Faster. Cheaper. Better? AI is being heralded as the latest solution to the perennial problem of workplace efficiency and costs. But as with any advance – technological or otherwise – it’s not that simple.
Sometimes AI makes a lot of sense. For instance, it’s perfect if you have a huge data set because it can analyzes data faster than any human ever could. After all, it’s designed to look for patterns and serve up answers based on them.
Artificial Intelligence
But while AI may provide answers like a human, it’s not actually thinking like a human. As explained by Walter Quattrociocchi of Sapienza University in Scientific American, “[LLMs’] architecture makes them extraordinarily good at reproducing patterns found in text. It does not give them access to the world those words refer to.”
In other words, AI lacks the human understanding that allows us to make nuanced and informed judgements based on experience, judgement and context.
And as many companies run to adopt AI, a significant number of professionals feel uncomfortable – not just about the impact on their jobs but on the decisions being made.
According to our survey of archivists, historians, brand managers and attorneys, 45% of respondents report feeling pressure to use AI for tasks they believe are better suited to human judgement, such as customer service and legal and corporate decisions.
The tension between technology and knowledge
The growth of AI is also breeding concern about the impact on human expertise and judgement. Our research found that 62% of professionals are concerned that AI could devalue or replace human expertise in their field.
So, there is a tension between technological advancement and domain-specific knowledge and judgment. And this is a pattern we’ve seen before.
Wasn’t the Internet supposed to replace librarians? Instead, librarians became more technologically savvy and evolved into information professionals. They are now part of every sector – not “isolated” in archives or libraries – and are leading the charge to embrace technology, not reject it.
We’re seeing this with taxonomists now. In theory, AI could analyze the entire Internet, but you need humans to catch the outliers and make sense of the data based on their frame of reference and the nuances.
AI can be additive
If we take the right approach to AI, it can actually add value and meaning to the work humans do – not just by freeing up their time but by allowing them to expand our professional contributions into new areas.
We already know that AI has limitations. For example, it’s only as good as what it’s scraping. Authentication and provenance have always been the first step in knowledge management. It’s the same with AI. You still have to check that the basis of the information is accurate.
Second is completeness. The world isn’t digitized. At the U.S. National Archives alone, more than 12 billion pages of records remain largely undigitized – about 96% of the collection. Some records are being digitized, but some, including information that we perceive as high value, will never be – WWII records, for instance. That makes a vast amount of information totally inaccessible to AI and LLMs. We need humans to add those missing pieces to any analysis.
Third is trust – not just in AI outcomes but in the organization using AI.
Today, people often automatically assume that the next development will expand and improve on what came before. AI warrants more scrutiny because of its ability to generate and hallucinate information. While inaccurate and biased information has always been a problem – whether due to bad sourcing or deliberate misconstruing of the facts – AI raises the stakes with its ability to be convincing and its speed; it becomes the “easy” way to get a summary or update.
How do we solve this? By aligning the deployment of AI with professional judgment, ethical standards, and sector-specific expectations, particularly in roles in which human expertise remains irreplaceable. We call this human-guided AI.
Human-guided AI
There are four principles to human-guided AI:
- Experts remain in the loop at every stage.
- AI augments research; it does not drive conclusions.
- Context, provenance, and ethical considerations are non-negotiable.
- AI is used to enhance accuracy and deliver efficiency and speed where appropriate, not to shortcut rigor.
Human-guided AI gives companies a framework for the appropriate and effective use of AI. Here are three examples:
- A politician is retiring and donating his records to his alma mater. He might have 500 boxes of letters from constituents. Only humans can weed through those boxes and determine which letters should be removed due to personally identifiable information. However, you can use AI to scan emails and other electronic records, speeding up the process.
- A company is publishing a book celebrating its 300th anniversary. Archivists discover records showing it bought and sold slave labor in the 1860s – how do you want to treat that information? Only humans can decide. But while AI may have assisted with the discovery, it isn’t going to determine if the fact should be included in the book or not.
- A tobacco company is sued – AI can find the relevant electronic records, but only humans can determine whether certain documents need to be included in discovery or not.
The boundaries are still being drawn
Ultimately, there will be pressure from boards and senior management to use AI. But by embracing a hybrid approach, companies can reap the benefits of AI with the safeguards of human judgement.
Equally, by applying AI to tasks that people can’t do as efficiently or as accurately, companies can free up their people to create more value and find greater meaning in their work – increasing loyalty and satisfaction.
Learn more about the intersection of AI and knowledge management in our 2026 AI Signature Research Report: Pressures, Possibilities, and Precautions.
Preserve and Activate Your Heritage
Organizations across sectors — from universities and museums to corporations, nonprofits, and government agencies — rely on strong heritage practices to protect the stories, collections, and records that shape identity and inspire future generations.
HAI's Heritage Services include:
- Archival Services & Collections Processing
- Digitization & Digital Preservation
- Records & Information Management
- Historical Research & Storytelling
- Exhibits, Interpretation & Digital Experiences
- ArchivalOne and PastView Partnership
- HeritageAdvantage CoLab
We help institutions safeguard their past so they can lead with clarity, credibility, and purpose. Let's preserve and activate your organization's heritage together.