ChatGPT has the Potential to Accelerate HR Transition to Web3

ChatGPT can write business application code. Low Code platforms that were designed for users, or citizen developers, to build small custom applications can now be used for a broader purpose. That is, to participate in the redesign and reconstruction of HR technology as it moves from Web2 to Web3. However, it is not that simple. It must be done in structured and methodical way. This article explains how and provides examples.

Introduction

Over the next couple of years HR system suppliers that wish to remain in the industry will need to transition to Web3. That means they will need to embrace DLT (Distributed Ledger Technology) and componentise their products to operate in a decentralised peer-to-peer architecture. Companies will not go shopping for monolithic systems, as they have in the past, but will purchase and deploy only the type of functionality that is needed and where it is needed in the network.

The old issue of whether to build or buy is now revisited by industry in light of new Low Code software products and AI (Artificial Intelligence) to simplify development. In this article I will look at the transition process from the Consortium for Decentralized HR perspective and Competitive Edge Technology’s (CET) prototype approach and how ChatGPT will play a very important role. In particular, I will look at:

a)??????Why now is the right time to rearchitect HR technology

b)?????Why it is advisable to transition to Web3 via a prototype built on a Low Code platform to ensure the business community involvement from the specification to deployment stages. The chosen Low Code platform is Salesforce.com because of the ease of use and ability to use native Salesforce custom development features without Apex code knowledge.

c)??????Can OpenAI’s ChatGPT (Generative Pre-trained Transformer) product actually write code? The answer to that is YES and I will show examples of how Salesforce code can be copied and pasted and stored for reusability.

d)?????Quantifying in hours and weeks how savings from the use of ChatGPT and a Low Code platform can be used in development to accelerate the build to a DLT environment. The DLT software chosen is Holochain because its’ framework structure is similar to Salesforce’s object relational design.

e)?????How reusable code is stored in a Metadata Repository at field level to copy and paste later.

f)???????Provide an example of how ChatGPT code can be reverse engineered for audit purposes to explain in simple language what an AI code is executing and whether there is any instance of bias.

The diagram below illustrates the transition process to Web3 and where the productivity savings promised by ChatGPT will occur.

High level description of the transition to Web3 process

To map legacy functionality and data from Web2 to Web3 without a staging area and prototype is full of danger. Competitive Edge Technology provides clients with templates and advice during the process.

a)??Opportunity for HR Tech to Rearchitect

From an HR perspective it is a great opportunity to re-architect and fix some of the wrongs of the past that resulted in monolithic, centralised, and proprietary systems that made millions of dollars for companies specialising in integration of systems with different data structures and applying fragile API (Application Programming Interface) interfaces to make multiple systems “talk” to each other.

From a functional viewpoint many companies were paying for functionality that they didn’t need and didn’t want. It is a perfect time now to revisit what is required for present and future needs as massive changes will occur over the next few years. The best way to rapidly define requirements and prepare for the future is for the HR business community to prototype using a Low Code platform. In that way HR can decide what type of “HR system” they wish to compose.

Outline of the Web3 Architecture

A new architecture is required that is both open source and standardised. The architecture should allow interoperability and component assembly of solutions. In a Web3 peer-to-peer environment only segments (or HR microservices components) of an HR system need to be deployed to network nodes (or servers) based on the business practice. Web3 does not enable huge SaaS (Software-as-a-Service) systems, currently residing in massive datacentres, to migrate in their entirety to a single network node in a peer-to-peer architecture. ?

Web3 uses a different operating system and protocol to Web2 and most current applications will need to be rewritten in order to run successfully on the new platform.

The reality is, if all the HR software running on a Web2 platform needed to be rewritten to operate on a Web3 platform it would take many years. A similar situation arose with the introduction of Web2 cloud computing and there was a cry out to citizen developers and Low Code platforms at that time to come to the rescue: But that didn’t happen. The tools developed by most Low Code software companies required a good knowledge of programming techniques and it still required a high level of scrutiny by IT before they would be “production” ready. Citizen developers now have new tools and all that may be about to change for Web3 transition with the arrival of ChatGPT and its’ ability to write computer code for some Low Code platforms. It just may be the Holy Grail for citizen developers and the HR software industry in their endeavours to narrow the Web3 transition time gap.

b)?ChatGPT can Write Application Code

ChatGPT from OpenAI is a most exciting development not only for HR but for all enterprise applications. ChatGPT has evolved from AI and search engines. Unfortunately, at present OpenAI’s Web2 data sources are not reliable for some features that require fact checking and up-to-date reporting. However, there is no doubt that ChatGPT is a technology whose time has come and future versions of ChatGPT will need to run on a Web3 platform as data security is already an issue and AI transparency has been brought into question as possible bias may be built into programming logic. Web3 and ChatGPT are mutually inclusive from a data privacy and availability perspective.

ChatGPT could fill the citizen developer programming knowledge void by generating application logic code in response to simple language commands. Code can be generated for Salesforce.com custom applications and for Holochain, the DLT and possible Web3 framework product of choice for HR in the future.

It is still too early yet to ask ChatGPT to write complete software applications. However, at a micro-level it is capable of writing reusable code for custom applications that can be assembled manually from microservices and plug in components.

The Salesforce and Holochain products are of particular interest to readers of this article because they are the two major players in HR’s transition to Web3 identified by Competitive Edge Technology. Add to that Atlassian’s Confluence software to build metadata repositories and there is a team that could lead the charge and see HR Web3 component applications be the popular choice for HR software purchase by the end of 2023.

ChatGPT Salesforce Code Example

Here is how it works: Imagine I am building a custom application on the Salesforce platform. I want to add a field called AGE to save me subtracting today’s date from a person’s date of birth each time. There is a formula field in Salesforce designed to perform that kind of calculation. However, if I wrote the code as a citizen developer and made one single syntax error it would not work. If I made a request to ChatGPT to “give me the code to calculate a person’s age” it would generate the code shown in the diagram below for me to copy and paste to my Salesforce custom development.

The code generated by ChatGPT is confirmed as correct by the existing manually compiled code (shown in the example below) and tested in a CET production environment.

At the very micro level there is not much commercial value in the IP (Intellectual Property) associated with the age code example and NFTs (Non Fungible Tokens) are not coming into play at this stage. It is more beneficial to collaborate with other programmers internally and externally to reuse that code and to share other code.

Note, there is place shown in the diagram above where a link to a Metadata Repository or Code Library / Store is recorded for easy connection during development.

ChatGPT Holochain Code Example

Here is a screen shot of a ChatGPT request to produce the same code for Holochain running on a Rust platform:

For audit purposes it is advisable to keep a narrative of what the code is doing in the company’s Metadata Repository. The text below generated from ChatGPT is an example of the output when ChatGPT is asked to explain.

Note, further on in this article the same principle is used to identify bias in code.

c)??Potential Saving

The table below illustrates where potential savings could occur in the Salesforce / ChatGPT development.

The table assumes the CET Salesforce framework product has been downloaded and installed and no programming is associated with building the framework infrastructure. It is coding only.

Some fields (such as Text) will not be affected by ChatGPT code generation to copy and paste: Others will be fully impacted, and another group will only be partially impacted.

In some cases where large computing calculations are necessary, and beyond the formula field capability, the code would be executed in a Salesforce Workflow facility and the value inserted in the regular number field. Those type of calculations, although within the ChatGPT capability, are not included in the calculations below. In situations where complex formulas are necessary, and performed by ChatGPT in Workflow, the savings would be enormous: Especially, if payroll and benefits were developed on the prototype platform and then recreated on the Holochain platform. ?

Analysis of Salesforce Saving

An analysis of the table above indicates:

a)??????Full Focus of ChatGPT. If a formula needed to be constructed manually to perform a calculation 768 fields would be affected. If an average saving of 90 minutes (written and tested) per field was realised a saving of 69,120 minutes or 1,152 hours would be saved.

b)?????Partial Focus of ChatGPT. If only 45 minutes was saved for the partially affected group a further 65,070 minutes or 1,084.5 hours would be saved.

c)??????Total saving. Based on one citizen developer working 40 hours per week, that would result in 28.8 weeks saved for the fully impacted and 27.1 weeks for the partially impacted group. A combined 55.9 weeks or over one year’s work would be saved.

Estimate of Holochain Saving

From a Holochain perspective: Even if CET build the application framework and ChatGPT code is used to construct the Zomes (Equivalent of fields) syntax, the Salesforce savings could be doubled because of the newness of the Holochain product and the learning time that would be saved.

Initially, it would be beyond citizen developer capability to program and test the Holochain application. For experienced developers the time saving could be close to two years work for one person.

Overall Saving

The pre-built preparation work represents a saving as well as ChatGPT coding time: The saving to companies by downloading the CET application framework prototype infrastructure (100+ objects, 2,000+ fields) and the Metadata Repository described in the following section amounts to hundreds of hours work for citizen developers or internal IT resources.

d)?The Metadata Repository to Record ChatGPT Code

Internally, each company should maintain an HR Metadata Repository or Registry where all information relating to data or programming logic can be stored. Also, when analytics is performed and the original raw data has been manipulated it is advisable to keep track of what took place for audit purposes. The Metadata Repository serves as the equivalent of GitHub used by developers for code storage and version control.

The CET Metadata Repository diagram below illustrates how a client’s Metadata Repository could be structured and the type of information that will be useful for continuity and consistency purposes. CET provides clients with a downloadable pre-built Metadata Repository on the Confluence platform ready to be company branded and start populating with company specific information. The CET downloadable “space” has all 2,824 fields arranged for easy menu navigation and ready to edit.

The formula copied and pasted from Salesforce in the earlier diagram is shown in the METADATA box in the diagram above.

e)?Using ChatGPT to Detect Bias

In the simple hypothetical example below an AI application has been used to select the best candidate for a position. If an inhouse program was written to select the best candidate most HR business professionals would not be able to read the code or would find it hard to gain access to the code to look for intentional or unintentional bias. In the example below bias was intentionally embedded in the code to only select males. However, the narrative generated by ChatGPT describes the content of the program and identifies the bias to select males in a format that is easy to read for HR professionals.

For audit purposes and management confidence in a fair selection process this tool is extremely valuable for HR. No knowledge of code is necessary.

The following shows the request to ChatGPT and the response.

The response below from ChatGPT is a combination of code and narrative to explain what is contained in the code. The ChatGPT response could be stored in the company’s Metadata Repository for easy reference.

ChatGPT Response

Note that this code is just an example and should be adapted to fit the specific needs and data structure of your Holochain application.

Disclaimer: The products mentioned in this article such as Salesforce, Holochain, Confluence, OpenAI (ChatGPT), and claims made in this article about their potential for Web3 transition involvement, is based on hands on experience in some instances and research and marketing material for others. The products will be used in a Proof of Concept project commencing soon.

My next article about Web3 transition will focus on the Web3 Network Layer, the Holo hosting environment and their agent centric governance model, the IPFS provisional network assembly and the Consortium for Decentralized HR governance role to ensure participant authenticity. ?

For more information please visit the Competitive Edge Technology or Decentralized HR (DeHR) website or contact [email protected]