Seven Ways to Jump-Start AI - InformationWeek

InformationWeek is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

IoT
IoT
IT Leadership // IT Strategy
Commentary
4/18/2019
07:00 AM
Harsh Dhundia, Director, Pace Harmon
Harsh Dhundia, Director, Pace Harmon
Commentary
50%
50%

Seven Ways to Jump-Start AI

Artificial intelligence can offer competitive advantages, but it's not easy to drive success without understanding how to overcome the challenges.

Artificial intelligence will change the way we live, work and play. In the process, it will challenge existing businesses, create new ones and fundamentally shift the economy. There’s no shortage of reports on enterprises investing in AI, but there are relatively few details about lessons learned and what can be done to improve the design of those AI plans. Here are seven ways to overcome inherent technical challenges, organizational barriers and a lack of precedent for AI pilot programs:

1. Pick a high-impact business problem

Transformative AI initiatives with significant organizational and cultural barriers require sponsorship from the highest leadership levels  as they often take longer and are more complex. Start by addressing a high-impact problem that will not just get one-time executive sponsorship but ongoing executive engagement. This will also prevent the initiative from getting cut when the next budget crunch hits.

2. Understand feasible AI use cases

While headlines about the possibilities and consequences of AI abound, the reality is that current AI techniques still have several limitations. Consider an early feasibility assessment. Does the organization currently generate, store and analyze the types of data required by the machine-learning algorithm? Is the underlying process that will be improved already software-enabled? If the answer to both questions is no, chances are that even if machine learning offers potential benefits, the time and investment may not be worth the risk.

3. Know your use case and your data

Not all machine learning techniques are the same. A clear understanding of the use case and your data is imperative in picking the right machine-learning technique (or company/product). For instance, the algorithms required for targeted online advertising differ from those in healthcare. Similarly, if your use case requires extremely fast decision-making, such as applications that make real-time decisions, you must pick the appropriate algorithm. Additionally, if you’re thinking of purchasing and implementing an out-of-the-box AI capability, it’s imperative to understand what data the system was trained on and how that maps to your situation. For example, an AI capability trained on American healthcare studies may not be the best bet for a European healthcare use case.

4. Plan for IT infrastructure requirements

The IT infrastructure (network connectivity and computational power) required for machine learning applications as well as the infrastructure architecture are keys to AI success. IT infrastructure is also expensive, and once put in place, often can be hard to replace. For example, you may need to provide the ability to perform computations on a remote device (edge computing) versus centrally in the cloud (core computing). Think about what computations need to be performed and where to avoid nasty surprises later – both in terms of performance as well as the investment required for the infrastructure.

5. Assemble the right team

AI initiatives require cross-functional, multi-disciplinary teams to succeed. This includes team members who are developing the machine-learning system, members who understand the data, and those who act on the data and execute the process that will ultimately be improved. Change management and people-impact champions, project management and financial management requirements all need to be met. Compliance or enterprise risk-management team members are also critical to ensuring compliance risks are identified early in the process. Further, implementing a machine-learning system is not a one-time event. These systems need to be re-trained on an ongoing basis. Defining who is responsible for the ongoing training of the system at the very onset will help clarify organizational responsibilities and ensure that the ongoing costs associated with the training are kept in mind.

6. Ensure you have an appropriate contract

Unless you’re building an in-house team of machine-learning engineers, chances are that you are either purchasing a pre-packaged system or partnering with a vendor to develop one. However, AI is a statistical model that is closer to a service than subscription software. That means it is important to ensure that issues are appropriately addressed in your contract. For example, will your data be used to make the vendor’s product smarter? And will that smarter product then be sold to other customers? If yes, you’ll likely need to consider how to monetize your data as well as implications related to intellectual property. What are your criteria to accept the system as fully functioning? What is your recourse if the system doesn’t function as you’d expect it to? Who is responsible for ongoing training of the system? All these factors need to be addressed in the contract.

7. Be aware of brand and reputation risks and social impact

It is still early days for AI. Cautionary tales such as Microsoft’s AI chatbot tweeting offensive remarks should give organizations pause before introducing a public-facing AI. Additionally, recent research has highlighted the risks of biased data sets used for training and those of biases encoded (mostly inadvertently) into algorithms. The most cited example is that of facial recognition systems being error-prone when it comes to minorities since the training data set mostly contained images of light-skinned men. Countering algorithmic bias and its social impacts is now a big area of research. It’s important to consider what impact a bad recommendation/decision (or a series of skewed decisions resulting in a pattern) made by AI could have on your business.

AI has the potential to offer competitive advantages to businesses by offering previously unseen insights from data, increasing efficiency and supporting decision-making. It has become one of the top priorities for many organizations. However, AI initiatives face several challenges. Considering the above-discussed factors will increase their likelihood of success.

Harsh Dhundia
Harsh Dhundia

Harsh Dhundia is a Director at Pace Harmon with a background in delivering complex, large-scale business operations and IT transformations. 

 

The InformationWeek community brings together IT practitioners and industry experts with IT advice, education, and opinions. We strive to highlight technology executives and subject matter experts and use their knowledge and experiences to help our audience of IT ... View Full Bio
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.
Comment  | 
Print  | 
More Insights
Slideshows
Top-Paying U.S. Cities for Data Scientists and Data Analysts
Cynthia Harvey, Freelance Journalist, InformationWeek,  11/5/2019
Slideshows
10 Strategic Technology Trends for 2020
Jessica Davis, Senior Editor, Enterprise Apps,  11/1/2019
Commentary
Study Proposes 5 Primary Traits of Innovation Leaders
Joao-Pierre S. Ruth, Senior Writer,  11/8/2019
White Papers
Register for InformationWeek Newsletters
Video
Current Issue
Getting Started With Emerging Technologies
Looking to help your enterprise IT team ease the stress of putting new/emerging technologies such as AI, machine learning and IoT to work for their organizations? There are a few ways to get off on the right foot. In this report we share some expert advice on how to approach some of these seemingly daunting tech challenges.
Slideshows
Flash Poll