ElementsGPT delivers Actionable Intelligence

AI stands for Actionable Intelligence

AI is having a profound impact on the whole Salesforce implementation lifecycle. It offers the opportunity to dramatically increase the speed, whilst at the same time improving the quality of implementations. Arguably the greatest impact is in the planning phases where a huge amount of documentation is created. This is where the power of GPT can give 100x productivity gains.

The Elements.cloud team have been working with AI for over 5 years and GPT for the last 12 months. They’ve been looking at every activity in the planning phase to see where it can be disrupted. This is accessed in 2 different ways inside ElementsGPT:

• AI Integrated which accelerates or enables an Elements feature.  e.g. drawing a UPN diagram from whiteboard sketch, or writing a user story from a process map
• Copilot where it is a coach skilled in a certain aspect of business analysis e.g. conversation with your metadata, or process best practices.

The recent changes to OpenAI, combined with the rich Org data that we store and calculate for an Org, means we can deliver staggering insights; these could be delivered as graphs, lists, images or text. We’ve been blown away by what is possible.

Over 5 years ago, we worked on an AI product that could take a sketch of a process diagram on a whiteboard and then convert it to a UPN process diagram. Whilst promising, the results were not good enough. 6 months ago, we looked at it again and developed the import capability using AI, but again the ability to understand the diagrams, was not good enough. The recent release of OpenAI’s APIs, now makes it achievable to take any image of a process and convert it to a UPN process diagram.

Another example is a Copilot that uses metadata, dependency and usage data.

A query could be, “if I change a picklist value on the Stage field on the Opportunity object, how long will it take to update all the configurations in the dependency tree that will be impacted?” BTW the answer was 96.45 hours for our Org, because there are Flows that call Flows and Apex that call…….etc etc.  We know all the metadata, dependencies, and complexity. A query like this, would have taken 1,000s of lines of code and been specific for this use case.  But now, we leverage the reasoning capabilities of the LLM (Large Language Model) and it can deal with a wide range of queries.

It is 100% accurate. No. It is taking your assumptions on how long it takes to make changes to each metadata type. What is important is you can quickly see the ballpark effort. It is in the 80-100 hours range. Not 5 hours or 500 hours.

AI Integrated and Copilots

We have a huge roadmap of ideas to implement. The recent OpenAI release, means that some of the features that we had thought would take 6-12 months to develop, are now only 4-6 weeks away. Below are some of the features that are GA (Generally Available) or are being worked on. They are listed in the context of the implementation change cycle:

Capture and validate requirements

• Generate a UPN process map from the photograph of a hand-drawn process [AI Integrated]
• Generate a UPN process map from an image from another drawing application such as Lucidchart, Visio, Draw.io, and Miro  [AI Integrated]
• Identify business requirements in an audio file or text transcription [AI Integrated]
• Identify improvements or missing steps from a UPN process map [Copilot]
• Generate a UPN process map from an audio file or text transcription [AI Integrated]

Define work items

• Generate user stories from activity boxes on a UPN process map [AI Integrated]
• Create acceptance criteria for user stories [AI Integrated]
• Create behavior-driven development (BDD) tests for user stories [AI Integrated]
• Recommend metadata changes / additional metadata to satisfy user story [AI Integrated]
• Query Well-Architected principles [Copilot]
• Automate impact analysis [AI Integrated]

Assess impact of changes

• Querying Org metadata using natural language [Copilot]
• Generating on-demand visualizations and graphs of Org metadata analysis [Copilot]
• Generate a schema/ERD diagrams for the Org [AI Integrated]
• Auto-generate metadata documentation [AI Integrated]
• Understand Well-Architected principles vs Org configuration [Copilot]

Operationalize Copilot Prompting

• Shared repository of Copilot prompts, with a security model that controls access at corporate, Space or individual level.

ElementsGPT is an add-on that is now GA.  That means it can be switched on for any customer for an Elements Space, or for an individual Consultant.

OpenAI charges to access their LLM on a usage basis. There is a cost for every ‘token’ used.  A token is about 1000 characters; tokens are accumulated when you send a request to the LLM (the prompt) and in the answer you get back (the result). In order to simplify the implementation, the cost of tokens will not be passed on to customers and consultants.

For launch, ElementsGPT will simply be a fixed annual cost. There will be a fair use policy, as we are all entering a new world where it is difficult to predict how the service will be used.  

Set up a call and talk to one of our team, to get your Space or consulting licenses set up with ElementsGPT.

What is GA now?

We will be rolling out enhancements to ElementsGPT in weekly releases. Currently, the following functionality is GA:

Create complete user stories

• Create a process diagram or drill down from a transcript, procedure or written notes [AI Integrated]
• Generate user stories from activity boxes on a UPN process map [AI Integrated]
• Create acceptance criteria for user stories [AI Integrated]

Understand Org complexity and technical debt

• Querying Org metadata using natural language [Copilot]

Recommend better Salesforce solutions

• Recommend metadata changes / additional metadata to satisfy user story

Security, trust and governance

Who has access to your data, is at the forefront of all of our architectural design work. Firstly, we do not access customer data so we only have your business analysis documentation and Org metadata stored inside Elements. View our Data Security & Compliance document here.

Whilst this information is proprietary and important, it is not necessarily the source of your organization’s competitive advantage. Nevertheless, there are natural concerns about what AI will do with any data. 

Firstly, Elements is architected, so that it can connect to any number of LLMs. Currently, we are using OpenAI which is the clear market leader, but when Salesforce makes the Einstein Trust Layer available to ISVs, we can connect to it on a customer-by-customer basis. We connect to OpenAI via their APIs – the OpenAI privacy policy explicitly states that they will not use the data, for training their LLM. OpenAI only stores the interactions for 30 days, for audit purposes.

We are keeping close to recent changes at OpenAI’s leadership and the impact it may have on our ability to use its LLM.

The final word

ElementsGPT represents a significant advance in using AI to accelerate Salesforce implementations. It offers a range of AI-integrated features and Copilot capabilities. They can streamline the planning process and enhance productivity, whilst maintaining high levels of data security and trust. ElementsGPT is available for customers and consultants as an add-on.

Set up a call and talk to one of our team, to get your Space or consulting licenses set up with ElementsGPT.