CIO Series: Why Does AI Matter when Migrating Content to Office 365? – Part 2
This post is part two of a two-part series. Find the first part here.
AI as Actual Intelligence
Think back to the story from the first post. One thousand people would be taking 15 years to digitise and order content. It’s not viable. The juice is not worth the squeeze.
If we can automatically give context, automatically enable findability, and automatically add value, the juice is sweeter. We have extracted.
AI makes this possible. Deep Learning (DL), Machine Learning (ML), and Cognitive Services (CS) enrich content. Enriched content has value.
Imagine being able to find all content of a given type, for a given project, about a given customer or a given product – all without having to have humans mark up or tag your content.
Imagine having the ability to identify trends in how information was received by customers (for example, common words used in successful proposals).
Imagine being able to immediately locate all information about a given subject in the case of litigation.
Imagine a world where transactional data (such as an invoice, purchase order, or prescription) was available to all relevant information workers as opposed to just those with access to finance systems.
Imagine never having to worry about document security ever again because permissions were automatically set based on categorisation – with no user intervention required.
I am sure you see my point. Enriched data adds value.
The holy grail for most organisations is improved decision making. Business Intelligence on transactional or structured information is complex to enable, costly to deliver, and often requires specialist skills. If we can extract business intelligence from our unstructured data, we can immediately amplify our sample sizes, narrow margins of error, and improve decision making.
AI makes value extraction possible.
Making the Impossible, Possible
It simply isn’t practical to attempt to add value to 3 miles of content manually. Indexing (such as that possible with OCR’d digitised content today) brings some value, but only where human intervention has done the heavy-lifting of determining the classification schemas, taxonomy, and geography.
Full-text indexes don’t analyse, don’t contextualise, and don’t interpret – they label based on brute-force lookup, and you have to tell them *what* to search for.
AI solves these problems, and more.
AI can discern patterns, similarities, contexts. AI can derive connections where (even to expert humans) none appear to exist. AI can detect sentiment, identify anomalies, and infer meaning.
The benefits are apparent.
A task that would be time-consuming, error-prone, and expensive can be automated with faster delivery, reduced error rates, and lowered cost.
Despite the vast amounts of information that may be involved, AI can enable value extraction at the speed of the cloud. One thousand operators may take 15 years to digitise and file billions of documents, but AI would be able to extract orders of magnitude more value in a fraction of the time. AI saves us time.
It gets better.
As the AI models learn, they will improve. What matters here is that the content that the AI as *already* classified will have its classification improved as new data points are understood in the models. Overall quality increases as the system gets holistically smarter. AI extracts more value.
It keeps on giving.
Humans are prone to errors. Evidence suggests that the categorisation aspect of content projects tends to run to error rates of 1-2% or so with ‘non-categorisation’ events running at 2-3%. This total error rate of 5% (and with good practice suggesting that if sample batch failures occur entire batches of content need to be re-categorised resulting in a net 100% failure rate for that batch) creates enormous cost.
Tiredness, irritability, social media, and boredom are not a factor for AI so we can reasonably expect the use of AI to drastically reduce error occurrance. AI lets us take the humans, with their fallibility and frailty, out of the loop. AI reduces errors, lowering cost.
AI and Migration to Office 365
Migrating to Office 365 is conceptually no different to the digitisation story I have been using to make my point. Moving content from the old to the new is at the very root of an Office 365 migration – it doesn’t matter what you are migrating from, the benefit opportunity from leveraging AI remains the same.
What if you could analyse content within its source system to provide insights that assist in determining both structure and taxonomy for improved IA?
What if content could be categorised as it is migrated, applying metadata to increase findability?
What if duplicate content could be removed before it migrates, reducing the amount to move?
What if sensitive information could be marked appropriately and permissioned correctly?
What if the categorisation of information resulted in smarter placement within the geography of the destination?
What if non-compliant or no-longer-needed content could be identified and removed prior to migration?
The opportunity for improvement is significant and almost endless – all thanks to the possibilities brought to the table by AI – and made practical due to the speed at which AI can perform not matter how much source data there may be.
Replacing 1,000 fallible humans with a single, relentless, goal-oriented, ever-energetic AI is possible today with Content Halo from Proventeq. Interested to learn how?