Generative AI in 2025: From Hype to Tangible ROI

2025 is shaping up to be a pivotal year for generative AI—a year where businesses will shift from experimentation and excitement to measurable impact. According to Sébastien Paquet, VP of Machine Learning at Coveo—an AI platform provider powering companies like Salesforce, SAP, United Airlines, and Zoom—organizations will prioritize generative AI solutions that deliver clear business value, seamlessly integrate into workflows, and empower both customers and employees.

The "Show Me the Value" Era: Measuring AI’s True Impact

Businesses are entering what Paquet calls the "Show Me the Value" era, a period where organizations demand measurable results from generative AI investments. As the dust settles on months of hype and experimentation, the focus will shift to practical applications that solve real problems and provide a clear return on investment (ROI).

“There were so many prototypes, and they are not cheap,” Paquet explained. “At scale, the costs go high very fast. Companies need to see clear value—whether it’s saving money or increasing revenue—before continuing to invest.”

To avoid implementing AI for technology’s sake, businesses must evaluate ROI based on key metrics such as productivity gains, operational efficiency, and revenue impact. Practical applications—like AI-powered customer self-service tools that reduce support tickets or internal copilots that streamline workflows—are prime examples of where value is already being demonstrated.

Key Question: How can businesses measure the impact of AI effectively? Paquet advises leaders to prioritize initiatives where outcomes—like reduced case submission rates, increased customer satisfaction, or time savings—can be quantified and tied back to the bottom line.

Unified AI Platforms: Breaking Down Silos

Another key prediction for 2025 is the rise of unified AI platforms. As companies race to implement generative AI, many have unintentionally built siloed AI initiatives across departments, leading to duplicated infrastructure, spiraling costs, and increased vulnerabilities.

“Many enterprises started with demos or proof-of-concepts,” said Paquet. “But when you scale siloed initiatives, the costs increase, and you lose efficiency.” Maintaining fragmented AI systems also creates challenges for data security, with vulnerabilities replicated across disparate pipelines.

To address this, businesses are consolidating AI initiatives into unified platforms that provide a shared infrastructure for data retrieval, model deployment, and governance. These platforms allow organizations to scale AI efficiently while reducing redundancy and managing costs.

Paquet pointed to retrieval-augmented generation (RAG) as a critical component of unified AI systems. “RAG setups allow businesses to ground AI outputs in their enterprise data. The retrieval layer is common to most AI applications, so investing in it provides a strong foundation for future use cases.”

Key Risks of Siloed AI Initiatives:

1. Duplicated infrastructure and rising costs.

2. Increased security vulnerabilities across systems.

3. Slower time-to-market for AI-powered solutions.

Overcoming Barriers: Successful consolidation requires not just technology investment but also cultural alignment. Paquet emphasized that mature IT organizations are better positioned to enforce horizontal, cross-departmental AI platforms.

Right-Sized LLMs: Efficiency Over One-Size-Fits-All

While early AI adopters often defaulted to using large, general-purpose language models, Paquet predicts that 2025 will see a rise in "right-sized LLMs"—smaller, more efficient, and task-specific models. This trend reflects a growing emphasis on cost optimization and performance.

“At first, everyone wanted the biggest and best models,” said Paquet. “But the largest models are expensive, slower, and often overkill for simple tasks like summarizing documents or generating emails.” Instead, businesses will deploy smaller models tailored to specific tasks, reserving the most powerful models for complex reasoning or orchestration.

This multi-model approach ensures businesses strike the right balance between performance and cost. Open-source models, Paquet noted, are also becoming viable alternatives for less demanding use cases.

Practical Example: A company might use a smaller LLM to generate internal meeting summaries while relying on larger models for customer-facing applications requiring deep contextual reasoning.

Key Trade-offs:

• Cost: Smaller models are cheaper to run at scale.

• Speed: Lightweight models often provide faster response times.

• Capability: Larger models remain essential for tasks requiring advanced reasoning.

To succeed, businesses must adopt platforms that support multiple LLMs, ensuring flexibility across various use cases.

Workforce Transformation: AI as a Co-Pilot, Not a Replacement

AI’s integration into the workforce will be transformative but not threatening, according to Paquet. AI copilots—tools that assist employees with routine tasks—will bridge the productivity gap, allowing teams to focus on higher-value work.

“We’re seeing tremendous productivity gains,” Paquet explained. “For developers, AI copilots can generate documentation, build tests, and improve code comments without added effort. The result is better collaboration and faster workflows.”

In 2025, the skills most critical for employees will include the ability to:

1. Work alongside AI tools effectively.

2. Adapt to new workflows powered by AI.

3. Leverage AI insights for better decision-making.

However, Paquet cautioned that organizations must invest in upskilling programs to prepare employees for AI-enhanced roles. While the benefits of copilots are clear, businesses must also address potential downsides—such as over-reliance on AI or reduced critical thinking.

Explainable AI: Building Trust and Ensuring Accountability

As AI becomes integral to decision-making, businesses must prioritize transparency and trust. Paquet highlighted the importance of explainable AI—systems that provide visibility into how outputs are generated and why certain decisions are made.

“Transparency is key,” Paquet stated. “If businesses can cite sources and explain the logic behind AI outputs, trust increases.” This approach aligns with tools like retrieval-augmented generation (RAG), which grounds AI responses in factual enterprise data, reducing the risk of hallucinations.

Paquet believes that governance frameworks will play a crucial role in ensuring responsible AI use. Organizations must develop policies to:

1. Monitor and evaluate AI outputs for bias or inaccuracies.

2. Provide transparency to employees and customers.

3. Comply with ethical and regulatory guidelines.

“Without trust, adoption suffers,” Paquet said. “Building accountable, explainable AI systems will be critical for businesses in 2025.”

A Pivotal Year for AI Adoption

As generative AI matures, businesses will move beyond flashy prototypes to focus on value-driven implementation. By consolidating AI initiatives, right-sizing models, empowering the workforce, and building transparent systems, organizations can unlock AI’s full potential.

Paquet’s predictions provide a clear roadmap for enterprises navigating this transformative year:

1. Focus on measurable ROI with practical AI applications.

2. Eliminate silos with unified AI platforms.

3. Deploy right-sized LLMs for efficiency.

4. Equip employees with AI copilots to drive productivity.

5. Build trust through explainable and responsible AI.

For businesses, 2025 represents both a challenge and an opportunity—a chance to transition from experimentation to tangible results and ensure generative AI delivers lasting value.

🧑‍🚀 Not a subscriber yet? Join Forward Future for timely news, accessible education, and a thriving community. Click here.