“The last 10 years have been about building a world that is mobile-first. In the next 10 years, we will shift to a world that is AI-first.”, Sundar Pichai, Google CEO, October 2016
Artificial Intelligence (AI) is at a tipping point, leading a watershed shift to digital intelligence by discovering previously unseen patterns, drawing new inferences, and identifying new relationships from vast amounts of data.
This shift from the current Programmatic Era to the new AI Era will be transformative and disrupt companies and entire markets. To accomplish this, AI/Cognitive solutions will require entirely new skill sets and job descriptions. Companies of scale need to gear up these new jobs and transform their organizations for this future.
5 new jobs you need to be hiring for to transform from programmatic to cognitive era:
| From: Programmatic Era | To: AI/Cognitive Era |
|---|---|
| Database Administrator (DBA) | Taxonomy (or Digital) Curator |
| Systems Engineer | Cognitive Architect |
| Data Analyst | DEADS: Data Engineer and Data Scientist |
| Content Administrator | Natural Language Processing / Cognitive Scientist |
| Programmer | Machine Learning Engineer |
As background, every person creates a gigabyte of digital exhaust every day – web/search, mobile, location, social, transaction, audio, vision, etc. Traditional IT departments are overwhelmed by Big Data and challenged to keep up. This coupled with advancements in cognitive services are creating a compelling business opportunity (and associated risk) for value creation.
We currently operate in the era of ‘programmatic computing’, where data analysis involves heuristically searching for patterns in limited data sets, then performing operations on the result. Conventional computers have difficulty working with Big Data because their programming requires structured information (data organized in spreadsheets, databases etc.), while 80% of the world’s information is this unstructured digital exhaust.
For illustration, let’s say you are an online retailer who wants a computer program that identifies images of a ‘boot’. Currently it’s not possible to algorithmically specify all features that will enable correct identification. Boot images vary by brand, type, style, gender, shape, size, color, background, lighting and a myriad of other attributes. There are too many variables to write a rules set. Even if we could, it wouldn’t be scalable, as we’d need to write a program for every type of boot and UPC we wanted to identify.
Enter AI/Cognitive which represents a new era in computing. The promise of AI is to shift the complexity of managing systems from the programmer to the program. These systems take a different approach – they Understand-Reason-Learn (URL).
Natural Language Processing (NLP) for instance is used to understand unstructured information. Developers do not program cognitive systems in a conventional sense, but rather a corpus of information is created for a specific domain set. These systems are built by curated value pairs. For example, you teach a cognitive system that Argentina is a country, that Patagonia is a region, and so forth. Cognitive systems tend to gain knowledge, build neural connections and improve via supervised-learning over time. As user interaction increases, experience is gained and mistakes are minimized. Significant corporate value is created via the formation of this domain specific corpus of IP information.
Deep Learning (a branch of Machine Learning) uses a ‘neural network’ which receives an input, analyses it, makes a determination and is informed if its determination is correct. If the output is wrong, the connections between the neurons are algorithmically adjusted, which will change future predictions. Initially the network will be wrong many times. But, as we feed in examples, the connections between neurons will be curated so the neural network makes correct determinations on most occasions.
A retail cognitive assistant is a sample Cognitive/AI application. The objective is to enable a consumer to ask a natural language query such as “What boot should I buy to go hiking in Patagonia this June?’” The digital assistant would understand/reason that the trip is during the rainy winter season and recommend accordingly. These systems: learn at scale, understand with meaning, reason with purpose and interact with humans in natural ways, with the goal of improved customer experience.
More detailed job description of new roles you need to hire for:
- Taxonomy (or Digital) Curator: This term is taken from the Health Science field, early AI adopters, where a Taxonomist identifies and classifies data. This position applies the process of content curation of a domain specific corpus of information. Data is ingested and content is curated in the form of question/answer pairs. Additional tasks include building of indices, knowledge graphs and ongoing curation based on user interaction. The process frequently involves sorting through vast amounts of public and proprietary information and presenting it in a useful and meaningful format related to a specific domain topic area. The curated content is frequently presented in digital format, such as in a Learning Portal.
- Cognitive Architect: Responsible for designing, creating, maintaining and communicating the overall roadmap and architecture of the Cognitive Solution. This includes Big Data strategy (both internal structured and external unstructured data); identify data components, select analytics engines and all the system and data interactions. The stack includes Big Data, Advanced Analytics and AI services. AI includes a broad range of techniques ranging from Machine Learning, Deep Learning, Cognitive Services, Natural Language Processing, and Speech and Vision recognition services. Cognitive Services are offered by leading software companies such as IBM (Watson), Microsoft, Google (DeepMind), Amazon (Lex), Facebook (FAIR), Salesforce (Einstein), start-ups and academic research institutions.
- DEADS (Data Engineer and Data Scientist: Responsible for design and implementation of processes for data sets used for modeling, data mining, and measurement purposes. This includes Big Data concepts such as Data Swamps. Lakes and Reservoirs, each with varying degrees of the 3 Vs of Big Data – Volume, Variety and Velocity. Hadoop and Spark are popular frameworks.
- Natural Language Processing (or Cognitive) Scientist: Uses NLP services such as speech-to-text to mine unstructured data. Data sources are frequently external unstructured data such as social, audio, web, weather, etc. This data needs to be converted to a structured format for purposes of reporting, analysis and insight creation. Techniques include: predictive modeling for sentiment analysis, clustering for document analysis, and anomaly detection. An example is Twitter, where Tweets are converted to measure sentiment, or life event detection (going to college).
- Machine Learning Engineer: The objective of ML/DL is to develop a predictive engine for a specific use case. An algorithm receives information about a domain (say, movies a person watched) and weights the input to make a prediction (probability of watching a different movie in the future). By passing the task of optimization (weighting) to the algorithm, the computer learns over time – predictive quality improves with experience. Deep Learning (subset of ML) itself has over 15 approaches such as Random Forest, Bayesian Networks and Support Vector Machines. Required skills include Python (popular ML language), probability/statistics, Big Data and distributed computing.
- Additional fields of AI with new job requirements include: Customer Experience (Digital) Agents building Chatbots, UI/IX (User Experience), Automation Engineer (robotics), Automotive (self-driving vehicles), with Healthcare having unique requirements.
Many of the positions in ML/DL require advanced-math skills. Curator jobs, on the other hand, value domain subject matter expertise, which does not necessarily require advanced degrees. Ginni Rometty, CEO IBM, recently introduced these ‘New Collar’ jobs in an open letter to President-Elect Trump (https://www.ibm.com/blogs/policy/ibm-ceo-ginni-romettys-letter-u-s-president-elect/).
In the coming years, Applied AI will be incorporated natively into most corporate functions. Consider for example the range of processes that will be incorporated into the Human Resource (HR) function as follows:
- Recruitment can be improved with enhanced targeting, intelligent job matching, skill alignment and partially automated assessment;
- External unstructured data such as social media publishing can be used to develop candidate personality profiles which can be more efficiently matched to job requirements;
- Workforce enablement can be improved as content better suited to the employee is recommended;
- Workforce management can be enhanced by predictive planning of personnel requirements and likelyprobable absences;
- Workforce allocation can become more efficient For example, retail staff can be more efficiently positioned based on: local conditions (E.G., rainfall lowers store foot traffic), local promotional events (E.G., Cinco de Mayo) create assortment and promotional opportunities;
- Employee churn can be reduced by predicting that valuable employees may be at risk of leaving.
Over time, expect the adoption of AI/Cognitive to become normalized. AI will become a standard part of the toolkit, initially improving existing processes and then reinventing them as part of a broader digital transformation process.
So where to start? Ultimately, it is strategy rather than technology that will drive value creation. A strategic assessment is recommended to build an over-arching vision, determine opportunities and identify skill gaps. From there, low hanging fruit opportunities can be incubated. The transformative path from incubation to integration must include leadership, culture and organizational change considerations. A key question is: Are you the disrupter, or do you want to be disrupted? Companies need to accelerate the Cognitive Digital Transformation process and acquire skills at pace for business value creation (or risk mediation).
As always, I’m interested in your thoughts or questions…
