Navigating Organizational Growth with an Interaction Record and LLMs

When an organization is young, it's natural to begin adopting tools to help run the business—creating social media accounts, adopting various tools for knowledge sharing, growing a codebase, connecting various integrations to help run the business, etc. The problem is exacerbated as the organization grows and teams need new tools to perform at their job. Before we know it, the organization's data is everywhere in different databases across the internet. This is data sprawl.

Let's take a look at some downsides of data sprawl:

Data is too far apart. We depend on other teams to accomplish anything, which further strains the team's limited bandwidth. The lack of planning and coordination makes things disjointed and feel slow. And now an existential fear permeates throughout: growth is bad and growth will hurt that which is already built.

Sound familiar?

The natural sprawl of data that occurs during growth is nearly impossible to wrangle across the myriad of tools and SaaS products that have been adopted.

However, combining two concepts may help us here:

1. Keeping an Interaction Record

2. Large Language Models (LLMs)

An Interaction Record (IR) is one or more data sources where you store historical events that impact your organization. This should include anything that adds context to the existence of your organization: financial transactions, messages, analytics, emails, documentation changes, and more. This record should include first-party information that the organization controls, like emails, documentation, product information, and organization structure. The IR should also include third-party information that relates to the organization; information that does not live inside owned data sources, but rather another organization's data sources. The more data, the better.

Once this data exists in a single accessible store, it becomes possible to train and fine-tune an LLM on the data. An organization would have incredible insight into its operations.

While this doesn't fix data sprawl, it weaponizes it:

  • Cross-app data synthesis for previously undiscoverable insights
  • Internal audits of employee performance at every engagement level
  • Artificial intelligence understanding of business operations, with recommendations
  • Potential bidirectional interactivity of artificial intelligence operations
  • Ease of data discovery in legal diligence
  • Employee-specific insights
  • An LLM personal assistant for each employee

Now an organization can use tools to understand this dataset, build on top of it, and communicate directly with it to have a fully-encompassing view into its operations.