By Dawn Fitzgerald, the AI Executive Leadership Insider
(“AI Holistic Adoption for Manufacturing and Operations” is a four-part series which focuses on the executive leadership perspective including key execution topics required for the enterprise digital transformation journey and holistic adoption of AI for manufacturing and operations organizations. Planned topics include: Value, Program, Data and Ethics. Here we address the Program.)
The Executive Leadership Perspective
The first article of this series described the fundamental responsibility of executive leaders to focus the enterprise Digital Transformation and AI Adoption on Value. AI Holistic Adoption drives the value focus and is the combination of addressing the needs of the people, processes, and tools associated with the AI solution. This is a critical perspective shift from the comfort of data science to the applicability of true customer engagement.
Once the executive leader has truly anchored their vision of the enterprise’s digital transformation and AI adoption in this perspective, they must enable their teams by addressing the formation of their analytics program.
Analytics Program Deliverable
A key deliverable of the analytics program is the analytics library. It is important for the executive leader to recognize however that enterprise-wide AI Holistic Adoption requires more strategic elements from the analytics program beyond the basic “library of AI algorithms”.
To achieve AI Holistic Adoption, the analytics program must address the analytics library development, library portal, analytics design package, security elements and implementation governance.
Analytics Library Development
In AI Holistic Adoption, the analytics library is defined as a library of value analytics (VA). Each VA is associated with user stories and brings specific value. Each VA also has a design package which enables holistic adoption and enterprise level scaling. For most enterprises, the initial step in analytics library development is to review the existing analytics inventory for value, determine alignment with the value objectives of the enterprise, and identify value gaps.
The executive leader looking to rapidly scale their analytics program will drive their analytics library development to leverage internal teams, cross enterprise development teams and third party analytics developers. Evolving your analytics program to support multi-sourcing is a complex enterprise program endeavor.
Typically, the starting point for most companies is a highly customized set of dispersed analytics development efforts with pockets of individuals creating analytics. These analytics are known primarily by the originating developing team, “hard coded” in the offer, and are very difficult if not impossible to maintain, upgrade, or reuse. If third party analytics are used or analytics-as-a-service (AaaS) purchased, then the “black box” effect can occur where the company stakeholders do not have visibility into aspects of the analytics performance that are necessary to confirm value and ensure integration, quality, and reliability.
To scale analytics library development, the analytics program must deliver a suite of mechanisms to give visibility and control over the analytics. This is accomplished with the portal, design package, security elements and implementation governance.
The library portal is used to hold the value analytics library and is critical for successful global visibility, access and management of the analytics library. These are key capabilities for the scalable growth of an enterprise analytics program which may include multi-sourcing of analytics.
The library portal provides visibility and utilization of library content across the enterprise. For the digital transformation and connectivity goals of the enterprise, this is invaluable as it provides teams with the insight into existing cross enterprise analytics capabilities and corresponding access as required.
The Library Portal must also provide accessibility to the key design elements of the value analytics (VA). The set of design elements is referred to as the analytics design package.
Analytics Design Package
The analytics design package defines the use of the value analytics and the combination of design elements necessary to execute, reproduce, transport, and manage the VA. This package addresses the transportability, reproducibility and quality of the VA.
The analytics design package is a supercharged analytics delivery mechanism which provides development communities with both a standardized analytics design methodology and modularity required to scale, supporting an enterprise level analytics program. Although specifics vary by business, the table below provides an example summary of potential design package components, associated value and stakeholders.
Table: High level example of component, value and stakeholder
|Analytics Design Package Component||Value delivered||Stakeholder|
|User Story & results Benchmark||Mockup demo & results example||Data Scientists
|Algorithm Code||Date assessment results||Data Scientist|
|Baseline Data Set||Reference input data defines assumptions and criteria associated with input date||Data Scientists
|Training Model||Analytics Quality and Security||Data Scientists/Development|
|Simulation Engine||Simulation recreation||Development|
|User Value Configuration||Configuration specific to target customer or business segment||Development
|Test Suite||Analytics results against predefined quality criteria.||Data Scientists/Development|
When transporting a VA within the enterprise or bringing in as an AaaS from a third party, the components of the analytics design package set the design standards. These standards reduce development time, shorten time to market, provide modularity and improve operational efficiencies associated with integration and reuse across the enterprise.
The security of the analytics program has multiple facets. It needs to address basic system-level security, analytics design package security and stakeholder control access security.
System-level security should follow the necessary enterprise level network and application security requirements. In most cases, the analytics program can leverage established enterprise system security policies.
Depending on the use cases addressed, the security considerations for the analytics design package are critical for certain elements of the package. The analytics design package must carefully consider each element and typically have a minimum of high security for the training model and the results benchmark.
Finally, stakeholder control access security enables specific stakeholder roles to have different visibility and control points in the overall analytics. This is key for initial and evolutionary solution execution.
The implementation governance of the analytics program addresses multi-source implementation, data access and integration. For each of these the corresponding topics of consistency, quality, deployment and run mode maintenance must be considered.
The implementation governance specifies how third party or other enterprise organizations’ analytics will go through three phases of development for incorporation into the analytics solution of choice. The three phases are 1) VA identification & review 2) VA integration into development 3) VA deployment and maintenance.
Identification and review include use of the analytics design package elements for analysis of fit-for-purpose and a deeper review in a proof-of-concept, if needed. VA integration into the offer development requires either code incorporation into the offer or creation of transferred data integration points in the system (as with AaaS). The VA deployment process incorporates key needs of marketing, deployment planning and launch execution topics. Maintenance is the run mode plan for maintenance of the VA long term.
Through the VA integration phase of implementation governance and the data suite component of the analytics design package, the analytics program addresses key aspects of big data for overall AI Holistic Adoption. Data will be the next topic of the AI Holistic Adoption series.
Dawn Fitzgerald is Head of Innovation, Platforms and Analytics for Schneider Electric, where she is focused on the development of new digital services. She is experienced with how to make a range of AI technologies work for multi-site, global organizations. Learn more at LinkedIn.