Conversation-as-a-Service: the big challenge and big opportunity of algorithmic knowledge

To combat the accelerating speed of change, there is a growing need for a step change that leads to more productive and smarter humans. The barrier to this transition is knowledge needed by humans, whether customers or employees, to make informed and timely decisions.

Knowledge is too broad a term.

The highest form of knowledge is for informed and timely decisions.

The largest concentration of this type of knowledge is where it contains algorithms involving choices, pathways and outcomes. Selecting an option within a choice takes the user decision journey down a different pathway. The more permutations of pathways, the greater the algorithmic complexity.

Algorithmic knowledge is pervasive with its highest concentration found in documents such as those containing regulatory, statutory, policy, standard operating procedures, instructional and assessment-based content.  Because algorithmic knowledge is typically contained in documents, there is no user decision journey data for measuring the efficiency and effectiveness of the algorithm.  By their very nature, these documents are well controlled and audited. But, and it’s an important but, there is no assessment of whether the algorithm is weak, strong or even complete. And to compound matters, these documents are not subject to understandability tests.

As the speed of change accelerates, it has become clearer that the state of algorithmic knowledge in content form has become weaker and weaker as the document is no longer fit for purpose to handle permutation complexity.

The complexity within algorithmic knowledge is primarily determined by the number of decision choices and options involved. For example, if the first choice involves three options and this is repeated at the next level and repeated again then it forms a geometric progression: Level 1 = 3 options; Level 2 = 9 options; Level 3 = 27 options; Level 4 = 81 options, Level 5 = 243 options. Typically, algorithmic knowledge does not have this symmetry, but a user decision journey may involve 2 to 42+ levels.  This is demonstrable today as algorithmic knowledge transcends from documents to chatbots for delivering a conversation-as-a-service.

For many organisations, public and private sectors, their documented algorithmic knowledge is no longer fit for purpose, often regulated to shelfware for orientation purposes and to gain the annual audit tick for acceptability. As a consequence, human decisions used to complete many input forms have gradually shifted towards subjective-based decisions, which inherently lead to increases in negligence, false positives, false negatives, errors, rework, delays and handoffs. At the same time, the cost of coordination to search, organise and synthesise knowledge in a contextual form to make informed decisions is now much higher that most organisations realise.

Even with the increasing number of AI successes, it is hardly making an inroad as algorithmic knowledge continues to accelerate in volume, velocity, volatility and variations.

At least the shift towards conversation-as-a-service, involving chatbots and humans working together, has highlighted the issues around documented algorithmic knowledge. This is both a big challenge and big opportunity.