Author: Charles Wright, Columbus.
AI curiosity is at an all-time high, similar to the failure rate of AI projects. I have at least a conversation every week with an organisation that wants to get started with AI but is unsure how do to so.
So I posed myself the question: “If I was a newly appointed C-suite executive with a fresh slate, some budget and a responsibility to integrate artificial intelligence (AI) into my organisation, how would I do it?.”
Initial Steps: Strategy and Responsible AI
It’s important when engaging an organisation in a new technology to demonstrate visible value as quickly as possible. Therefore, you need to begin building AI solutions soon. However, doing it without a plan would be a disaster. I would start by executing a short AI strategy to produce a use case focused roadmap.
1. Identifying the Right Use Cases
This is the biggest game changer and the first point where organisations fall short. If you pick the wrong use cases; ones that aren’t feasible and don’t drive business value, they’ll never do well in production. This involves:
- Identifying AI Uses: An exercise run across the organisation starting by splitting the organisation into domains (HR, Finance…), interviewing middle leaders from each domain to identify opportunities and then detailing the use cases with process operators.
- Assessing Feasibility: It’s important to understand if the use case can be implemented successfully depending on factors like data availability and quality, but also with business related factors such as the amount of change or existing transformation projects impacting the same processes.
Developing a Value-Based AI Roadmap: This roadmap will prioritise AI use cases focused on delivering significant value with minimal complexity. However, transformational use cases that can fundamentally change business operations can take priority
2. Producing a Responsible AI Policy
In parallel with the strategy project, it’s crucial to develop a responsible AI policy. This forms the foundation of AI use within the organisation and needs to be set up to enable all your AI users to safely use the various technologies. The policy should:
- Define Ethical Guidelines: Establish clear ethical standards for AI implementation.
- Ensure Compliance: Align with regulatory requirements and industry standards.
- Promote Transparency: Foster a culture of transparency regarding AI decisions and processes.
- Secure Data Privacy: Establish robust data protection protocols to safeguard sensitive information used in AI systems.
- Encourage Accountability: Set up mechanisms to hold individuals and teams accountable for AI-driven decisions and outcomes.
Crafting the AI Roadmap
1. Don’t just focus on Generative AI!
“AI” or what people actually mean by the term “Generative AI”, receives a lot of focus across media sources, however is not a magic bullet to every challenge. The roadmap should include a mix of AI use cases, such as:
- Traditional Machine Learning: Often where most business value is derived. Organisations collect huge amounts of data in systems like their ERP and CRM. Here I would focus on use cases which drive financial benefits such as demand forecasting and customer churn. I would also include Computer Vision use cases (such as quality control) and also more traditional Natural Language Processing (Invoice Processing) here as well.
- Bespoke Generative AI: Tailored solutions which use generative AI can be used to automate creative processes and interactions. For example, a bespoke customer service chatbot can reduce customer service costs, while also answering questions faster and more effectively than a human.
- Generative AI tools: Utilising tools like CoPilot and ChatGPT to drive employee efficiency. It’s important to roll these tools out with training, usage guides and clear use cases which are adapted specifically to their roles. As a result I would do this in waves, focusing on specific domains and teams to start. Sales and marketing are often the best candidates as the creative nature of their work can utilise AI to streamline much of their processes.
2. Balancing Change Across Domains
People can only handle so much change, implementing multiple AI use cases within a single domain at once is a recipe for failure.
To avoid overwhelming any single group within the organisation, target different AI use cases across multiple domains, implementing no more than 2 or 3 at once to start. This effort can be scaled to work on more use cases in parallel once initial successes have been delivered and consistent processes are established.
Treating Use Cases as Products
Turn your roadmap into a product portfolio. Each AI use case should be treated as its own product, following a consistent development process and roll-out:
- User-Focused Design: Prioritise the end-user experience, remembering that theirs often multiple end-user profiles!
- Proof of Value: Demonstrate the tangible benefits of the use case towards the metric of business value you’re trying to drive.
- Production Implementation: If the Proof of Value is positive build the full AI product and deploy it.
- Managed Roll-Out: Push the product out to the relevant end-users with change management factored in.
Consistent metrics for success, such as Objectives and Key Results (OKRs), should be applied across all stages and products. These shouldn’t include just measures of value, but also of product adoption and other indirect drivers of success.
Engaging Expertise and Building Teams
Hiring for AI is hard and there’s a talent shortage; in order to get started and deliver quickly its likely you’ll need external support.
1. Short-Term Consultancy to Get Started
Engage an experienced consultancy to design and deliver the initial AI roadmap. They should do this under your guidance, it’s crucial to be the face of AI, actively leading the project and engaging across the organisation.
You should look for breadth and depth of expertise across all of the previously mentioned areas. They should also have implementation capabilities as they will be required to build the first products. They will need the ability to scale resource and switch out skillsets as dictated by the stage of the effort and products that are to be produced.
2. Building a Permanent Team
Once the roadmap is established, begin hiring permanent roles to maintain and develop AI products while dramatically lowering costs. Aim for a blend of onshore and offshore resources if possible to maximise return on investment (ROI).
Conclusion: Driving AI Success with Strategy and Responsibility
Implementing AI in an organisation is a transformative journey that requires a strategic approach, responsible practices, and strong leadership. By identifying high-value use cases and a roadmap, developing a responsible AI policy and treating AI projects as products, you can accelerate your journey and increase success rates dramatically.
Our Columbus AI approach contains everything you need to get started in AI.
Join me at Smart Factory Expo, where I’ll be hosting a speaking session on: Simplifying AI Adoption in Manufacturing: A Product-Based Approach, 6th June, Industrial Data & AI Theatre, 11:30am to 11:50am. Find out more here.
Charles Wright, Director of Data & AI, Columbus
Charles possesses first-hand knowledge leveraging AI within major industries, including Financial Services, Life Sciences, and Oil & Gas. His proficiency in modern cloud technologies and machine learning makes him ideal for guiding manufacturers towards AI maturity.
Contact him directly: [email protected]
For more articles like this, visit our Industrial Data & AI channel. [/breakout
