Building a Cognitive Center of Excellence

7 minute read

Forget certainty, forget traditional business cases with year one payback, and forget how things have always been done with existing technology and processes.

Maximizing the potential of cutting-edge cognitive platforms like Amelia requires more than carrying out technical integration with other systems. During the course of the numerous engagements we have underway with enterprises across the world, we are frequently asked for our thoughts on the best way to establish and nurture in-house expertise. Customers are keen to learn about successful approaches to increasing adoption of cognitive technology. Our response to these important questions often involves a recommendation to build out a cognitive Center of Excellence (CoE).

A cognitive CoE is a working organization focused on driving the adoption of cognitive technologies throughout a company. It serves as a central control point for evaluating new technologies and keeping track of ongoing projects. Most importantly, it provides leadership, best practices and support for projects and teams implementing cognitive solutions within the business.

While the specific details of building a cognitive CoE will vary from company to company, over the course of multiple Amelia implementations, we have developed a set of best practices. In this blog I’ll share the insights we’ve gained into the organizational structure and methodologies best deployed. I will also include thoughts on how we assist our customers with building a backlog, roadmap and utilize tools to facilitate the actual work that needs to be done.

CoE organizational principles
While some companies prefer to build a distributed organization with central leadership, we see more success with a dedicated and centralized organization operating as a CoE. Once the use of cognitive technologies has been established within the enterprise, these teams can become more distributed, but initially, we recommend a centralized model. There are two practical reasons for initial centralization:

  1. Flexibility: The initial interest and business case usually comes from a specific business unit. As the this first implementation becomes visible across the organization, other business units or functions are quick to build their own business cases. They want to draw on the emerging CoE skills rather than re-invent the wheel.
  2.  Impact: AI technologies have many transformational properties, however businesses rarely transform themselves from the inside. When carried out within the business unit itself, the tendency is to working on improving current processes rather than changing them entirely.

Don’t be tempted to classify the project as an IT program. Although implementing digital labor involves a significant amount of IT and application integration, these projects need to be recognized as business transformation initiatives. Making that distinction usually clarifies where the CoE should report.

Team Structure and Roles
Drive the CoE by establishing a small management team, including a cognitive lead who is responsible for managing the whole operation and members responsible for planning, project management, resource tracking and reporting. This team should also take responsibility for communicating to the broader organization as well as training and educating new team members. Some team also have data scientist, linguist and user-experience designers to augment the implementation teams with special insight.

Depending on the size of the transformation initiative, we also recommend setting up a steering committee. The steering committee reviews proposed changes to the roadmap along with implementation schedules, and provides strategic guidance and insight to the cognitive team.

Reporting into the cognitive lead will be three distinct teams, which are as follows:

  • The cognitive implementation team(s). This group is responsible both for configuring and building the knowledge base, in conjunction with the SMEs, and improving and maintaining it once the use cases have been deployed. In large transformation projects, we recommend a central team dedicated as a traditional sustaining engineering team for completed projects so the individual implementation teams can maintain velocity for new projects.
  • The cognitive architect team, which will work with the subject matter experts (SMEs) within the business to clarify use cases, determine business value and provide initial scoping of the effort required. This team will also own the technical architecture and be responsible for customer experience.
  • The cognitive dev ops or operational team, which has overall responsibility for deploying, maintaining, testing and upgrading the solution, including integration into the IT and business architectures. This team always has a dotted line to our internal dev ops team to coordinate changes and updates.

The cognitive implementation teams have their own project lead, architect and QA/tester, but the main work is performed by cognitive engineers for training and importing knowledge and Subject Matter Experts (SMEs) from the business for design and validation. The numbers might change over time, but as with any agile project, team consistency is important.

When staffing these teams, particular attention should be paid to the technical and business acumen of the individual members. As these kinds of transformation projects happen at the intersection of business process and technology, they require team members to understand both sides of the equation, to ensure efficient communication and rapid implementation. As we will discuss later, we recommend running the implementation teams using Agile methodologies, so experience with Agile is beneficial but not required.

Embracing agile methodologies
Utilizing Agile methodologies, such as Scrum and Kanban, is vital if the implementation of cognitive technologies is to be successful. Especially as many of these projects involve a significant amount of rethinking of existing processes and this is where more iterative approaches are beneficial. The projects must be able to handle the challenges posed by the fast-moving digital world and the CoE itself must be able to evolve in response to new opportunities and threats.

For each use case, or series of use cases, a dedicated implementation team should be assigned.

These teams will consist of:

  • One cognitive project lead to manage, track and report progress
  • One cognitive architect to ensure consistency with other projects, integrations etc.
  • Multiple (3) cognitive implementation engineers to teach Amelia
  •  Multiple (3) subject matter experts from the business to define and validate functionality
  • One QA/tester to create test cases

Depending on the scope of the use cases being developed, the team may be augmented with business architects, customer experience specialists, linguists, web developers and integration developers.

Structured Meetings
While the basics of Agile, such as two-week sprints and short daily stand-ups are working well, we also added more structured and pre-scheduled meetings to the sprints. We actually combine two sprints, so currently the optimal unit for completing an epic is four weeks. In addition to sprint planning sessions, sprint review and retrospectives, we have added several pre-scheduled alignment, validation and integration sessions to ensure we communicate and collaborate better with various business stakeholders and subject matter experts.

The meetings are scheduled throughout the four weeks and allow dedicated time for doing the actual work, but also specifies defect tracking and testing sessions so the teams quickly get into a cadence.

Some of the individual meetings have different participants and facilitators, and we have prepared initial agendas and templates for all the meetings so we can drive the appropriate discussion and resolution.

Templates and Sample Documents
Building your first cognitive solution can be a daunting task. You have gone through the training, and have a fairly good idea about teaching Amelia how to perform the roles. However, suddenly this technology needs to fit into a larger picture or solution.

We provide several templates and good examples for key documents, such as:

  • Functional and technical design specifications
  • Architecture and integration design specifications
  • Grammar and automation templates
  • Activity diagrams, conceptual models and process flows

The documents, together with the pre-scheduled meetings with agendas helps provide the structure for the teams to start running the sprints and delivering the prioritized epics and user stories.

Tools that support active collaboration and communication between team members and stakeholders are critical. While, we have our chosen tools, the actual tools are not as important as making sure that the team uses them consistently. The tools should be available to the extended team, i.e. business SMEs working with the implementation teams should have their own tasks and stories so they can be fully immersed in the implementation. There is a need for the following functions:

  • Agile planning tool to manage the development work and process. We internally use Jira, but any similar tool will suffice.
  • Project/task management solution to coordinate higher level deliverables with the other organizations. Internally, we use Clarizen, but any modern project management tool will suffice.
  • Document management and sharing. We currently use Clarizen for managing documents for the individual projects.
  • Social collaboration/team chat capability. We have used several tools, and currently Slack is the preferred solution.

As noted above some tools have multiple of these capabilities and could overlap. Our recommendation would be to investigate what is currently in use within your company for these functions.

Developing a Roadmap
Having a clear roadmap, showing which functions, services and processes will be initial targets for cognitive enablement, is key to establishing the size and composition of a CoE. Rather than a wish list or laundry list, the roadmap should be a prioritized set of use cases based on clear fact-based analysis. The data to drive the analysis comes from the business systems, including building small business cases attached to each major roadmap item.

A number of questions need to be answered as part of this analysis, including:

  • What is the volume of the activity and at what rate is it growing?
  • What is the cost per activity and what are the cost drivers?
  • How does this activity impact the quality or timeliness of the service we are delivering?
  • Are there technical or business restraints impacting the ability to deliver on specific use cases?
  • What development efforts are associated with the use cases?
  • What will be the organizational impact of implementing these use cases?

Obviously, looking at high value and/or volume processes is the first step, but many other aspects should be taken into consideration, especially if you can improve service or reduce the time end-users have to wait for resolution. It is key to find a healthy mix of cost savings and transformational projects that will show the versatility of the cognitive technology. On the other hand, completing entire functions or processes via digital labor is often more impactful, so prioritizing the backlog can be a formidable tasks with many conflicting opinions.

The roadmap should be updated at regular intervals to take into account any changes in the corporate environment and feedback on those use cases already being implemented. Determining the efforts, benefits and likelihood of success for each use case should allow them to be plotted into a visual roadmap.

Final recommendations
2017 is poised to be a year of technology driven disruption. AI and cognitive technologies have come of age and are being infused into existing businesses and processes as well as creating completely new opportunities. Furthermore, the technologies are continuing to evolve rapidly, late last year, AI beat humans in lip reading, scoring 93.4% accuracy versus 52.3% for humans. So, what was not possible last year, may be available the next. Having your finger on the pulse have never been more important.

Forget certainty, forget traditional business cases with year one payback, and forget how things have always been done with existing technology and processes. Things have changed, and things will change at an ever-increasing pace, and if you believe your business won’t be impacted by these technologies… well, then you are probably not reading this blog in the first place.

My advice is to get your feet wet. Whether you are looking into our cognitive technology (Amelia) or another technology provider, 2017 should be the year where your company establishes a Center of Excellence focused on evaluating and educating the rest of the company about these game-changing technologies.

Previous Next

The Intelligent Contact Center

Companies have spent decades implementing Interactive Voice Response (IVR) systems in their call and customer care centers, but they've proven unable to keep up with customers' expectations.

In this white paper, we examine the benefits of an Intelligent Contact Center, where companies utilize Conversational AI-powered virtual agents to provide first-line resolution and support for customers, and augment human employees through AI and automation.

Download our paper to learn the benefits of this approach and why current IVR systems simply will not cut it in today’s hyper-paced digital landscape.

Learn More