The utility and applications of large language models (LLMs) such as ChatGPT are growing with every passing day. It offers opportunities as well as challenges in relation to using technology in an appropriate manner. One of the key mechanisms to maximize the benefits of using LLMs is to develop an understanding of prompting. In the LLM context, prompting is a strategy to provide instructions to the machine that maximizes the likelihood of eliciting a correct response.

Prompting occurs in many forms. Role prompting is one such type, where a human or machine interacts with a machine in a certain role (e.g., project manager) to raise prompts to get the desired responses. The idea is not very different from using the right keywords to find relevant information, such as when using a search engine, with one key difference though. LLMs differ in the sense that they have enhanced capabilities in natural language processing, so they offer opportunities to ask complex questions in an interactive, natural, and conversational setting to find answers. Building rapport with machines through effective interactions in the form of prompting is vital to benefiting from LLM.

Role prompting can occur in all types of work situations and businesses. Project management is no exception. For example, one of the key tasks in project management is the creation of project documents, and role prompting can be used for it. It is no secret that the creation of project documents often takes a lot of time and effort. People find it cumbersome and boring, resulting in either error-prone documentation or a lack of documentation altogether. Using role prompting, one can work with LLM to build information that can go into the creation of project documents. Particularly, people with little experience but some knowledge of project management can use LLMs to get started on building information for creating project documentation.

With that in mind, we decided to put theory to the test and find out how ChatGPT could help in the creation of information that can be used for writing a project management plan (PMP). The idea was to raise role driven prompts that could induce ChatGPT to give better and more relevant information that can be instantly used (with little refinement) for the creation of a PMP.

Below are the actual prompts raised to help create a PMP. As answers by ChatGPT were often detailed, so keep it short, I have summarized some answers in brackets). Any text that is not in brackets is actual interaction.

Role: Project staff (PS) interacting with ChatGPT to build information for creating a PMP for a project

 PS: I am new to project management, and I have been asked to create a project management plan for a project that is to install 100 desktop computers at three office locations. Can you help me create a project management plan. I want to emphasize I need to create a realistic and quality project management plan so keep that in mind.

 GPT: Of course, I can help you create a realistic and quality project management plan for installing 100 desktop computers at three office locations. A project management plan is crucial for organizing and executing a project effectively. Here’s a step-by-step guide to creating your plan:

(Here GPT gave a 15 point list of various aspects including scope, risk mitigation, budget, and quality assurance.)

 Remember to be flexible and adapt the plan as necessary throughout the project’s lifecycle. Continuously communicate with the team and stakeholders to ensure everyone is aligned and committed to the project’s success. Good luck with your project management plan!

PS: This is very vague and not helpful. Ok, lets go step by step. First I want you to create a project objective statement.

GPT: Apologies for the vagueness. Let’s start by creating a clear and concise project objective statement:

Project Objective Statement:

The objective of this project is to successfully install 100 desktop computers at three office locations in Sydney, ensuring they are fully operational and ready for use by the respective office staff. The installations must be completed within the agreed-upon timeframe and budget while adhering to the highest quality standards. Effective communication, risk management, and stakeholder coordination will be emphasized to achieve a seamless and efficient implementation process.

PS: thats a good start. Ok, I need you to write specific requirements for the project. I need you to ensure that every requirement is verifiable and measurable. I also need you to write how requirements will be verified.

GPT: Specific Requirements for the Project: Installing 100 Desktop Computers at Three Office Locations in Sydney

Hardware Specifications:

Requirement: The desktop computers shall be equipped with a minimum of Intel Core i5 processors, 8GB RAM, and 256GB SSD storage.

Verification: Hardware specifications will be verified by physically inspecting each computer upon delivery and checking the specifications against the purchase order.

Software Setup:

Requirement: Each computer shall be installed with the latest Windows 10 Professional operating system and necessary office productivity software, including Microsoft Office Suite.

Verification: The software setup will be verified through manual inspection of each computer and confirming the presence of the required software and licenses.

 (GPT also provided various other requirements including Network connectivity, Functional testing, UAT, Health and safety compliance, and User testing)

PS: Ok now lets create scope. Can you pls tell me what to include in scope. Pls be specific.

 GPT: The project scope defines the boundaries and extent of the project. It outlines the work that needs to be accomplished to achieve the project objectives. The scope includes both what will be delivered (inclusions) and what will not be delivered (exclusions). Here are the specific elements to include in the project scope:

Inclusions:

Procurement of 100 desktop computers with specified hardware specifications (Intel Core i5, 8GB RAM, and 256GB SSD storage) and peripherals (monitor, keyboard, mouse).

Exclusions:

Any changes or upgrades to the existing office network infrastructure beyond basic connectivity setup for the new computers.

Custom software development or modifications to existing software beyond the standard software setup.

(GPT lists many other Inclusions and Exclusions when describing scope)

 PS: ok now pls help create a WBS based on the scope and requirements

(GPT creates an indented WBS which was quite detailed. On a further query, GPT also created an activity list)

 PS: ok lets estimate time for each and every activity first

(GPT created an estimated time for each activity. Though time estimation by GPT has no real basis and is a mere assumption)

PS: ok, your initial estimated total duration for the project was 12 weeks, but when estimating time for the activities, you did not keep that in view. Can you pls ensure that the project plan does not have these types of errors.

 GPT: You are absolutely right, and I apologize for the oversight. It’s essential to ensure that the estimated time for the individual activities aligns with the overall project duration. Let’s revisit the estimates and make sure the project plan is accurate:

(GPT re-estimates the time duration for each activity. GPT was further asked to estimate the cost, and it provided a ballpark figure as an estimated cost for each activity)

 PS: ok. now can you create a project gantt chart given the list of activities, their dependencies and time estimates.

 GPT: Sure! Below is a simplified Gantt chart based on the critical path activities, their dependencies, and time estimates. Please note that the actual Gantt chart used in project management software would have additional details, such as task assignments, progress tracking, and milestones.

(GPT creates a simplified version of Gantt chart in tabular form)

PS: since I am new to project management, how do I know that the project management plan you have generated is reliable?

GPT: As a newcomer to project management, it’s natural to have concerns about the reliability and effectiveness of a project management plan. Here are some steps you can take to evaluate the reliability of the project management plan:

(GPT lists various steps such as cross-check with experts, review assumptions and constraints)

 PS: Ok. Now this project is going to be delivered in Australia. Can you pls provide me the list of approvals that I need to get from various agencies such as government departments to ensure project is delivered in a sound manner.

 GPT: Delivering a project in Australia may require obtaining various approvals and permits from relevant government departments and agencies to ensure compliance with local laws and regulations. The specific approvals needed can vary depending on the location, nature of the project, and other factors. Here’s a general list of approvals that may be required for an IT project involving the installation of 100 desktop computers in an office:

(GPT lists various types of approvals that may be needed. It also provide names of specific agencies for seeking approvals)

 Concluding thoughts:

Working with GPT to build information for a hypothetical project was quite an interesting experience. Here are some of the learnings, positives, and areas for improvement in using LLM for the creation of information for project documentation purposes.

 

1. The information that one can build using GPT is worth the time and effort. While one can use previous project files to extract information and refine it to current project needs, the experience showed that using GPT is potentially more effective as it is like a one-stop information portal for every type of information that one needs for creating project documentation. One does not need to look for multiple archived files to build project information for creating relevant documents.
2. The value produced by GPT in the form of its responses to prompts goes well beyond traditional means of producing information. For instance, for the hypothetical project explained above, GPT was able to provide information about things like the type of regulatory approvals one needs to use the product produced by the project.
3. If the prompts do not produce the desired answer, changing the prompts is always an option to get more relevant and accurate information. It saves time and is easy.
4. One needs to have some domain knowledge and familiarity with keywords used in a particular domain to induce GPT to provide better information.
5. Information that requires logical calculations and mathematical operations to produce is not handled well by GPT. So, estimation of time, cost, and other similar tasks need to be done outside of GPT. However, over time, it is expected that LLMs will be integrated with project management tools, which can then work in sync to handle complex tasks such as performing calculations behind the scenes and producing results through the GPT interface.
6. It is important to understand that GPT is producing the next most likely word based on probabilities when forming its response to a prompt. Therefore, using the right keywords is vital for GPT to pull together a better and more relevant response to a prompt.
7. Learning prompting can really help everyone, be it experienced or not-so-experienced project staff, get the most out of GPT for building project documentation.
8. It is also important to check the information produced by GPT before using it for project purposes, as the information is not always relevant and accurate.
9. The fast-paced developments in relation to the integration of LLMs in various AI-based applications are indicative of LLMs potentially becoming a mainstream technology. Hence, preparing for such an eventuality should be part of the futuristic vision of every project- and non-project-based organization.

 

© 2023 Jiwat Ram, All Rights Reserved.