Artificial Intelligence is becoming a key component of business transformation. Virtually, any business leader seeking to unlock value and develop new capabilities using technology is at some stage of the AI journey. For example, those at the leading edge have incorporated machine learning insights into business processes and are building functionality such as Natural Language Processing (NLP) and preventative maintenance diagnostics into their products. Others are experimenting with pilot projects or developing plans to get started.
In this edition of our newsletter, we briefly discuss six (6) pillars that every CTO must consider for building a comprehensive AI strategy for the firm.
AI Strategy – Some Key Considerations
1. Data Governance: Since machine learning algorithms are trained on the data provided, both structured and unstructured, to make predictions, the output accuracy greatly depends on the quality of data that is fed to the system. A comprehensive data governance forms the linchpin of a firm’s AI strategy that continues to deliver positive ROI in the long run. This includes, building of data lakes, data analytics, data security, ethical use, and other aspects of that may be considered important.
2. Pilot Project vs. Transformation: While long-term objective should always be to transform the business, firms must learn to take baby steps rather than take on large scale projects that come with risks that can be mitigated by taking on pilot projects and then applying lessons learned to bigger initiatives.
3. In-house Computing vs. Cloud: While firms use both on-premise equipment as well as Cloud, the latter approach has become wide spread. The reason for this is the ease and speed with which compute instances can be provisioned for pilot projects in the cloud that may take time to procure on premise, and then scaled up as work load climbs. However, in some case, consideration for special purpose hardware that may be difficult to acquire in the cloud as well as long-term variable costs (monthly expenses) may cause some firms to stick to on-premise computing or adopt a hybrid strategy.
4. Technology Choice: This includes both software and hardware. In terms of software, firms must decide between open source frameworks and packaged software, and also the choice of programming language (such as Python, C++, etc.). In terms of hardware, companies must also decide between general purpose CPUs and special purpose GPUs. As mentioned above, this may also include choice of Cloud provider (for example AWS vs. Azure) and the choice of high-levels APIs and other AI frameworks they provide.
5. Scalability: Many companies begin with a pilot AI project in one area, but soon realize that putting that into production or applying this across the organization is much more complex than they had envisioned. The more successful companies, on the other hand, never lose sight of AI at scale even in their pilot projects, which then allows them to apply solutions and reap benefits without having to re-architect the solution.
6. Team: Successful AI implementation typically involve a business leader who has both vision and clout to drive that solution. The firm must also decide what the makeup of the team would look like in terms of data scientist(s), analysts, and programmers, and whether it wants to build and train its internal team, join forces with an outside partner, or adopt a hybrid strategy.