|
Putting
Knowledge to Work
Knowledge
management has spawned a lot of theoretical discussion, but how do you
put it to work to achieve – and quantify – real benefits for
businesses? Peter Turnbull, SER’s UK head of marketing, goes back
to Shakespeare to bring to life working solutions for companies today.
It is a proposition almost universally acknowledged today that companies’
ability to exploit the knowledge of their people, of their processes and
of the raw information flowing throughout their businesses is a –
if not the – major source of competitive advantage in the
global economy (old and new).
This applied knowledge, or intellectual capital, lies at the heart of
today’s frenetic transition from industrial to knowledge-based economy
and from physical assets (such as land and materials) to the intangible.
Ten years into the longest recorded economic upswing in US history, management
gurus daily preach the gospel of technology, globalisation and competition.
We celebrate the speed of change and the potential for rapid growth and
innovation.
But the keyword is potential: where and what are the tools and
techniques which will allow us to fulfil it?
The Knowledge Barrier
To be of value, data needs to be captured, stored as information, organised
into knowledge, personalised, distributed and shared, adapted and exploited.
In itself this statement is true of all human historical development and
it applies to all industries – not just those with high technology
flavours.
But what is different today is that we have focused information and communications
technology with great efficiency to deliver a quantum leap in both the
volume, and speed, of new information generated. Problem
is: it threatens to engulf us all.
In this context, the manual knowledge-labourer is actually a barrier.
He/she can only read, absorb and apply 50-100 pages of information each
day and, quite literally, sometimes cannot see the wood from the trees.
Daily, major innovations – and whole series of dependent innovations
– may remain unknown unless we augment humankind’s intellectual
processing power to learn, share and target knowledge. Arguably this challenge
- to build a complementary enabling technology which will facilitate a
radical improvement in information management and decision-making –
is the most important current research task. We need to break decisively
free from the procedures of the past and the deployment of conventional
rule-based software.
Beyond biology
Somewhat elegantly, the key to escaping this biological constraint lies
in one of nature’s great survival tools – mimickry. In this
case to mimic, and accelerate, the human learning process in the form
of electronic neural networks and by using a special patented algorithm.
At the heart of this software is an integrated classification and search
engine. It classifies stores and sorts information in ways that were previously
only possible if individuals read it personally. Crucially, this engine
learns by example: it is trained by people who show it how they would
read and make decisions on the basis of available information.
The basis of the software learning process is a number of groups of sample
documents. These are collectively known as a ‘learn set’.
What we are talking about is much more than document management or workflow
(although they are part of it). Rather it is ‘knowledge logistics’
– that is bringing learned and applied decision-making to a mass
market for the first time.
The Bard in question
But what does this mean in practice? To focus the potential, a recent
experiment tasked a new system with solving the riddle of Henry VIII –
a play whose authorship has been in dispute for 150 years but which is
conventionally attributed to a partnership between Shakespeare and his
highly popular contemporary, John Fletcher.
The system ‘learnt’ Shakespeare, Fletcher and (as a control)
another contemporary, Christopher Marlowe by being introduced to selected
works from their canons – the ‘learn set’ defined above.
Then it was shown a variety of plays by other authors, including Henry
VIII.
The result, where no other play scored above 16% probability of authorship
by the Bard, the verdicts for Henry VIII were more or less conclusive
– even in scenes often allocated to Fletcher.
This finding, however, is not the punchline. Once inputting was complete,
the system delivered its results in just five minutes. By contrast, to
undertake much more limited and inconclusive textual analysis manually
would take many man-months - if not man-years.
Scene-change: the Bard at work
If now we time-shift from the turn of the seventeenth to the beginning
of the twenty-first century and move from dusty folio editions to electronic
mail, we can begin to assess the commercial impact of such a system.
Simply step into the average company mailroom. Watch the staff there sort
and classify a daily flood of information. Follow that information as
it trickles through the organisation during the day and observe it conflict
with heterogeneous, often non-compatible sources flowing from websites,
email, fax and voice.
In such a context, implementing knowledge-enabled software to learn highly
repetitive actions such as reading, evaluating, routing and indeed re-defining
work processes adds dramatically to an organisation’s efficiency.
This is illustrated specifically, for example at AXA Colonia, one of Germany’s
leading building and loan associations, which processes over 300,000 items
of customer correspondence each year. Previously this was a highly labour
intensive mailroom activity requiring the four-strong staff to spend several
days each month manually to scan and classify incoming mail and then to
distribute it to relevant departments for further processing.
Now using knowledge-enabled software, all incoming correspondence is classified
into 50 distinct categories and the policy number extracted from each
to provide a link to the policy history. Documents are then distributed
electronically and tracked through the remainder of the process.
A second more specialised example is provided by the Mannheim head office
of German DIY pioneer, Bauhaus which operates over 170 facilities in nine
European countries. Daily the recipient of over 6000 supplier invoices,
it has deployed knowledge-based software to implement automatic invoice
processing and integrated document management and to deliver automated
data entry and auditing processes.
From efficiency to value
There are also enormous ‘effectiveness’ benefits to be harvested.
Release your staff from their roles as ‘information caretakers’
and they can work with the technology to become highly proficient ‘knowledge
generators’.
This process is underway at Union Krankenversicherung (UKV), Europe’s
eighth largest private health insurer where a similar system was recently
introduced – initially to process an estimated 800 daily foreign
travel insurance forms, later to manage an estimated 10,000 claims forms.
The ultimate goal of the UKV implementation is to set up an information
management infrastructure by which data will be sorted entirely independently
of data type (e.g. fax, email) and style (form, free writing). This will
support delivery of the high volumes of information required in health
management – specifically in the claims arena.
The price of intelligence
Meanwhile, another major insurer is going the next step and using the
system to learn, and analyse, the often highly complex and technical content
of medical claims documentation. Where previously manual clerks could
only validate simple key points (e.g. date, source of claim, type of illness),
knowledge-enabled software is learning and assessing the core content
– and checking for fraud or medical error.
This is the equivalent of moving the Shakespeare debate (reviewed above)
beyond simple stylistic issues such as word frequency into an understanding
of complete patterns. But in this context, such intelligence has a price:
potentially a 20% saving in a multi-billion dollar annual bill, according
to the company.
Riding the information wave
As we put knowledge to work, it’s critical to appreciate that the
type of intelligent learning engine which I have described here is not
limited to storage and retrieval of language-based knowledge. It can also
learn how knowledge workers move information within an organisation –
that is procedural knowledge which, in turn, can open up the DNA of any
business.
Meanwhile, as humans gain more time so they can add real value to the
knowledge development of the organisation. Finally, traditional IT can
assume a new role as a powerful tool supporting knowledge workers in their
personal and collective knowledge creation and managements.
Today, in the aftermath of the dot.com collapses, it’s fashionable
to decry those who predicted enormous new economy gains. Yet given learning
engines which can allow humankind to capture and deploy knowledge thousands
of times faster than at any period in recorded history, their pronouncements
may come to seem ultra-conservative. Our ride on the information wave
promises to be an exhilirating one.
- ends -
Peter Turnbull is UK head of marketing for SER Systems. He can be reached
on peter.turnbull@SERuk.com
|
