Abstract
Developing AI solutions, including the training and deployment of ML/DL models, remains an important and often IT resource-intensive task. The integration of AI artifacts, such as ML/DL models and data engineering modules, into an existing enterprise IT infrastructure and application landscape represents an additional challenge. The productization or operationalization of AI and the inference of AI-based analytical insight into consuming applications are further explored in this chapter, where we focus on the productization and operationalization of AI specifically in an enterprise context. Furthermore, we shed light on the key challenges of operationalizing AI and describe essential goals for an efficient and sustainable productization of AI solutions, particularly ML and DL models and data engineering artifacts.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
See [1] for more information on the operationalization of AI deployments.
- 2.
A subset of the scope of AI operationalization as used in this book is often referred to as inference or inferencing.
- 3.
See [2] for more information on requirements and challenges of AI operationalization.
- 4.
See [2] for more information on configuring IBM Watson ML for z/OS scoring services in a CICS region.
- 5.
Please see Chapter 4, “AI Information Architecture,” where we have described ML model accuracy and precision in the context of the ML workflow.
- 6.
This sequence of challenges is not within a particular order.
- 7.
See [4] for more information on deploying AI, especially in regard to developing scorecards and performing comprehensive self-assessments.
- 8.
See [5] for more information on PMML, ONNX, and PFA.
- 9.
See [6] for more information on data and AI on IBM Z.
- 10.
See Chapter 4, “AI Information Architecture,” where we have described some of these vendor offerings.
- 11.
See [7] for a brief description of the most prominent AI open source tools, libraries, and frameworks.
- 12.
This is related to numbers 1, 2, and 3 in Figure 6-2, key AI operationalization domains.
- 13.
This is related to number 4 in Figure 6-2, key AI operationalization domains.
- 14.
This is related to numbers 2 and 5 in Figure 6-2, key AI operationalization domains.
- 15.
This is related to numbers 6 and 7 in Figure 6-2, key AI operationalization domains.
- 16.
This is related to numbers 3 and 8 in Figure 6-2, key AI operationalization domains.
- 17.
See [8] for an example of AI operationalization of ML pipelines with PFA.
- 18.
See [9] for more information on Apache Spark ML pipelines.
- 19.
See [10] for an example of an integrated online scoring services.
- 20.
We would like to point out again that the term inference is often used for a subset of the AI operationalization scope as used in this book.
- 21.
Please refer to Chapter 8, “AI and Governance.”
- 22.
See Chapter 13, “Limitations of AI,” where we elaborate more on autonomous ML and DL.
- 23.
See Chapter 13, “Limitations of AI,” for more information on bias.
- 24.
See Chapter 13, “Limitations of AI,” where we elaborate more on the explainability of decisions.
- 25.
See [11] for an example of measuring outcome from AI against business KPIs.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 2020 Eberhard Hechler, Martin Oberhofer, Thomas Schaeck
About this chapter
Cite this chapter
Hechler, E., Oberhofer, M., Schaeck, T. (2020). The Operationalization of AI. In: Deploying AI in the Enterprise. Apress, Berkeley, CA. https://doi.org/10.1007/978-1-4842-6206-1_6
Download citation
DOI: https://doi.org/10.1007/978-1-4842-6206-1_6
Published:
Publisher Name: Apress, Berkeley, CA
Print ISBN: 978-1-4842-6205-4
Online ISBN: 978-1-4842-6206-1
eBook Packages: Professional and Applied ComputingApress Access BooksProfessional and Applied Computing (R0)