be combined with Computer Vision AI, which detects patterns in images and classifies them based on text. This allows companies to optimise their quality control processes and stay ahead of the competition.
- Easy and straightforward (predictive) maintenance: When combined with other machine learning models that detect anomalies in sensor time series data, ChatGPT can provide clear and easy-to-understand instructions for maintenance technicians to work faster and more efficiently. This enables maintenance teams to proactively address issues and avoid costly downtime.
Drive pilots and scaling with design thinking
These possibilities show that there are already countless enterprise-specific AI use cases across all major operations areas. Technology is currently on the verge of broad implementation – however, disruptive breakthroughs call for bold steps. The key challenge for companies now is to define suitable scenarios for piloting and rapid scaling.
We offer our customers a design thinking approach to deploying ChatGPT/OpenAI solutions.
Through our Design Thinking workshops, we follow an iterative process that aims to look at challenges from the user’s perspective and identify solutions based on ChatGPT/OpenAI capabilities. The overarching goal: initial pilot applications and prototypes.
The key to the design thinking approach's success is the stakeholders involved. For the best possible result, we recommend a domain-specific set-up (e.g. procurement, supply chain, production, R&D) and clear guiding principles (e.g. increase in user satisfaction, reduction in throughput times).
Our five-step approach
We accompany and guide our customers through a proven five-step approach from understanding the current challenges to final implementation:
- Understanding
- Generate ideas
- Prototyping
- Testing
- Implementation
Understanding
First, we collate the requirements of users and departments in order to better understand individual needs. At the same time, we create a basic understanding of the ChatGPT/OpenAI solution within the organisation.
Generate ideas
Once we have understood user needs, we work with our customers to develop initial ideas for using ChatGPT. In this phase, the imagination has no limits and we also use various techniques for brainstorming, such as mind mapping, sketching, role playing and user stories.
Prototyping
From the ideas generated, we create prototypes that are orientated towards the users' needs. Quick and easy assembly is important to enable rapid testing and reworking of the prototype. We usually bring in experienced technology experts in order to keep up the pace.
Testing
We test the prototype in terms of user experience and suitability for target group. The feedback is collected and used in the further development of the desired ChatGPT/OpenAI solutions.
Implementation
After successful prototyping, we implement the most successful ideas and roll them out as part of customer-specific release management.
The greatest advantage of design thinking is not only the quick identification of improvement opportunities, but also the rapid validation and prototyping. This is critical for a technology as rapidly scaling as ChatGPT to stay ahead of the market.
Within 14 days, our experts are able to verify the product range, the needs it addresses, and industry knowledge, and translate all this into a helpful design thinking approach.
“ChatGPT harbours great potential for a wide range of use cases for business operations. However, in order to use this, we need a good sense of the needs of the specialist departments and a structured approach in the development of prototypes.”
Related Content
Follow us
Contact us
Director, Data & Analytics, Operations Transformation, PwC Germany
Tel: +49 151 15535019
