Protecting Human Intelligence With AI

Russ Kennedy discusses how modern enterprises can uncover and extract knowledge from their distributed data through AI services while protecting human intelligence.

May 15, 2024

The tight labor market in the U.S. is accelerating a technical exodus. Skilled knowledge workers have more employment options, so more people are moving around to advance their careers. In the process, they are taking unique business and operations knowledge with them.

Yet that specialized human intelligence isn’t entirely walking out the door. Some of it is left behind in structured and unstructured data, or what we all know as files. Enterprises have not had the tools to uncover and extract knowledge from this distributed data, but I believe this is going to change very soon with the help of AI services.

Countering Technical Exodus

In my last few articles, I introduced a few essential steps organizations need to follow to prepare for the enterprise artificial intelligence (AI) tools of today and tomorrow. This piece is slightly more speculative but points to the sort of possibilities that may be available if you consolidate and curate your data. Imagine enterprise AI services capable of uncovering, collating and synthesizing the knowledge and history of employees, effectively creating a digital twin ready to share that person’s key ideas and contributions.

Where would this tool find such information? Consider the output of a standard knowledge worker over the course of a year. They create and edit documents, spreadsheets and presentations. They correspond with colleagues, prospects and customers via email and messaging apps. Maybe they work in specialized design applications or comment on files in a platform like Microsoft Teams. There will be calendar notices, notes, videos and all kinds of back-and-forth within groups and departments.

The data the average person generates simply by working inside your organization is going to be incredibly diverse and widely distributed across different systems and applications. No former coworker of the departed employee is going to want to pull all that together—if you assign someone that task, you risk prompting another resignation. But if you consolidate all of your organization’s data in a way that makes it easily accessible to approved enterprise AI tools, then the diverse, distributed nature of this data is no longer an obstacle. After consolidation and curation, those scattered digital breadcrumbs become easily accessible to AI.

Preserving Human Intelligence

I don’t believe the virtual representation of a departed worker’s knowledge would be a digital twin in the sense that it could perform that person’s job. However, this specialized AI tool could represent a departed worker’s knowledge, generating information from their past work data. This approach could assist in onboarding new hires by providing summaries of predecessors’ contributions, accelerating the learning process.

In an engineering firm, for example, the tool could consolidate complex renderings and digital models, aiding new hires in navigating previous projects efficiently. Beyond engineering, this concept could benefit various industries by acting as an internal research assistant, enabling easy access to past employees’ work-related content, and benefiting both new and existing team members.

Investing In Data Stewards

To make this work, an organization will need relevant data to train the AI tool, which will in all likelihood be a large language model (LLM). You don’t want to expose this tool to old and outdated files and information, so you will have to curate the data associated with the individual who has left the organization. If you feed AI tools outdated data, they will generate outdated results. Curation is critical if you want useful, relevant output.

There are various data intelligence tools available to help curate unstructured data and ensure the data that you feed into an LLM is timely and accurate. These tools provide information about the unstructured data in your environment. Another tool to consider is a sophisticated search engine that will index the data based on its content and allow administrators to query the resulting index to identify the most relevant content. Be sure to experiment with data intelligence tools and search tools to find the ones that meet technical and financial requirements.

The final piece necessary to institute this sort of system would be human talent. Data consolidation and curation tools will get you ready to leverage AI services, but organizations will need to invest in data stewards responsible for understanding the different data types and the relevancy and recency of data. You will still need people to make the critical decisions and ensure that you’re leveraging the right data.

This process is typically referred to as the “human in the middle.” As stated previously, results generated by AI tools are only as good as the data that they have available. In most cases, organizations will want to have humans verify the results generated by AI tools for accuracy and the absence of bias or hallucinations. This “human in the middle” concept will ensure that the results generated from a previous employee’s work product will be usable by those who take on that individual’s tasks.

Final Thoughts

This will all require some work, but the long-term rewards will be significant. By consolidating and curating your data and then building and maintaining a team of data stewards who intelligently manage that data, you will be setting your organization up for the AI services of the future—the knowledge tool described here or one of the many other exciting solutions yet to come. I suspect you will also reduce the risk of technical exodus, as you will be building the kind of company that people do not want to leave in the first place.

Ready to dive deeper into a new approach to data infrastructure?