Five Steps to a Successful AI Strategy
While the potential opportunities for AI in an enterprise are vast and the technology more accessible than ever, focusing on the highest-impact projects that move business strategic goals forward is more important than ever.
The key to developing an effective AI strategy with low TCO and high ROI is to start with an AI strategy that’s business-aligned and considers organizational strategy and capabilities. The five steps detailed in this post provide a roadmap and help ensure IT and business leaders invest in AI initiatives that have high business impact, low risk, and fast time-to-market.
AI in the Enterprise
Marc Andreesen famously coined the phrase “software is eating the world” in 2011, and recently this phrase has been accurately restated as “data is eating the world”. In the last 10 years we’ve witnessed the introduction of inexpensive and scalable cloud data storage, cloud-based data lake platforms, and growth in streaming IoT sensor data.
Organizations have access to a broader range of data sets that provide more detailed information on all aspects of their operation than ever before. In a 2022 survey, IBM found that 35% of global enterprises have already adopted AI in their enterprise IT systems, while another 42% are considering adoption.
Source: IBM Global AI Adoption Index 2022
In this post we’ll discuss the five key steps to develop a strategy to leverage AI technologies that drive cost-efficiencies and competitive advantage for enterprise and mid-market.
Why your Company Needs an AI Strategy
While data analysis and visualization tools have advanced, developing actionable insights and deep understanding of increasingly diverse unstructured, streaming, and multi-media data sources has remained challenging. Often, valuable insights contained in diverse data repositories goes under-utilized.
Organizations are increasingly turning to AI and machine learning technologies to transform high volumes of structured and unstructured data into actionable insights that help drive cost efficiencies, increase the velocity of decision-making, and create competitive advantages.
Source: IBM Global AI Adoption Index 2022
Yet choosing the most valuable use cases for AI investment and selecting the most promising AI technologies leverage remains challenging.
The Five Steps to Developing an AI Strategy
AI technology is evolving quickly, making it imperative that organizations develop a process and plan for incorporating them into their overall data analytics and IT portfolio. We recommend following a systematic approach to developing an AI strategy.
1. Prioritize Business Goals
2. Identify AI Use Cases
3. Prioritize Use Cases
4. Assess Skills and Organizational Capabilities
5. Measure Outcomes
Step One – Prioritize Business Goals
As with any business investment, technology must serve to move forward the overall business strategy and advance organizational objectives. Understanding strategic business goals and priorities helps avoid falling into a state of “perpetual proof-of-concept”, and to gain the executive sponsorship needed to move from ideation to implementation.
Key questions to ask when identifying business goals AI can serve:
1. Are there existing data-driven business transformation planned or in process?
2. What employee and/or customer experience goals are important to the business?
3. What are the primary KPIs driving the business? Sales growth? Cost reduction? Customer Satisfaction? Product/Service Innovation?
4. What competitive pressures does the business need to effectively respond to?
Knowing which outcomes the business most values helps us seek AI use cases that provide the best leverage in advancing key organizational goals.
Step Two – Identify AI Use Cases
Once the strategic business goals are understood, we can survey which business processes and existing data projects can be enhanced using AI technologies. Again, asking the right questions can lead us toward the highest value project initiatives.
1. Does the business have streaming or IoT data that is under-utilized?
2. Are business users manually processing unstructured data—such as product reviews, customer service feedback or scanned documents?
3. Are there quality control processes that involve high levels of human labor inputs?
4. Do consumers of BI reports and dashboards often refer to external data sources (especially unstructured sources) to fully understand what’s driving the results they see in the BI systems?
Most likely there will be more ideas than resources to address them all, and we’ll need to prioritize ideas.
Step Three – Prioritize Use Cases
Having a list of ideas where AI can provide leverage to new or ongoing data-driven processes, we can prioritize potential initiatives to identify the best candidates for initial pilot or full implementation.
Key prioritization questions:
1. Does the use case align with a high-value organizational priority?
2. Does a high-quality data source exist that provides the basis for the AI implementation?
3. Can we envision which technology and development process would meet the needs of the use case?
4. Is there an executive sponsor and budget to make the project viable?
5. Can we develop a proof-of-concept to ensure the solution is viable to reduce cost before committing to a full implementation?
If the answer to all the prioritization questions is “Yes”, the candidate AI initiative should be retained for further investigation. However, if any question’s answer is “No”, we should consider exploring the candidate project further, or focus on other ideas that have a higher likelihood of success.
Step Four – Assess Skills and Organizational Capabilities
If a proposed AI initiative is viable and has a high correlation with business strategy, we still need to consider that any IT project is at its core a product of the people who deliver it. We need to have the right people with the right skills to be successful.
Artificial Intelligence is an advanced technology, but recent developments have made it more accessible to more IT practitioners. Questions to ask when evaluating staff available for design and implementation:
1. What skills are required -- Data Scientist, ML Engineer, AI Engineer?
2. Do internal teams or trusted services partners possess the necessary?
3. Do existing cloud or on-premises platforms support the type of AI solution envisioned?
4. Do our existing operations teams have the skills to support AI in production-- if not how can they acquire those skills?
Step Five – Measure Outcomes
When introducing new types of IT systems into an organization, it’s important that business leaders understand the value the investment has delivered, and how strategic business outcomes are moved forward through the new investment.
The key to successfully measuring the impact for a new AI project is spending time at the beginning planning how to measure success after deployment. Develop key metrics during project planning, and plan how to collect success baseline and post-implementation metrics at the start. For example:
1. For employee efficiency AI projects, have a plan to measure the time savings achieved by incorporating AI/ML models the data processing process.
2. For QA projects, compare baseline to post-implementation metrics. Were more defects caught by AI than human operators alone?
3. How many new data elements were added to a Business Intelligence system through AI/ML techniques? How are business users incorporating these elements into their decision-making process?
As always, collecting quantitative measurements is preferred when measuring outcomes—but qualitative feedback, surveys, and success stories are often equally valuable.
What can a Successful AI Strategy do for your organization?
Artificial Intelligence technology has moved beyond specialized and exclusive organizations and is now available as a feature within most tier-1 data platforms used by enterprise and mid-market companies.
Adding chatbots, using large language models, and leveraging computational AI in data analytics and line-of-business applications is now a few mouse clicks away, but as with any new technology having a strategy and measuring the strategic impact is crucial. The five-steps to developing an AI strategy detailed in this post help both enterprise and mid-market organizations assess the best candidates for investment, and measure outcomes to ensure continued success in AI programs.
Rob Kerr is VP, Artificial Intelligence at DesignMind.
Learn about our AI and Data Science solutions, including AI Strategy and Roadmaps, Machine Learning and AI Model Development, and Large Language Models (LLM).