It's been three years since ChatGPT brought large language models to the mainstream. Most of us have experimented with generative AI by now, yet we remain in the early stages of adoption. During this time, I've observed how organizations create practical results with AI, and I've noticed a pattern that might help you think about your own initiatives.
Three Levels of AI Impact
Personal Productivity: Useful But Not StrategicLink to Personal Productivity: Useful But Not Strategic
At the first level, people use tools like Copilot, Claude, or ChatGPT to write documents, respond to emails, and organize their calendars. Sometimes this works smoothly, other times you find yourself in a rabbit hole of prompt engineering. While generally useful, this level isn't strategic. Personal productivity gains are real but rarely translate to strategic advantage. If Paula from HR saves three hours daily, she'll likely fill that time with more operational work rather than strategic thinking. Even if she did focus on strategy, organizational hierarchies mean her insights need to travel up through management layers to reach decision-making forums. Individual efficiency doesn't automatically create organizational capability or competitive advantage.
Business Process Improvement: Measurable Gains Through IterationLink to Business Process Improvement: Measurable Gains Through Iteration
The second level involves taking existing workflows and adding AI capabilities to improve them. This could be a small departmental process or a cross-functional workflow spanning multiple teams and organizational silos. You review the current state, then introduce automation, AI agents, or generative AI to enhance performance.
I've seen practical applications in B2B sales processes, such as target group creation based on ideal customer profiles or automated prospect research. Here you can identify measurable improvements in data quality and analysis, flow efficiency, speed of execution, and cost reduction. These gains are real and often quantifiable.
However, let's recognize this for what it is: continuous improvement, not innovation. This is the "AI on top" or bolt-on approach. There's nothing wrong with it, and the business benefits are tangible, but you're working within the constraints of existing processes rather than reimagining them.
Value Chain Innovation: Starting With a Blank SlateLink to Value Chain Innovation: Starting With a Blank Slate
The third level is where things get interesting. Instead of improving legacy processes, you start with what's possible using AI technologies. Every few decades, a transformational technology arrives that creates this kind of opportunity. The dot-com era commercialized HTTP and web technologies, scaling online payments and enabling companies like Amazon and Netflix to fundamentally redesign value chains.
Amazon bypassed traditional importers and brick-and-mortar retail entirely. Netflix first disrupted Blockbuster with DVD-by-mail, then later eliminated physical video rental stores entirely by introducing streaming in 2007, replacing physical distribution infrastructure with direct digital delivery to consumers.
We're looking at a similar scale of opportunity with AI technologies, including the adaptive capabilities of machine learning, generative AI, and agentic AI. Academic research is already showing how autonomous AI agents on both supply and demand sides can independently reshape value chains in online commerce. We're seeing AI-native companies like Lovable and Cursor emerge and grow rapidly.
We're still in the early days of seeing true value chain redesign. While AI-native companies show what's possible when you build from scratch, examples of established industries fundamentally reshaping their value chains are only beginning to emerge. Large enterprises have similar opportunities for innovation, and right now everyone has access to the same technological building blocks.
Domain Knowledge Matters More Than TechnologyLink to Domain Knowledge Matters More Than Technology
Regardless of which level you're working at, one principle holds constant: domain knowledge matters more than technical expertise. Applying AI in logistics is a logistics challenge, not a technology challenge. AI in sales is fundamentally a sales question. The technology is increasingly accessible and commoditized, but understanding where and how to apply it within your domain creates the real competitive advantage. This is where your expertise becomes leverage.
Moving ForwardLink to Moving Forward
Think about your current AI initiatives. Are they clustered at the personal productivity level? Are you systematically improving business processes? Or are you asking fundamental questions about how AI enables entirely new ways of creating value? Each level has its place, but strategic advantage comes from clarity about which game you're playing. The technology is available to everyone now. Your domain expertise and strategic choices will make the difference.
This framework should help you plan and design your AI initiatives for positive ROI. It connects to the AI adoption maturity model I've described earlier, providing a way to think about where you are and where you're headed.
I don't have all the answers, and significant questions remain open. What I do know is that the best learning comes through experimentation and testing.
What are you going to build?
