The temptation to chase the "shiny new toy" in AI is costing companies millions. Instead of buying the latest tech, leading companies avoid costly missteps and achieve sustainable results by carefully crafting their AI strategy and plan. Discover from our Simon-Kucher Elevate experts how to turn your AI potential into long-term growth.
Let's talk about the AI demo that blew everyone away in your company's last AI workshop. You’ve seen it – the one with the seamless interface, the mind-bending reasoning – the promise to solve nearly all your most pressing problems. It’s captivating. It’s also the start of a massive challenge.
Because after the demo ends and the real implementation work begins, a harsh reality sets in. Research suggests that over 80% of AI initiatives fail, often due to a lack of clear strategy or foundational readiness (RAND Study). The initiatives get stuck in pilot mode or produce unreliable results – leaving a trail of wasted budget and disillusionment.
Why? Because too many companies are buying AI tech without a plan. They're chasing the "shiny new toy" instead of crafting a well-thought AI strategy that fits their purpose.
A five-step journey to real AI success
So what does a successful AI journey really look like? Based on our work with clients, it typically follows five clear steps:
- Start with a clear job-to-be-done – What is the AI actually solving?
- Build a strong data and technology foundation – Quality, connected data and scalable architecture.
- Test, pilot, and validate before scaling – Avoid rushing to deployment.
- Maintain oversight and governance – Human control and safe guardrails matter.
- Prepare your people and processes – AI success is ultimately a human transformation.
Let’s explore why so many companies get it wrong – and how to avoid the most common pitfalls.
The 'shiny new toy' syndrome is real (and why it's costing you)
There is great temptation to jump on the latest AI trend – but without structured planning, it’s a high-risk game.
Consider the healthcare AI model designed to diagnose COVID-19 from medical scans. Researchers discovered it was trained on a dataset where patient position – lying down or standing up – correlated with illness severity. The AI mistakenly learned to predict disease based on posture, not medical indicators, rendering it unfit for clinical use (MIT Technology Review). Similarly, a major law firm made headlines after using generative AI for legal research, citing entirely fake cases in a court filing– highlighting the risks of unverified AI outputs that can be extremely damaging for reputation (CIO.com).
These examples underscore a critical lesson: adopting AI technology without planning and careful examination of its fundamentals can lead to unreliable results and reputational damage.
Step 1: Define the job-to-be-done
A successful AI journey starts with a simple question: "What problem are we actually trying to solve?" Without a clear, measurable goal, even the most powerful technology is just a solution in search of a problem.
AT&T is a prime example of getting this right. They didn't just "implement AI." They targeted specific, high-cost business challenges: By adopting an AI-as-a-service platform from H2O.ai, AT&T achieved an 80% reduction in fraud and saved $17 million annually through predictive maintenance ($7 million) and technician route optimization ($10 million) (H2O.ai Case Study).
Most companies stop here – focusing on automating existing processes. But the most transformative AI strategies go a step further: imagining what AI could enable that wasn’t possible before. Think of AI not just as a faster, cheaper version of today’s workforce – but as a superhuman partner with perfect memory, constant uptime, and access to a lot more knowledge. In this mindset, you’re not just improving operations – you're redesigning value chains, customer experiences, or even your business model.
To gain a true competitive advantage:
- Identify pain points and opportunities: Focus on areas where AI can deliver high ROI - not just cost reduction or process automation, but also in unlocking new capabilities.
- Define ambitions and measurable goals: Use quantifiable indicators to track success, but allow room for AI to push boundaries beyond current expectations.
- Start small, but think big: Pilot projects can prove feasibility, but visionary thinking early on ensures you’re building towards long-term strategic innovation.
Step 2: Build a solid data and technology foundation
Even with a clear goal, your AI's performance depends entirely on the data you feed it. Think of an upscale engine: feeding it low-quality fuel will inevitably lead to poor performance, no matter how sophisticated the design. In AI, your data is your only fuel. High-quality, well-structured data is essential for every reliable AI model.
This is where most long-term AI strategies fall apart. In many companies, the best data is trapped in silos – an important spreadsheet on someone's laptop, a legacy database no one has touched in years, or a departmental app that only a few people have access to. But you can't build a powerful engine on disconnected, messy data. Learning how to effectively manage data isn't just a technical task – it's the foundational work required to make AI successful.
As an example, consider Alibaba’s AI-driven recommendation engine, AIRec. The engine processes data from active users to deliver personalized recommendations on Taobao – driving significant sales growth (InData Labs). In contrast, inappropriate and bad-quality data led to the slowdown of the IBM Watson project, which struggled with inconsistent medical data that resulted in unsafe recommendations (STAT).
Equally critical is a robust tech architecture. AT&T’s adoption of H2O.ai’s platform enabled rapid integration and deployment. Without such infrastructure, AI workloads can falter under computational demands or fail to integrate with existing systems.
In addition to technical readiness, companies must also ensure their AI systems and data practices align with regional laws and global standards. Neglecting compliance can create legal and reputational risks just as serious as poor data quality.
Key considerations for a strong foundation:
- Data quality: Ensure data is accurate, diverse, and well-curated
- Scalable architecture: Invest in platforms that support AI workloads and integration
- Regulatory compliance: Ensure your data handling and AI applications comply with evolving legislation in your jurisdiction (e.g., EU AI Act, GDPR, state-specific laws)
- Continuous monitoring: Regularly assess data and system performance to maintain reliability
- Test thoroughly: Pilot AI solutions to uncover flaws before rollout
- Verify results: Confirm that models work for real-world scenarios
- Measure impact: Use trials to understand actual business value
Step 3: Test, pilot, and validate before scaling
Even with solid data and goals, jumping straight into full implementation can backfire. Before deploying AI at scale, thorough testing is essential. Pilot solutions to identify flaws, validate assumptions, and assess performance in a controlled environment.
Testing is more than QA — it’s about uncovering edge cases, measuring real-world reliability, and learning how AI interacts with your existing processes.
Consider Stitch Fix, the online personal styling company. Rather than immediately automating customer recommendations across its entire platform, it initially ran controlled pilots to compare human-curated and AI-generated outfits. The results were tested with real customers — enabling iterative improvement and gaining trust across business and tech teams before broader rollout. The outcome: an AI-human hybrid model that increased customer retention, drove sales and operational efficiency (Harvard Business Review).
Make sure to:
- Test thoroughly: Pilot AI solutions to uncover flaws before rollout
- Verify results: Confirm that models work for real-world scenarios
- Measure impact: Use trials to understand actual business value
Step 4: Maintain oversight and AI governance
Governance ensures that AI augments human judgment rather than replacing it blindly. It's the safeguard that protects against bias, misuse, and overconfidence in machine decisions.
As AI systems become more autonomous and complex, embedding human oversight becomes critical – not only for safety and ethics, but also to ensure AI continues to align with your business and brand values.
Take Zillow’s “Zestimate” example. The company once leaned heavily on its proprietary home value algorithm to guide pricing decisions — including for its iBuying business. But when the model misfired during volatile housing market shifts, Zillow had to write down hundreds of millions in losses and shut down the iBuying unit altogether (CNN). The lesson: even high-performing AI models require continuous human review, especially when used in high-stakes financial or operational decisions.
Key elements of strong AI governance:
- Maintain human oversight: Use AI as a decision-support tool, not a decision-maker, especially in high-stakes situation
- Ensure transparency: Understand how models make recommendations or predictions
- Establish governance: Define clear policies for usage, responsibility, and accountability
Step 5: Prepare your people and organization
Even the best strategy, data, and models can’t succeed if your people aren’t ready. AI adoption is ultimately a human challenge.
It requires new capabilities, new ways of working, and often a new mindset. Data scientists and engineers may design the models, but business teams must understand, use, and trust them. Change management is not a side activity – it is core to adoption.
Look at Lemonade, the insurance company: They used AI to personalize insurance quotes, support policy sign-up, facilitate payments, and automate claims processing (Lemonade Blog). Their staff didn't just disappear; they evolved from manual processors to AI overseers, armed with new skills and new workflows. Similarly, Carnegie Learning’s AI-driven personalized education tools required teachers to adapt their methods – using AI insights to tailor instruction (Carnegie Learning).
Success requires:
- Upskilling employees: Provide training to work alongside AI tools
- Process redesign: Adjust workflows to integrate AI effectively
- Cultural shift: Foster a mindset of collaboration between humans and AI
Three important levers: How Simon-Kucher Elevate can enhance and optimize your AI strategy
At Simon-Kucher Elevate, we help clients turn the five-step AI journey into real results. Our three-pillar approach safeguards your AI strategy against the most common and pressing problems, ensuring a clear path from investment to impact:
- Building a winning data strategy: Our experts join your teams to break down data silos and design a proper data management strategy that turns scattered information into a powerful, unified asset – from sophisticated dashboards to a whole data warehouse. We achieve the critical milestones towards enabling responsible AI – with the right data governance
- Designing your architecture: If legacy systems are holding you back, we help you re-architect your technical foundation, providing a clear technology roadmap for the right AI systems and data architecture needed for advanced applications like contextual marketing
- Ensuring your models deliver long-lasting value: As markets shift, algorithms need to be fine-tuned. We combine deep industry expertise with ML know-how to tailor AI use cases for optimum AI capabilities - making sure your machine learning initiatives continue to deliver long after launch
Contact Simon-Kucher Elevate to discuss your AI strategy
No other consultancy knows more about growth than we do. We live, breathe, and drive tangible growth, running over 1,500 projects every year.
Within our digital practice, Simon-Kucher Elevate, we have true digital practitioners who leverage data, technology, and design to drive digital growth and sales excellence helping your organization accelerate AI projects with a high impact on reduced time horizons, delivering a high ROI through customized and individual AI strategies. We have:
- Experience of over 500 digital strategy and use case implementation projects with positive EBITDA impact
- Combination of industry expertise and AI know-how to tailor use cases for business impact
- Database with a longlist of use cases and KPIs with categorization by industry relevance
- Translation of strategic business goals into data-related KPIs to monitor progress and the impact of your AI strategy implementation
Want to know more? Contact us today to learn how we can elevate your AI strategy. It's time to join the AI revolution!