Three Keys to Successful AI Deployments

By Jeff Karmiol, IBM Product Marketing

July 27, 2018

As organizations begin to explore how Artificial Intelligence (AI) can transform their operations and improve customer experiences, the questions and concerns about how to successfully deploy emerging AI technologies can be considerable. To provide insights based on real-world AI initiatives, IBM worked with IDC to create the IDC Technology Spotlight “Accelerate and Operationalize AI Deployments Using AI-Optimized Infrastructure.”  Here are excerpts of that report.

“AI has grabbed the center stage of business intelligence, despite having been around for decades, due to the growing pervasiveness of data, the scalability of cloud computing, the availability of AI accelerators, and the sophistication of the AI Machine Learning (ML) and Deep Learning (DL) algorithms. IDC predicts that by 2019, 40% of digital transformation initiatives will use AI services; by 2021, 75% of commercial enterprise apps will use AI, over 90% of consumers will interact with customer support bots, and over 50% of new industrial robots will leverage AI.

While the power and promise of AI is exciting, deploying AI models and workloads is not easy. Despite all the buzz, most organizations are struggling through proof of concepts (POCs) and only a few have made it to full production.

AI Model and Workload Deployments: Challenges and Needs

In January 2018, IDC surveyed 405 IT and data professionals in the U.S. and Canada who had successfully completed an AI project, had budget control or influence, and were responsible for evaluating or architecting a platform to run AI workloads. The survey sought to determine how organizations use and manage AI-enabled technologies and to identify the infrastructure used for running cognitive/ML/AI workloads, the deployment location of the technology, and the associated challenges and needs. As shown in Figure 1, survey respondents identified their key AI deployment challenges as dealing with massive data volumes and associated quality and management issues.

Poor data quality has a direct correlation to biased and inaccurate model buildout. Ensuring data quality with large volumes of dynamic, diverse, and distributed data sets is a difficult task as it is hard for the developers to know, predict, and code for all the
appropriate checks and validations. View the full IDC report here.

Decision Criteria for AI Solutions

When respondents were asked by IDC for their top decision criteria in choosing an AI solution, they identified security, cost effectiveness, and operationalization (building, tuning, optimizing, training, deployment, and inferencing) of data models/intelligence, as shown in Figure 2.

Keys to Successful AI Deployments

With all these factors in mind, the essential keys to successful AI deployments are:

Data Scientist Productivity

Building, testing, optimizing, training, inferencing, and maintaining the accuracy of models is integral to AI workflow. The selection and installation of the open source frameworks and the initiation of the modeling processes can be a cumbersome affair, and it may take weeks or months to get things working. Building and optimizing models can require manually testing thousands of combinations of hyperparameters.

Training models may take weeks or months to complete in some use cases: for example, a healthcare organization took a year to build and train a medical model to detect an early-stage cancer.

Tools to automate manual model building, optimizing and training help data scientists to increase model accuracy in less time.

Optimized Infrastructure and Efficient Data Management

Both training and inferencing are compute-intensive and require high performance for fast execution. AI and DL require a new class of accelerated infrastructure primarily based on GPUs. For the linear math computations needed for training neural network models, a single system configured with GPUs is significantly more powerful than a cluster of non- accelerated systems.

Enterprise Readiness

When bringing emerging technologies and frameworks into an enterprise setting, it is critical to ensure enterprise readiness – security, reliability, support, and other criteria, as noted in Figure 2. Most of the AI/ML/DL frameworks, tool kits, and applications available do not implement security, relegating their use to disconnected experiments and lab implementations. An additional challenge in DIY-built systems is the difficulty in getting enterprise-grade support from multiple vendors.

Conclusion

To help organizations accelerate AI-driven business outcomes and overcome deployment obstacles, IDC offers the following guidance:

  • Focus on the business outcomes, keep the project timeline well defined, and prioritize projects with immediate revenue and cost impact.
  • Seek out software tools to simplify and automate data preparation and accelerate the iterative building, training, and deployment of AI models to drive improved business outcomes.
  • Look for dynamically adaptable, simple, flexible, secure, cost-efficient, and elastic infrastructure that can support high capacity along with high throughput and low latency for high performance training and inferencing experience.
  • Embrace intelligent infrastructure, leverage it for predictive analytics and valuable insights, then slowly phase in task automation once the trustworthiness and quality of data is established.”

View the full IDC report here.

Learn more about the IBM AI Infrastructure Reference Architecture that helps reduce the complexity of AI deployments, improve data science productivity and efficiency, and accelerate adoption of AI.

Return to Solution Channel Homepage

IBM Resources

Follow @IBMSystems

IBM Systems on Facebook

Do NOT follow this link or you will be banned from the site!
Share This