Shaping Winning AI Strategies: A Guide for Executives

Executive Summary

Artificial intelligence (AI) promises immense opportunities for efficiency gains, innovation, and revenue growth. However, most companies struggle to translate AI's potential into reality. Our research shows that only 16% of AI initiatives achieve scale and deliver material impact.

To buck this trend, executives must get the AI strategy right from the start. This requires clearly defining the intention and framing of AI, choosing the right implementation approaches, identifying high-potential use cases, securing buy-in across the organization, and managing change sensitively.

This article provides a playbook for executives to craft an AI strategy poised for success. We outline best practices across five key dimensions:

1. Set the intention and frame AI appropriately
2. Balance top-down and bottom-up implementation
3. Carefully select use cases with clear success metrics
4. Secure sponsorship and nurture internal champions
5. Manage change through communication and celebrating small wins

With the right vision and game plan, AI can transform how your company operates and serves customers. The recommendations in this guide will pave the way for your AI initiatives to achieve the coveted scale and business impact.

1. Setting the Intention and Framing AI

The first step is clearly articulating the intention and framing for AI in your company. How you position AI sets the foundation for securing buy-in, identifying use cases, and managing expectations.

We recommend focusing the AI narrative on efficiency, speed, and augmenting employee capabilities. Terms like "artificial intelligence" often create misperceptions about autonomy and job displacement. It is better to use language around intelligent workflows, better decisions through analytics, and elevating what employees can achieve.

The intention should align with your overall corporate strategy. Is AI intended to drive more efficiency in operations and cut costs? Or is the focus on empowering innovation and new data-driven offerings? Ensure alignment on AI's role amongst the leadership team early on.

2. Balancing Top-Down and Bottom-Up Implementation

With the framing set, next focus on implementation approaches. Successful AI strategies blend top-down and bottom-up efforts:

Top-down: Projects mandated from the C-suite, focused on high-value operational use cases. Governance frameworks established around risks, ethics, and technical standards.

Bottom-up: Grassroots initiatives where employees directly apply AI tools to enhance their workflows. The focus is on capability building through training and platform access.

Balance is key - overly top-down efforts often fail from lack of adoption, while pure bottom-up leads to fragmentation. We recommend a 70/30 mix of top-down and bottom-up initiatives.

3. Carefully Selecting Use Cases

The most critical lever for impact is choosing the right use cases. This requires aligning stakeholders on success metrics and risk considerations upfront:

• Financial impact: What is the direct cost reduction or revenue upside from solving this problem? Factor in any additional investments needed.

• Risk profile: What risks does the use case pose around ethics, biases, regulatory compliance, brand reputation, and explainability?

• Scalability: Can the solution expand across the organization? Will it require change management for broader rollout?

• Data readiness: Does the use case rely more on internal or external data? How complex is managing this data at scale?

Focus first on a few use cases that score highly on financials, are low risk, leverage strengths around data, and easily scale. Avoid spreading efforts too thin or pursuing "moonshots" with unclear returns.

Here is one framework that can be used to evaluate some of these competing characteristics.

VIABLE

V - Value

‍What is the potential financial value this use case could generate? Consider cost savings as well as revenue upside.

I - Impact

‍Beyond financials, what is the broader impact? Does it align to corporate priorities around customer experience, innovation, etc.?

A - Adoption

‍How easy will it be to drive adoption across the organization? Assess change management needs and complexity.

B - Barriers

‍What are the biggest barriers to success? Consider biases, technical limitations, data readiness, regulatory issues.

L - Learnings

‍What learnings can be extracted even if the use case fails? Can insights about our data, operations, or customer needs be reused elsewhere?

E - Explainability

‍Can the AI system explain its working and recommendations? What level of interpretability is needed?

By assessing these VIABLE dimensions upfront, executives can quickly prioritize high-potential AI use cases and pivot or tweak those less likely to deliver value. This acronym provides a simple but effective evaluation framework.

Worked Example: Augmenting customer service with Generative AI

Here is an example of evaluating a generative AI use case for customer interactions in an investment management context, using the VIABLE framework:

Value:

• Reduces call center costs by 20% through automating common queries
• Drives 10% increase cross-sell and upsell revenue

Impact:

• Improves customer experience with 24/7 access to updates
• Aligned to firm's strategy of leveraging AI for growth

Adoption:

• Frontline employees likely to resist over job security concerns
• Training on effective handoffs from AI chatbot will be key

Barriers:

• Regulatory requirements around suitability apply here
• Must ensure financial advice accuracy or have defined human handoffs

Learnings:

• Rapid prototyping to uncover what content / personalized insights customers find most useful
• Apply text mining to identify additional cross-sell opportunities

Explainability:

• Black box models won't suffice given high-risk guidance
• Globally accepted principles for responsible AI also necessitate explainability here

So while high strategic value opportunity, the use case needs to address crucial regulatory, ethics, and adoption barriers through governance frameworks and change management. The overall risk-reward merits an initial controlled pilot.

Quantifying the value of use cases

Here is a framework for evaluating the economic value of an AI project.

The potential value is:

Direct value from time savings

• Additional revenue from faster outcomes
• Premiums charged for enhanced quality
• Sales conversion uplift from automation
• Agility and flexibility value
• Upside from employee time savings

Development costs

• Change management and productivity dip costs
• Wrong outcomes and legal/regulatory costs
• Costs associated with marginal error rates(vs. human baseline)
• Long term brand reputation impact

To elaborate:

The potential value considers both top line upside as well as cost savings. This includes quantifying direct time savings and faster outcomes, pricing power from performance lifts, sales improvements, strategic options value from agility, and freeing up employee time for judgement-intensive work.

Balancing this are the costs associated with building AI systems - both direct development expenses and indirect productivity dips from change management. Further risks come from inaccurate AI predictions that have economic or brand losses. Comparing the marginal error rates to human baselines is crucial.

Summing up the positives and negatives provides a realistic perspective on the net value potential for the AI system. The framework works for quantifying both cost-savings operational use cases as well as analyzing product-focused AI initiatives.

Quantifying the value of customer service:

We return to the previous use case of a AI-powered customer interaction system use case, using the framework outlined earlier.

The potential value is:

Opportunity

• $2M in annual cost savings from 20% improvement in call center productivity
• $5M revenue upside from 10% increase in cross-selling success
• $3M from faster response times and 24/7 availability
• $1M from call center agents reassigned to specialized roles

Costs

• $500K upfront development costs
• $200K change management and training costs
• $150K per serious regulatory breach (@2% likelihood = $3K expected value)
• $80K costs from marginal errors vs. human agents
• Brand reputation impact valued at $1M

Net Potential Value =

$2M + $5M + $3M + $1M

• $500K - $200K - $3K - $80K - $1M

= $9.22M

With ~$9M in net value, the opportunity exceeds costs by an order of magnitude. Likelihood of minor regulatory issues and brand risks, but overall strong case to proceed with AI system.

Will focus pilot on easiest queries and integrate compliance filters to minimize policy breach chances. For scale up, targeted hiring in customer analytics to handle handoffs and continue personalization.

4. Securing Sponsorship and Internal Champions

The most brilliant AI strategies fail due to poor adoption. Winning executive sponsorship and cultivating internal champions is critical to driving change.

Identify a senior executive to sponsor the AI program and orient internal messaging around their priorities. Find use cases that matter to this sponsor to secure their commitment. Beyond them, nurture middle managers and frontline staff who get excited by AI tools. Leverage their expertise for proofs of concept and lean on them to evangelise AI initiatives amongst their peers.

5. Managing Change Through Communication and Celebrating Small Wins

People ultimately determine the success of any technology transformation - including AI. Change management is vital, specifically:

1. Transparent communication around AI objectives, timelines, intended impacts, and how jobs may evolve. This reduces uncertainty and fear.
2. Inclusive training programs on AI fundamentals, tools, and ethical practices open to all employees. This builds capabilities and trust.
3. Publicly celebrating small wins and pilot successes. This reinforces commitment and sparks desire to participate.
4. Incentives for experimentation and starting bottom-up initiatives. This encourages innovation.
5. Avenues for employee feedback through surveys and open forums. This surfaces adoption barriers early.

Embracing these change management best practices separates AI strategies that languish from those that flourish.

The Path Forward

AI holds immense promise, but realising its potential requires meticulous strategy crafting and flawless execution. We hope the playbook in this guide charts a path for your AI initiatives towards scaled impact and transformational outcomes.