15/29

AI – opportunity and danger

As we delegate more to machines the routine of collating clean data - and the application of our nuanced judgement of it - gives rise to the opportunity for hidden misdirection, cautions Antworks’ Mike Hobday.

AI – opportunity and danger

As we delegate more to machines the routine of collating clean data - and the application of our nuanced judgement of it - gives rise to the opportunity for hidden misdirection, cautions Antworks’ Mike Hobday.

Mike Hobday, CEO, Antworks

mike.hobday@ant.works

There is no question data grows and human knowledge expands at an exponential rate every day. Increasingly we are concerned about knowledge and insight - is it real, is bias and misinformation steering our decisions? Instead of insight public data can be at best a blind fold at worst a danger.

Professionals with a duty of care to their firm’s reputation must have concern for the data quality and efficacy represented in knowing its source.

Artificial Intelligence too, like human intelligence, is open to manipulation by misdirection through source data. As we delegate to machines the routine of collating clean data and the application of our nuanced judgement to the assessment of its relevance and quality (in the detail) the opportunity for hidden misdirection becomes an increasing concern.

Professionals in law and accounting firms have long understood the importance of the source.

The ultimate source invariably is unstructured and non-standard documents, contracts, reports, invoices, deeds. For at least the last two centuries aspiring partners and managers have cut their teeth in the routine of studying and collating information and presenting an analysis or insightful structuring of the important data contained. This is expensive and valuable work and forms the foundation of most practices even today supplemented by professional back offices.

Enter Artificial Intelligence. It is not limited to supersmart application to data sets by a new generation of data scientists but to a bridge between the analogue world of contracts, deeds and accounts and the harvesting of critical data into a form for computation and analysis. Sometimes referred to as dumb AI it is far from this and offers perhaps the greatest opportunity for firms to accelerate the harvesting of large-scale data at source and at the same time reducing costs.

Care should be taken in the newfound lexicon of Intelligent Document Processing, but a new generation of Cognitive Machine Reading solutions offers the opportunity to train AI to emulate how a paralegal or accounts clerk speed reads with confidence substantial and often complex documents to find specific information that is relevant to the purpose of the task.

Case study

Thousands of leases for property on which 4G telephone masts stand. The task of reviewing each one to determine their suitability for 5G support is considerable. Can Cognitive Machine Reading be trained to recognise the patterns in the text that answer the questions:

  • How high is the mast?
  • Does the lease and structure allow for more equipment for 5G to be fitted?
  • Can vehicles get to within 20 metres of the mast?
  • Lessor details?
  • Location?
  • Length of lease?

A cognitive machine reading solution was trained to do just that. Emulating a professional speed reading a lease, recognising word patterns in the text, and drawing inference from the document to address the questions, at scale, at speed and with high levels of accuracy. Poor quality of documents, unexpected characters and low confidence in data reported to a professional for checking.

“AI will transform the routine of professional services firms. Given many senior professionals cut their teeth in the profession by undertaking such routine early in their career there will be implications for job structures, training, and development.”
“AI will transform the routine of professional services firms. Given many senior professionals cut their teeth in the profession by undertaking such routine early in their career there will be implications for job structures, training, and development.”

AI: Art or science?

Of course, it’s a science but in art we find a clue. Art, after all, is in its beauty and sophistication, unstructured like a contract. A painting can engage us not as a portrayal of beauty at face value but, in its construction, it may carry a deeper meaning.

In order to understand that meaning, you need a key, and that key is context. The context in which a painting was created unlocks the message the artist intended to communicate to the viewer.

This is a widely admired painting by Hans Holbein the Younger. Beautifully executed for his benefactor Henry VIII. For those of us not educated in the 16th Century, this might just be a beautiful painting but add the context of its date, the knowledge of the thinking, symbolism of the time and we can unlock the real message of the artist.

To the 16th century European eye

  • Here at two men of great distinction with fraternal pride
  • From the scabbard and detail on a book we are told the men are in their twenties.
  • The foot firmly placed in the carpet circle tells us he is bold and confident. The other modest in poise and dress is a bishop
  • The anamorphic skull distorted from the front, is perfectly proportional viewed from the side perspective, tells of the presence of death amidst life.
  • Christ on the cross behind green curtains is also symbolic of the presence of death
  • A terrestrial globe upside-down beneath the table speaks to turmoil in Europe with Lutheran Reformation shaking the continent.
  • Proudly on the table a celestial globe speaks to the strength and constancy of heavenly spheres versus earthly affairs.

There is so much more to this painting but let’s pause here. As we can see the 16th century context unlocks the meaning of objects and text.

Expert training from (an art historian or) a business user

Taking non-standard, unstructured documents and extracting knowledge, facts, intent is not dis-similar to the role of an art historian. Deciphering text and images to get to the root of meaning and presenting it in a coherent form for a reader.

If the context, semantics, and purpose of a document is understood, AI can be trained by a knowledgeable business user to extract the data that routine would have an army in the back office working on.

Cognitive Machine Reading explained

The workflow is simple and illustrated here by the young clerk which AI is seeking to emulate.

Step 1 – Having extracted an image of the document from a file, AI is deployed to enhance the quality and position of the document image for computer vision. Much as we do, correcting the orientation and focus of the page.

Step 2 – Recognising the document, unstructured as it may be, the application is looking for recognisable patterns much as this experienced clerk might

Step 3 – Knowing what the document is triggers search and recovery of the data points required built on a foundation of training with a representative sample of documents. No templates here. It emulates the techniques that an experienced clerk might use to find data in a lengthy contract. Afterall no one reads a document word for word they scan read looking for paragraphs that exhibit a pattern of words or objects that reflect the subject matter being sought. Cognitive machine reading does the same. Association and inference models can provide nuanced information, whether the subject matter for example is included or excluded.

Step 4 – Assessing the quality of the information gathered and seeking corroboration from trusted sources where needed. Consulting an expert for human review where appropriate.

Step 5 – Transforming the data extracted. Data from a document is rarely extracted verbatim but transformed into the semantics of the firm and input to systems of record.

AI will transform the routine of professional services firms. Given many senior professionals cut their teeth in the profession by undertaking such routine early in their career there will be implications for job structures, training, and development. But that is a subject worth a larger discourse.

The image contained in this article - and provided by Antworks - is a copy of ‘The Ambassadors’ by Hans Holbein the Younger which hangs in Room 12 of The National Gallery in London.

ant.works

Quicklinks

NEWS & EVENTS
IN CONVERSATION
INDUSTRY
BEYOND BORDERS
FOOTNOTES

Visit the PSMG website Find us on LinkedIn Find us on Twitter Contact

Share this page