The AI world is buzzing about DeepSeek, the Chinese AI company making waves with its high-performance open-source LLMs. Industry leaders and policymakers are debating its implications - what it means for the geopolitical balance of AI, the future of open-source innovation, and whether it signals a fundamental shift in technological leadership from the U.S. to China.

But amid all the excitement, we’re missing the bigger picture. DeepSeek is just another LLM. The real issue isn’t which model is winning - it’s why only 4% of companies are generating substantial business value from AI.

DeepSeek’s rise - A major shift, but not a game-changer

There’s no denying that DeepSeek’s success is impressive. It signals:

  • A Geopolitical Shift: China is no longer just catching up - it’s positioning itself as an AI leader, challenging the U.S.’s dominance.
  • Competitive Performance: DeepSeek’s models rival those of OpenAI and Google, proving that cutting-edge AI can be built efficiently and cost-effectively.
  • Transparency in Development: Unlike many Western firms, DeepSeek is embracing open-source AI, fostering collaboration and accelerating innovation.
  • Cultural Dynamics: It challenges outdated perceptions of Chinese innovation, showcasing a unique approach to technological progress.
  • Export Control Limitations: Despite U.S. tech restrictions, DeepSeek’s rapid advancement underscores the limited impact of such controls on AI leadership.
  • Cost Efficiency: DeepSeek is building high-quality AI at a fraction of the cost, setting a new benchmark for efficiency.
  • Exemplary AI Research: Its breakthroughs reinforce China’s growing influence in fundamental AI research.

Yet, none of this fundamentally changes how AI delivers value in business. The reality is that most companies today are still struggling to implement AI at scale.

The real challenge - Only 4% of companies are getting it right

While the world obsesses over the latest LLM breakthroughs, most companies remain stuck in pilot mode. According to multiple studies:

  • Only 22% of companies have moved beyond proof-of-concept to extract some value from AI.
  • Just 4% are generating substantial business impact from their AI investments.

Why? Because instead of focusing on execution, companies are distracted by the next big AI model. Whether it’s DeepSeek, OpenAI, or Gemini, the core challenge isn’t the model - it’s how businesses are (or aren’t) integrating AI into their operations.

What the 4% are doing differently

At SparkWise Data & AI by Ducker Carlisle, we’ve worked with some of the world’s leading companies on AI adoption at scale. Based on the experiences of our co-founders from Google and Artefact, we’ve found that the companies truly winning with AI do three things right:

They start where AI has the greatest impact: Top-Line Growth

Most companies misallocate AI resources by starting with efficiency plays rather than revenue-generating opportunities. The biggest ROI? Sales, pricing, and customer engagement.

A Success Story for us to illustrate this point: A $1B aerospace company struggled with a growing volume of complex RFPs. Using AI, we helped them reallocate efforts to high-value opportunities and Boost efficiency and win rates with AI-powered proposals. The Result: $36M/year in additional revenue.

They align incentives to drive AI adoption

AI success isn’t just about tools - it’s about behaviour. The companies seeing real impact have structured incentives to drive AI adoption across their teams.

On one end of the spectrum: Klarna. AI adoption directly affects employees’ equity and cash compensation, creating an internal race to maximise AI efficiency.

On the other end: Our AI Champion Program at Ducker Carlisle. Instead of financial incentives, we created a career-growth pathway where employees:

  • Develop their own AI initiatives.
  • Receive training and mentorship.
  • Become AI leaders within the company.

The result? Widespread engagement and a culture of AI-first thinking.

They build on a strong data foundation

No AI transformation succeeds without data. Yet many companies rush to deploy AI before ensuring they have:

✅ Clean, structured, and accessible data.
✅ A unified data strategy that spans departments.
✅ Clear data governance and security protocols.

If you want to dive deeper, check out our article dedicated to data foundations—because without good data, even the best AI is useless.

Stop chasing the hype, start implementing AI

The excitement around DeepSeek is understandable, but let’s not lose focus. AI adoption isn’t about which model you use today—whether it’s DeepSeek, OpenAI, or Gemini. If you build the right foundation, you can switch between models in just a few clicks.

The real challenge isn’t about having the best AI model—it’s about embedding AI into your organisation in a way that drives measurable impact.

Fabien Cros is Chief Data & AI Officer at Ducker Carlisle & Founder of SparkWise Data & AI by Ducker Carlisle.