Since the launch of ChatGPT and its peers in late 2022, generative artificial intelligence (GenAI) has captured both headlines and boardroom agendas. But as the novelty of broadly-capable chatbots wanes, we are entering what many analysts call the second wave, one focused not on flashy proofs-of-concept but on real-world business value, domain specificity and new operational paradigms.
In this article we analyse where GenAI goes next: what this second wave looks like, how enterprises should respond, which business models will dominate, and what emerging risks to watch. We draw on recent data, expert commentary and global cases to inform entrepreneurs and corporate leaders alike.
Defining the Second Wave of Generative AI
The first wave of GenAI was marked by rapid breakthroughs in large-language models (LLMs), diffusion models and image generation systems. These tools, typified by ChatGPT, captured public imagination by producing text, art, code and more. According to a recent piece from McKinsey & Company, this first phase is giving way to a second wave “focused on implementation and value creation.”
Another framework describes:
- Wave 1: General purpose models that generate new content from prompts.
- Wave 2: Models embedded into workflows, domain-specific fine-tuning, agentic/autonomous behaviour and tool orchestration.
For example, one blog describes second-wave activity as:
“The story is no longer about chatbots, it’s about transformation across entire value chains.” – Supertrends
Key hallmarks of this second wave include:
- Domain-specific models (e.g., legal, healthcare, finance) rather than general chatbots.
- Embedded AI inside operational workflows (not standalone UI tools).
- Agentic systems that can reason, plan, call APIs, use tools.
- Focus on measurable business outcomes: productivity, cost reduction, new services.
- Greater maturity: governance, trust, explainability, integration.
In short: the second wave is less about what generative AI can do in a demo, and more about what it delivers in the enterprise.
Where the Business Value Multiplies
Domain innovations
In this phase, companies are no longer asking “What can GenAI do?” but rather “How can GenAI improve my business processes?”
Consider the following examples:
- A legal-services firm fine-tunes a foundation model on internal contracts and past case-law, enabling rapid clause generation and risk-flagging.
- A pharma company uses generative AI to propose candidate compounds, simulate outcomes and surface high-value leads faster.
- A supply-chain logistics firm deploys agentic AI that monitors live data feeds, predicts disruption, and triggers adjustments automatically.
According to a startup-market analysis from NFX, the second wave is characterized by companies “not pitching AI, but using it inside their businesses to make them faster, cheaper and higher quality than competitors.”
New business models
Several model shifts are becoming evident:
- AI-as-workflow-engine: Deploying GenAI capabilities inside core processes (R&D, customer service, supply-chain).
- AI + tools ecosystem: Combining LLMs with tool-orchestration (APIs, knowledge graphs, vision models) to deliver autonomous agents. (See next section.)
- Vertical-native AI providers: Instead of horizontal chatbots, we will see specialist AI for diagnostics, compliance, design-automation.
- AI platform infrastructure: Companies that offer fine-tuning, model-serving, and orchestration for enterprise workloads will become important.
Measurable outcomes matter
The shift to second wave demands metrics. McKinsey emphasises that businesses must move from “experimentation” to “scale and value creation.” McKinsey & Company In practice this means:
- Defining KPIs (time saved, errors reduced, revenue uplift) before deployment.
- Ensuring proper change-management (employee training, process redesign).
- Embedding governance: model monitoring, data bias assessment, audit trails.
The Rise of Agentic and Autonomous AI
One of the most transformative aspects of wave 2 is the growth of what is called “agentic AI” systems that not only generate content but act, plan, and execute. According to NTT DATA:
“Intelligent software solutions (agents) that extend the capabilities of generative AI models by giving them the ability to ‘reason’ to solve business challenges.”
What this means practically
- Multi-agent architectures: multiple specialised agents coordinate to solve complex tasks (e.g., research → design → execution).
- Tool-integration: the model calls APIs (search, databases, CRM systems, workflow engines) rather than just respond with text.
- Memory and context: agents track conversations, history, goals over time.
- Autonomous execution: minimal human input required after goal-setting.
Business impact
For entrepreneurs and enterprises, agentic AI unlocks these opportunities:
- Seamless customer-journeys: bots that can research, author a proposal, arrange a meeting, and follow up.
- End-to-end process automation: from request to fulfilment with minimal human hand-off.
- Adaptive workflows: AI that re-plans, learns from outcomes, adjusts dynamically.
As one analyst put it: “The next frontier in Generative AI is autonomy.”
Global Implications & Strategic Considerations
Regulatory and trust-aspects
As GenAI moves into mission-critical roles, the stakes rise. For instance, public sentiment towards AI has shifted: a recent large-scale survey found that the proportion of Swiss respondents saying AI is “not acceptable at all” rose from 23 % to 30 % over one wave.
Also, national regulators are already acting. For example, Interim Measures for the Management of Generative AI Services in China introduced oversight of public-facing generative AI in 2023.
Opportunity in emerging markets
Companies in the Middle East, Africa and Asia that embed GenAI early into their industries (oil-&-gas, logistics, healthcare, finance) could leapfrog legacy players. The second wave is less about media coverage and more about operational advantage meaning regional players can move fast.
Strategic roadmap for business leaders
- Audit existing workflows: Where are the repetitive, high-volume, low-value tasks?
- Choose a pilot: Select a domain with clear business value, measurable KPIs and domain data available.
- Build capability: Assemble interdisciplinary teams (data scientists + domain experts + operations).
- Govern & scale: Establish guardrails (data-privacy, bias mitigation, human-in-the-loop), then plan for scaling once the pilot shows results.
- Adapt culture: Educate staff, redefine roles, GenAI won’t replace professionals but will augment them.
- Monitor value: Rigorous measurement of business impact to justify further investment.
Risks, Constraints and What to Watch
While wave 2 promises much, the path is not without obstacles. Key risk vectors include:
- Quality & training-data degradation: As more AI-generated content filters into the training pool, there is a risk of model degradation or “model echo-chamber” effects.
- Over-hyped pilots: Many organisations still struggle to move from POC to production; McKinsey found many GenAI projects fail to deliver measurable value.
- Regulatory backlash & trust issues: Rising negative sentiment and calls for human oversight may slow adoption. (See recent public-opinion survey.)
- Skills & integration challenge: Embedding AI into workflows often requires redesigning processes, re-skilling staff and governance capabilities.
- Ethical/intellectual-property challenges: Generative systems can inadvertently reproduce copyrighted material, bias or misinformation raising legal and reputational risks.
Entrepreneurs and leaders must therefore proceed with both ambition and caution: wave 2 is about impact, but the margin for mis-execution is higher.
Case Study – How a Global Logistics Firm Leveraged Wave 2
Consider a hypothetical scenario inspired by real-world practices:
A global logistics company handling freight forwarding, customs clearance and supply-chain management recognised that a large portion of its cost-base came from manual exception-handling, document-review and routing decisions.
Implementation steps:
- The company fine-tuned a foundation model with 5 years of customs and shipping data, enabling automated document classification and anomaly detection.
- It deployed an agentic-AI architecture: once an anomaly is detected, the AI triggers a tool to search regulatory databases, then generates an alternative route, books the shipment and notifies human staff for approval.
- KPI tracking: within six months, the exception-handling volume dropped 32 %, cost per shipment declined 14 %, and shipment throughput increased by 22 %.
- Governance: Daily reports flagged model decisions for human review until a trust-threshold was reached; bias and error logs were maintained.
- Scale-up: After success in one region, the model was rolled out globally, integrating local customs rules and languages.
Lessons learned:
- The value wasn’t in the “cool chatbot” but in changing how work is done.
- Domain data and operational context were necessary for success a generic chatbot wouldn’t have delivered.
- Governance and human-in-the-loop oversight were critical in gaining stakeholder trust.
- Measuring business outcomes early allowed management to build confidence and secure further investment.
Conclusion: What Comes After ChatGPT?
The question “what comes after ChatGPT” is no longer speculative, the answer lies in embedding generative AI into real-world value chains, making it operational, domain-specific, autonomous and measurable. The second wave of GenAI is about transformation, not hype.
Actionable takeaways:
- Companies should evaluate which workflows are ripe for GenAI embedding and pick pilot programmes with clear KPIs.
- Build interdisciplinary teams that merge domain expertise, data science and operations.
- Develop governance frameworks now, regulation and public sentiment are evolving fast.
- Prioritise measurement and scale only where value is proven.
- Look globally: regions and industries that adopt wave 2 early may gain competitive advantage.
Looking ahead, the third wave may bring interactive multimodal AI, autonomous ecosystems and possibly general-purpose autonomous agents but for now the key is pragmatic: turn generative AI from novelty to value.
