Building a successful AI strategy for your company – A holistic approach
July 21, 2023
Return to blog list
This year the floodgates have truly opened for Artificial Intelligence (AI) and Machine Learning (ML). Generative AI and Large Language Models have been a hot topic, and the release of ChatGPT showed the world just how accessible AI/ML programmes really are. Companies like OpenAI have allowed the public to use AI and experience this cutting-edge technology first hand.
With a rise in popularity, AI tools have become essential in today's business landscape, offering companies the opportunity to enhance products, streamline processes, and unlock new revenue streams. However, deploying AI/ML solutions requires a well-defined strategy that aligns with business goals and ensures successful implementation. Today, I wanted to explore five key pillars that should help guide a company's AI implementation plan:
- strategy,
- structure,
- lifecycle,
- data-centricity,
- and cross-functional collaboration.
1. Strategy (direction)
I am a firm believer in having to take a step back to understand what a company or team requires to succeed in their objectives. The following guidelines will help us achieve just that.
Firstly, identify and select the right use case – prioritise the most lucrative solutions that align with business goals and provide measurable benefits. This simple statement is easier said than done. It’s quite an art to be able to connect innovative perceptions to what is possible in your tech and data environments.
I have frequently found myself in a situation where the best idea for AI requires extensive investigation in terms of process, technology and data required for it to be practical.
There are several frameworks available that can assist you in ranking 'AI Initiatives' based on their complexity and potential business opportunities. To effectively utilise these, you'll need to consider several key factors that can influence the overall scoring of your AI initiative. I'm going to offer some insights on the factors that I believe play a significant role in providing a clearer understanding of opportunity versus possibility. Here are seven crucial factors that will guide you in achieving this goal:
- Strategic alignment to company vision
- Business impact
- Data accessibility and availability
- Algorithm/solution required – Prebuilt or custom and the accuracy required from AI predictions
- Change impact (process/system)
- Skills required – ask: “is it something we have done before, or completely new to the experience of the team?”
- Project duration estimate
Secondly, equip your teams with the appropriate tools and techniques - it's essential to empower your teams with the relevant technology, methodology, and training that will allow them to carry out AI/ML projects effectively.
Thirdly, guarantee that end users are able and willing to use AI - it's important to consider the usability and acceptance of AI/ML solutions by the end users. This is fundamental. There's no benefit in creating a technically flawless solution if it's not user-friendly for the intended audience.
Lastly, consider scalability and reusability - make sure to plan for scalability and reusability from the beginning of each use case. By crafting solutions with scalability in mind, we can enhance the value and lifespan of our AI/ML deployments. Moreover, there are certain components that can be reused from one AI proof-of-concept (POC) to another.
2. Structure
Each AI/ML project should be treated as a product or part of an enhanced offering to the market. To ensure clarity and success, the following aspects should be addressed:
- Problem Statement + ROI estimates: Clearly define the problem we aim to solve with AI/ML and estimate the return on investment (ROI) we expect to achieve.
- Change and Implementation requirements: Identify the changes and implementation steps necessary to integrate AI/ML into our existing processes and systems.
- Cost of delivery and accuracy required: Evaluate the costs associated with AI/ML deployment, including infrastructure, training, and ongoing maintenance. Determine the desired level of accuracy to meet user expectations.
- Stakeholder and product expectation: Engage stakeholders and product teams to align their expectations with the capabilities and limitations of AI/ML projects. Effective communication ensures a shared understanding of goals and outcomes.
3. Lifecycle
Adopting a structured lifecycle approach ensures consistency and adaptability throughout AI/ML projects. You should consider these phases:
- Ideation (business case creation): Develop a compelling business case that outlines the problem, proposed solution, expected benefits, and feasibility analysis.
- Experimentation (data science): Leverage data science techniques to explore and analyse potential AI/ML approaches. This phase allows us to refine our models and algorithms.
- Isolated POC (prototype): Build a proof-of-concept prototype to validate the viability of our proposed solution in a controlled environment.
- Integrated POC (technical integration): Integrate the prototype into our existing systems and assess its performance in a real-world context.
- MVP Production deployment (pilot + prod): Gradually roll out the Minimum Viable Product (MVP) in pilot programs, gather feedback, and make necessary improvements before full-scale production deployment.
4. Data-centric
To maximize the value of AI/ML, we need to adopt a data-centric approach that leverages existing AI capabilities and accelerates outcomes. Consider these when adopting your approach:
- Focus on data engineering: Prioritise data engineering to ensure high-quality, well-structured data that aligns with prebuilt AI models and existing frameworks.
- Deploy on SaaS platforms: Utilize Software-as-a-Service (SaaS) platforms to avoid upfront investments and accelerate delivery. This approach reduces the need for extensive infrastructure setup and maintenance.
- Improve human and AI coworking opportunities: Encourage collaboration between humans and AI systems, providing end consumers with user-friendly interfaces and empowering them to attain the benefits of AI in their daily requirements.
5. Cross-functional collaboration
Success in AI/ML deployment requires collaboration across different functional teams. Engage with the relevant stakeholders, such as:
- Business sponsor: Ensure strong sponsorship from the business side to align AI/ML projects with strategic objectives and provide the necessary resources.
- Tech and data: Collaborate with technology and data teams to build robust infrastructure, implement data governance practices, and ensure data security and compliance.
- Business SMEs: Involve subject matter experts from various domains to provide valuable insights, domain knowledge, and validate the AI/ML solution's effectiveness.
- Governance: Establish governance frameworks to monitor and manage AI/ML projects, including ethical considerations, privacy, and compliance with regulations.
Creating a successful AI strategy necessitates a comprehensive approach that considers clear direction, well-established structures, flexible lifecycles, data-centricity, and cross-functional collaboration. By aligning your AI/ML projects with these cornerstones, we can enhance our products, optimize processes, and foster innovation within your organisation. Viewing AI/ML as a strategic instrument position us for success in the constantly changing digital environment.
This guide was not written by ChatGPT nor an AI/ML application.
Citations
-
Tiaan Taljaard, Chief Data Officer at Ozow
