Why an AI Center of Excellence is the Key to Success


The following is a guest post by Wilson Pang, Chief Technology Officer of Appen.

Becoming an AI-first Organization in Finance

Many global organizations are recognizing artificial intelligence (AI) as a core component of their business. In fact, three out of every four companies surveyed in The 2020 State of AI and Machine Learning Report consider AI critical to their success. This is no surprise: there has never been a more opportunistic time to invest in AI given the breadth of people, budget, and other resources available to devote to these efforts.

Financial services firms are likewise integrating AI into their businesses to enhance operational efficiencies, bolster customer experience, and obtain competitive advantages. With several AI projects already under their belt, many financial services providers have started asking, now what?

Invest in an AI Center of Excellence

Becoming an AI-first organization will be crucial to long-term success. Organizations with this goal should invest in an AI Center of Excellence (CoE) (and in truth, more than a third of large firms already have). A CoE is a team of experts in a given discipline that manage resources and provide counsel within that field. With an AI CoE, firms benefit from a growing body of knowledge and set of best practices that enable scalable AI initiatives to launch with proven success.

Think of an AI CoE as a core machine in your organization. This machine contains the accumulative learnings from past AI initiatives and a clear vision for use of AI in your business strategy. It enables teams to continuously deliver solutions consistent with your business needs. It can drive revenue, create cost efficiencies, enhance customer experience, and give you a competitive edge.

In financial services, an AI CoE can help establish data infrastructure to ensure projects launch successfully at scale and are leveraging high-quality training data to do so. An AI CoE will support the structuring of the right engineering team to deliver on the increasing volume, quality, and speed requirements for training data. Few financial services firms have developed an AI CoE, and as a result aren’t fully leveraging the latest best practices, putting at risk the success of their AI ventures.

How to Build an AI Center of Excellence

Building an AI CoE involves several key steps:

  1. Make the case for AI

Identify the business use cases for AI and how your organization will benefit from an AI initiative. Determine what kind of data you have, and what kind of data you’ll need. Establish the scope of your CoE.

  1. Obtain stakeholder buy-in

Building a CoE requires a team effort. Share your case for AI with relevant stakeholders across your organization, particularly your executive team. Survey results indicated that 80% of AI projects are being managed by VP level or higher.

Many organizations struggle with alignment between business leaders and technologists, particularly on data challenges, core problems, and budget allocation. Keep in mind that an alignment is instrumental in creating strong AI infrastructure.

  1. Build your CoE team and architecture

Consider which teams are critical to success and have domain expertise. You’ll likely require teams across product, product management, machine learning, data analytics, and DevOps (or its next evolution, AIOps).

DevOps deserves particular mention—these teams ensure everything runs smoothly within the company infrastructure and their support is required to launch a model and manage post-production delivery pipelines. Like DevOps, AIOps monitors whether the model is working as intended, but with the added leveraging of AI through machine learning and advanced analytics technologies.

  1. Build a flywheel to launch your AI initiatives

A flywheel is a self-reinforcing loop made up of best practices. Your CoE should act as a flywheel, a core machine that drives revenue. To build scalable practices and create initial momentum, start small with quick wins.

Identify success metrics for each initiative, which could include saving money and time, generating revenue, or improving efficiencies. These metrics will guide your launch process and determine the data you need.

Gather high-quality data—data that is clean, complete, and reliable—and have the ability to collect, store, and annotate it before developing your algorithm(s) that address a use case. Don’t overlook the importance of this step; training data is the foundation of AI, and a key indicator of a model’s success or failure.

Depending on your in-house resources, you’ll choose to build your AI model using one of the following options:

  • Pay for a vendor-produced model – cheap and fast, but limited use cases
  • Build a model in-house – more control and alignment with use cases, but most expensive and resource-intensive
  • Outsource model build – customizable and requires few in-house resources, but expensive

An AI CoE will serve as a well-oiled machine for repeatedly launching scalable AI initiatives that support your core business strategies. Most importantly, building an AI CoE will take you further down the path of becoming an AI-first company, a critical next step in developing a competitive edge in financial services.

Wilson Pang has been with Appen since November of 2018 and has more than nineteen years’ experience in software engineering and data science. Prior to joining Appen, Wilson held positions at CTrip, eBay, and IBM.

Photo by George Becker from Pexels