Embracing the AI Future: Practical Steps for Executives

With AI advancements dominating headlines, many executives feel pressured to integrate AI swiftly to avoid being left behind. If I’ve taken anything away from previous tech waves, it’s not to jump in headfirst but instead to take a considered approach that creates long-term value whilst delivering short-term gains, controls costs, and avoids tech debt that you could be living with for years. Here’s how you can navigate the AI landscape effectively.

Is the Sky(net) Falling?

If you’ve been following some of the headlines from the Australian Financial Review AI Summit, you might feel like a failure if you haven’t already completed your AI transformation.?

• Adapt to an AI world or be replaced, executives warned - AFR
• 30pc of board execs to go in two years - AFR
• Big business braced for AI wave – and plenty of uncertainty - AFR

Visualising the AI Future

I recently attended an insightful conversation hosted by Concentrix and led by Gareth Sutton . One of the livelier topics was what AI will look like in 3-5 years. For me, the easiest way to visualise this future is the concept of virtual agent-to-agent experiences.

Book me a holiday

Today, this involves talking to friends, using Google, or asking ChatGPT for an itinerary. Then comes the joy of researching, booking, and organising all the details.

In the not-too-distant future, I imagine saying, “Hey, Harriet (my virtual agent), book a holiday for next Christmas.” My virtual agent would then work with your virtual agent to assess our budget, availability, and interests, find the best options, and book everything seamlessly.

Similarly, I could ask Harriet to find savings in our family budget, adjusting our insurance, mortgage, and discretionary spending to realise those savings.

The list goes on: Organising a family movie night, planning a birthday party, or setting fitness goals.?

Harriet’s going to be busy, and I’m going to be enjoying the high life!

The Risks of Rushing AI Integration

However, getting ahead of yourself can be risky, as Westpac’s $70m+ real-time decisioning flop demonstrates. According to this Mi3 report, the Westpac EMT backed a management decision?to accept a pitch from EY to build the bank a decisioning solution from scratch in a project code-named In the Moment.?The circa $70m project failed. Much of the work was written off in 2021, buried amongst almost $344m?in software impairment charges that hit the balance sheet that year. They attempted to implement a complex system without sufficient groundwork and adaptive learning processes, pushing them into chaos.

The Stacey Model for Managing Complexity

The Stacey Model is a valuable framework for understanding and managing complex systems. It helps determine the appropriate management actions based on the level of certainty and agreement.

• Simple Domain: High certainty and high agreement. Processes are straightforward and predictable. AI applications here include automating routine tasks where outcomes are well understood.
• Complicated Domain: High certainty but low agreement. Requires expert analysis and systems thinking. For AI, this could involve decision-support systems where data is clear, but strategic choices are diverse.
• Complex Domain: Low certainty and low agreement. Experimentation and adaptive learning are key. Westpac’s real-time decisioning system likely fell into this domain. The project's complexity and lack of outcome agreement led to a chaotic implementation without sufficient iterative learning and adaptation, resulting in a $70m flop (Mi3).
• Chaotic Domain: Low certainty and high agreement. Immediate action is necessary to stabilise the situation before learning can begin. AI applications here would involve crisis management and rapid response systems.

Lessons from the Westpac Case Study

Westpac’s failure with its ‘In the Moment’ real-time decisioning project highlights the risks of moving from an idea to implementation without sufficient groundwork. As other innovation failures have highlighted, it seems they lacked a clear understanding of their business and technology requirements and, critically, a pathway to unlock value earlier that would enable them to control costs and adapt and learn as things changed, pushing them into chaos.

A Balanced Approach to AI Integration

To avoid costly mistakes while embracing AI, follow a design-led, agile approach with good governance to develop the capability to deliver complex use cases:

Do Your Research

• Stay informed by talking to vendors, attending conferences, and networking with knowledgeable peers.
• Conduct thorough market analysis to understand trends and best practices.

Define Your Strategy and Roadmap

• Create a Bold but Achievable Vision: Set ambitious yet realistic goals that align with your long-term business objectives.? This is your BHAG (Big, Hairy, Audacious Goal).
• Establish Guiding Principles: Define clear principles to guide decision-making and ensure alignment with your vision.
• Define Immediate Use Cases: Start with use cases that offer clear, measurable benefits, such as enhancing customer experience through personalisation or automating workflows.??
• Identify Foundational Capability: Do you have the right people? What’s your Data Quality like? Do you have a Customer Data Platform?

Design a Composable, Extensible Architecture

• Avoid Putting All Your Eggs in One Basket: Be wary of monolithic tech stacks.?They often have huge integration costs and gaps in their ability to deliver on your vision.
• Ensure Optionality: Design flexible systems that can adapt to changing needs and technologies. Find your no-regrets decisions. You don’t want to buy everything upfront.
• Understand Your Vendors' Capabilities: You need good partners. Assess your current vendors and explore new partnerships to ensure you have the right tools and support.
• Assess Build vs. Buy Decisions: (Hint: Building your own GenAI capability is probably not a good idea.)

Ensure the Foundations are in Place

• Implement Strong Data Governance and Security: Ensure robust data management practices to protect and leverage your data.
• Establish Your Team: Identify those with the skills and will to advance your AI capability. Bring in experts to fill known gaps.
• Improve Data Quality: Invest in data cleansing and enrichment to enhance the accuracy and reliability of your data.
• Manage Ethical Considerations: Establish clear policies for ethical AI use, including data privacy and bias mitigation.
• Foster a Supportive Culture and Brand Identity: Engage your workforce and align your brand values with AI initiatives to build trust and support.

Get some runs on the board

• Secure Buy-in and Investment: Gain support from key stakeholders and allocate necessary resources to your AI projects.
• Deliver Initial Use Cases: Focus on delivering quick wins to demonstrate value and build momentum.? Pilots are a good thing.

Gain momentum

• Test & Learn: Continuously experiment, learn, and refine your AI strategies to stay ahead of the curve.
• Scale Up: Gradually expand your AI initiatives based on the success of initial projects.
• Develop In-House Capabilities: Build internal expertise through training and hiring to sustain and grow your AI efforts.

Horizon View

When we look at this in terms of horizons, it should look something like this:

Conclusion

Navigating the AI landscape requires a careful balance of innovation and risk management. By following these steps and understanding the complexity through the Stacey Model, you can harness AI's potential while mitigating its risks.?

Harriet, take a deep breath. It’s going to be okay.