AI is undoubtedly the buzzword of this decade. Its capabilities, and potential to change the way we live and work, create headlines both inside and outside the tech sector. AI is creating opportunities (and concerns) in most industries around the world, and content management is no exception. In the next ten years, the way we manage critical business records, and digitally preserve valuable cultural items for future generations, will see a step-change.
Algorithms capable of making decisions have been in existence for decades. But with the rise of AI, machines are developing the ability to process data and react like a human being. Thanks to this ‘machine learning’, computers are capable of remembering, processing and learning from data sets too vast for one human to analyse.
However, these machines are also capable of picking up un/conscious biases displayed by the people feeding data into them, and of making decisions at pace on a vast scale that if even slightly flawed can have rapidly damaging consequences. So, what can humans do to ensure that we manage the rise of the ‘robot archivist’ and make the most of a combined human/digital workforce?
Show me the data: How organisations can make best use of their most precious asset
Ask not what you can do for AI…
The most valuable way AI can be put to use in content management across a business is in categorising vast amounts of information, making it more findable, driving actions instantaneously, or managing security settings.
Emails, for example, represent many thousands of potentially valuable or sensitive items, and added to the mix the fact they combine personal and professional communication, it’s clearly an increasingly challenging issue for human teams. It simply takes too long to trawl through every email and decide whether it needs to be kept for either security or value-adding reasons. Also, businesses shouldn’t underestimate the expense that comes with storing large amounts of useless material – it can make a real difference to the bottom line.
Enter AI. By utilising the machine learning aspect of the technology, humans can ‘train’ AI to do this work for them to a degree of accuracy which matches human efforts. Giving the system a thousand emails, say, half of which are marked to keep and the other half to delete, allows the software to learn a much more developed system of classification. The need for human guidance is lessened, and over time only decreases, as we feedback into a system and support its natural learning process.
The key to harnessing unstructured data? Visualisation
Robert Dagge, managing director at Dynistics explains to Information Age how unstructured data exists in every organisation and how businesses can make use of it through data visualisation tools. Read here
Humans, meanwhile, can put our non-replicable skills (decisions that require a high degree of empathy, or heavily experienced-based choices), to better use and drive more value across the system.
Another really exciting way records managers can put AI to work in their systems is the process of image recognition. Whereas it would take a human the best part of a decade to work through a small company’s legacy database of pictures (perhaps to decide which images are relevant to a certain collection, or even more challengingly to find a specific one), companies like Microsoft, through their Azure tools, are enabling AI systems to do that job in hours. ‘Computer Vision’ is revolutionary AI that has obvious benefits to organisations, from charities to security agencies.
Ethical challenges in digital preservation
One of the hottest topics in the AI debate is how it will be governed, and, essentially, who carries the can if things go wrong – we’re well aware that AI can and will make errors. Bias in the training process can appear in the finished product, and with digital preservation, these mistakes can be replicated on a much larger scale than was previously possible.
This is where AI’s capacity for ‘heavy lifting’ can work against us. Mistakes made in inputting can lead to inappropriate record descriptions, the wrong tags, or the incorrect action assigned. Do we blame mistakes on the algorithm, the training dataset or the host? How does AI explain its decision? More importantly, how do we ensure that we don’t reach that situation in the first place? The power of human users to correct the assertions of AI changes future evaluations and creates future improvements, but it’s a future where unconscious bias can creep in.
Some ideas include making training datasets public and enforcing industry-agreed standards on what AI is permitted to handle; a narrower remit minimises the potential risks.
AI ethics: Time to move beyond a list of principles
More than the sum of its parts
The real solution to empowering businesses to make the most of AI lies in involving the human in the right part of the process. Rather than replacing the human workforce, AI works best alongside the preservation experts, enhancing their work, speeding up the process and opening up a new world of possibilities for the world of content management. In this way, the technology can be monitored and work within the parameters that it is needed, without running the risk of it gaining too much control. Technology can do the heavy lifting, with a boost up provided by the sentience of human-beings.
In a world where almost all information is digital, businesses know they need to address the issue of enterprise content management. Now, new advents in processing power and emerging technologies like AI are empowering organisations to build a feasible strategy for the future. Those that are embracing artificial intelligence the fastest now will see the greatest benefits soonest in a rapidly-developing business landscape.
Written by Jon Tilbury, CTO, Preservica