Generative AI Moves Into the Mainstream

Not since the internet boom has a technology been adopted by enterprises as rapidly as generative AI. Global spending on enterprise AI has surged from roughly $1.7 billion in 2023 to an estimated $37 billion in 2025, making AI one of the fastest-scaling software categories in history. Venture analysts count dozens of AI solutions that have quickly grown into major businesses, including at least ten with over $1 billion in annual revenue, from foundational model providers like OpenAI, Anthropic, and Google to industry-specific AI software in fields such as coding, sales, and customer support. This explosion was catalyzed by the public debut of tools like ChatGPT, which shattered adoption records by reaching over 100 million users within months and consumerized AI’s reputation for ease of use. Enterprise employees now expect similarly intuitive AI capabilities in their workplace applications.

Surveys confirm that AI is no longer a fringe experiment but a mainstream element of business. In the latest McKinsey global survey, 88% of companies said they use AI regularly in at least one department. Industry reports also show that generative AI is driving this trend: by 2024, three in four knowledge workers were already experimenting with genAI tools for writing, research, and productivity tasks. Major software vendors have quickly embedded generative AI copilot features into enterprise products, from Microsoft’s Office 365 and GitHub to Salesforce’s Einstein GPT, bringing AI assistance into daily workflows. According to Gartner, by the end of 2025 the vast majority of enterprise software applications will include built-in AI assistants, and by 2026 about 40% of enterprise applications will feature integrated AI agents, up from less than 5% in 2025. In short, AI capabilities are becoming a ubiquitous layer within the enterprise tech stack.

Yet for all the enthusiasm and investment, the enterprise AI boom has also seen its share of growing pains. Early reports of dramatic productivity gains abound. For example, AI customer-service assistants have boosted agent productivity by up to 30%, and generative writing tools can cut document drafting times in half. However, translating scattered wins into broad P&L impact has proven challenging. A recent MIT study found that only about 5% of enterprises have integrated AI tools into their workflows at scale, with most others stuck in experiment mode and seeing minimal structural change from AI deployment. Gartner similarly estimates that fewer than 20% of AI proofs-of-concept ever advance to scalable production deployments. This underscores a key point: simply adopting AI tools is not enough. To unlock transformative value, companies must knit these tools into an intelligence layer that fundamentally changes how work gets done.

From Isolated Tools to an Enterprise Intelligence Layer

In the early phases of adoption, many companies treated AI as a collection of point solutions, discrete tools tackling specific tasks in silos. A retailer might use a predictive model for inventory in one department and a chatbot for customer service in another, with little connection between them. This piecemeal approach yields local optimizations but misses the bigger prize. As Infosys research observes, the next stage of AI maturity is not to create additional isolated models, it is building a cohesive, integrated intelligence layer across the enterprise. In practice, this means AI is woven into the fabric of processes and systems, rather than bolted on as an afterthought.

When done well, AI ceases to be a set of standalone tools and instead acts as the digital nervous system of the organization. Data, predictions, and insights flow seamlessly across business functions, enabling systems to sense and respond in real time. For example, consider a traditionally fragmented process like order fulfillment: a new customer order in the CRM system would historically require human handoffs to check inventory in the ERP, then coordinate shipping. In an AI-enabled enterprise, an intelligence layer can automatically bridge these silos. The AI could cross-reference the order with inventory data, trigger a restock or delivery workflow, and even personalize a follow-up message to the customer, all without manual intervention. Multiple formerly disconnected steps become one continuous, intelligent workflow.

This enterprise intelligence layer is powered by a combination of generative AI and what are increasingly called agentic AI systems, AI agents able to perceive context, make decisions, and execute actions autonomously within set parameters. Crucially, the intelligence layer draws on knowledge from across the organization. Rather than operating with generic training data alone, the AI is enriched with enterprise-specific context: proprietary documents, transaction data, customer history, and domain policies. Companies are finding that foundation models, while powerful in general, need business-specific grounding to deliver relevant results. This has spurred approaches like retrieval-augmented generation, where AI models query internal knowledge bases, and fine-tuning AI on proprietary data. The goal is embedding intelligence where work happens, integrating AI deeply into core business processes rather than treating it as an add-on.

Early adopters who successfully build such an integrated AI core report significant benefits. In manufacturing, linking AI systems on the factory floor with supply chain and ERP systems can create a self-optimizing production line, automatically reordering parts and rescheduling production in response to a predicted machine failure. In retail, a unified AI layer might connect customer browsing data with inventory and marketing, so that pricing, promotions, and even product recommendations adapt continuously to real-time demand signals. In effect, the enterprise operates as a live learning system, what Infosys dubs a live enterprise that evolves and responds to changes dynamically. Organizations that reach this stage see AI move from being a helpful accessory to being a central driver of operations.

AI Agents and Autonomous Workflows

A defining feature of the new intelligence layer is the rise of AI agents that can carry out autonomous workflows. Unlike basic software automation or static algorithms, agentic AI systems leverage generative models and reasoning to handle complex, multi-step tasks with minimal human intervention. These agents can interact with applications, tools, and even other agents to accomplish goals. For instance, an AI sales assistant agent might automatically draft a customized proposal, send it to a client, book a follow-up meeting, log the interaction in the CRM, and update forecasting spreadsheets, all by orchestrating across various enterprise systems. Such scenarios are quickly moving from hypothetical to operational. McKinsey reports that 62% of companies are already at least experimenting with AI agents, and nearly one-quarter have begun scaling agent deployments in one or more functions.

Generative AI provides the brains behind these agents, allowing them to interpret natural language, generate content, and make inferences. But equally important is the integration of these AI agents with enterprise data and software. An agent is only as smart as the context it can access. That is why forward-looking organizations are investing in a vectorized knowledge infrastructure, transforming their vast stores of documents, records, and messages into AI-readable embeddings indexed for semantic search. With a solid data foundation in place, an AI agent asking a question or looking for information can retrieve trusted, relevant knowledge in milliseconds, rather than hallucinating an answer. Companies that thrive will treat search and knowledge retrieval not as a backend function, but as a core intelligence layer, one that turns data into decisions and insights into action.

The capabilities of AI agents are advancing rapidly. Today, many enterprise agents are essentially advanced assistants confined to single tasks, for example an AI that only handles invoice processing or only coordinates meeting scheduling. But the emergence of collaborative agent frameworks is already in view, where multiple specialized AI agents work in concert. Gartner predicts that by 2027, a third of enterprise AI implementations will involve groups of agents with different skills collaborating within applications. By 2028, networks of AI agents could dynamically cooperate across business functions, so that users achieve complex outcomes without needing to manually use dozens of separate apps. In other words, the vision is an autonomous enterprise workflow: a customer request or internal objective can trigger a swarm of interoperating AI agents that handle everything from data gathering to execution, supervised by humans at a high level.

Even in 2025, compelling examples of agentic AI in action are emerging. One global bank, for example, has deployed generative AI agents to automate over half of its IT service-desk requests, interpreting tickets and executing resolutions without human help in many cases. In healthcare, early agent systems assist clinicians by summarizing patient notes, pulling in relevant medical literature, and drafting documentation, saving significant time. And in a striking illustration of potential efficiency, participants in an Alvarez & Marsal executive roundtable noted cases where AI agents replaced entire teams of workers: at one business, 120 software developers were effectively replaced by 10 AI agents, accomplishing a coding workload that previously required an army of staff. Such anecdotes underscore how profoundly AI-driven automation can reshape operations. It must be noted that these are early instances and not yet the norm across industries. According to a recent McKinsey analysis, generative AI and advanced agents could technically automate 60% to 70% of employees’ typical work activities in sectors like banking and insurance, though realizing that at scale will depend on careful integration and change management. Even if only a fraction of that potential is captured in the next few years, the impact on productivity and enterprise economics will be transformational.

Rewiring Productivity and Decision-Making

By embedding AI as an intelligence layer, companies are beginning to redefine how work gets done and how decisions are made. One immediate area of impact is knowledge work productivity. Generative AI has proven adept at handling many rote cognitive tasks that consume employees’ time, drafting emails and reports, summarizing lengthy documents or meeting transcripts, populating forms, writing and debugging code, creating first drafts of marketing copy, and more. Surveys of early enterprise adopters show measurable efficiency gains. For example, customer service representatives assisted by AI chatbots saw productivity improve by 14% to 30% in field trials. Government agencies piloting generative AI report saving employees nearly half an hour per day on average through faster information retrieval and document generation. These kinds of micro-efficiencies, minutes shaved off here and there, add up significantly at scale across a large organization.

More importantly, the intelligence layer elevates the quality of work outputs and decisions. AI systems excel at sifting through large volumes of data to surface insights that humans might miss. In the enterprise context, an AI-augmented decision means that a manager is armed with comprehensive, up-to-date information and even recommendations when making a call. In finance functions, for instance, AI models can continuously monitor transactions and flag anomalies or opportunities in real time, enabling data-driven decisions rather than after-the-fact analysis. One CFO cited by Alvarez & Marsal noted success using an AI agent for sales training. The agent was trained on the company’s product knowledge base and was able to coach salespeople far more quickly and consistently than traditional methods. In supply chain management, AI forecasts can dynamically reroute shipments or adjust inventory levels in response to predictive analytics, optimizing costs and service levels beyond what manual planning achieved. Essentially, decisions that once relied on static reports and human intuition are increasingly informed by always-on AI analysis operating in the background.

The intelligence layer also fosters innovation by freeing humans from drudgery and enabling focus on higher-order problems. Early studies suggest that when knowledge workers delegate routine drafting or data processing to AI, they can spend more time on creative and strategic activities, brainstorming new products, engaging with clients, and solving complex exceptions. This shift aligns with the concept of the Age of With, where humans work with intelligent machines to achieve more than either could alone. Many executives see generative AI not as a replacement for human expertise, but as a multiplier of it: junior employees can perform at a higher level with AI support, and senior experts can tackle a broader scope when relieved of low-value tasks. The next generation of enterprise AI will be judged not by larger models or more impressive demos but by real-world results, by how it amplifies business outcomes. The ultimate promise of the intelligence layer is not just doing the same work faster, but enabling qualitatively better decisions and unlocking new opportunities, for example hyper-personalized customer experiences or entirely new data-driven services, that were previously out of reach.

It is worth noting that quantifying the impact of this AI layer remains a work in progress. Many benefits, such as improved decision quality or increased innovation capacity, are somewhat intangible or indirect in the short run. Traditional metrics might understate AI’s contribution. Saving 10 minutes on a task does not show up on quarterly financials, but it could cumulatively allow for a major strategic project to be completed faster. Early adopters caution that capturing the full value of AI often requires redesigning workflows and roles rather than just inserting AI into existing processes. Companies seeing the most value are those using AI to fundamentally re-engineer how work is done, not merely to speed up old workflows. This can involve retraining staff to work alongside AI, adjusting KPIs to recognize AI-driven outcomes, and fostering a culture that trusts AI-generated insights. Achieving breakthrough performance with AI requires moving from scattered initiatives and pilots to embedding AI directly into the fabric of operations. Organizations that manage this transition are already starting to pull ahead of competitors in efficiency, time-to-market, and adaptability.

Data, Trust, and Organizational Challenges

For the intelligence layer to function effectively, enterprises must get several fundamentals right. Data infrastructure is paramount. AI is only as good as the data it can learn from and act on. Most large organizations today struggle with fragmented, siloed data landscapes. Critical information lives in disparate systems, from CRM databases to file drives to email archives, often without a unified view or consistent quality standards. It is estimated that 80% to 90% of enterprise knowledge exists in unstructured formats, documents, PDFs, emails, chat logs, that historically were beyond the reach of traditional IT systems. Feeding this chaotic trove into AI models without preparation can be a garbage in, garbage out scenario. Indeed, in many early genAI projects, data quality and access proved to be the biggest bottlenecks. In a 2024 Deloitte survey, 62% of business leaders cited difficulties with data access and integration as their top obstacle to AI adoption. Enterprises are responding by investing in what Alvarez & Marsal calls a unified data foundation, modernizing their data architectures to break down silos and allow AI to draw from all relevant sources with speed and security. This often involves cloud-based data lakes or warehouses that consolidate structured and unstructured data, along with governance frameworks to ensure quality, lineage, and privacy.

Trust and governance are another critical layer of foundations. As AI becomes more central to decisions and customer interactions, boards and regulators are asking tough questions about explainability, bias, security, and compliance. Leaders need confidence that the AI’s recommendations are based on reliable logic and that using them will not create legal or ethical liabilities. This has led to an emphasis on AI governance programs: setting clear policies on data usage, establishing review processes for AI outputs, monitoring for unfair biases, and ensuring transparency in how models make decisions. For instance, in financial services, any AI that influences lending or trading decisions may need to show its workings to comply with regulations. Likewise, companies in Europe must consider data residency and the upcoming EU AI Act requirements for high-risk AI systems. Many organizations are implementing human in the loop controls for critical AI-driven processes, meaning that human experts validate AI outputs before actions are taken, at least until the AI has proven its reliability. As an executive quipped, AI has moved from the IT department to the C-suite as a concern. Issues of model ownership, data leakage, and intellectual property are now board-level discussions. Earning trust in the AI layer is both a technical challenge and a change-management exercise.

Organizational readiness is the third major hurdle. Deploying an intelligence layer is not just a tech upgrade. It is a transformation of how people work. Workforce adoption and skills have emerged as make-or-break factors in AI success. Employees need training not only to use new AI tools, but to interpret AI outputs critically and adjust workflows accordingly. Change management is essential, since some employees may resist or mistrust AI recommendations at first. Surveys by BCG and others have found that fewer than half of employees regularly use AI tools without targeted training and incentives, even when the tools are available. There is also the challenge of redesigning jobs and teams: as AI takes over repetitive tasks, companies must redefine roles to focus on higher-value activities or reskill staff for other positions. Leading firms treat AI deployment as an opportunity to upskill their workforce, teaching skills in data analysis, prompt engineering, and oversight of AI systems. In parallel, new roles are emerging, from AI model trainers to ethicists, to support the intelligence layer.

Finally, measuring the impact and managing the ROI of AI initiatives remains difficult but crucial. Many organizations lack frameworks to quantify AI benefits beyond obvious cost savings. As a result, some executives undervalue the gains, like improved decision speed or customer satisfaction, that the intelligence layer brings. To justify ongoing investment, forward-looking companies are developing AI performance metrics, for example tracking the percentage of decisions augmented by AI, the reduction in process cycle times, or revenue uplift from AI-driven product recommendations. A cultural shift may be required at the leadership level: adopting AI at scale often involves taking a portfolio view, accepting some experiments will fail, and a long-term horizon for returns, which can be at odds with quarterly financial pressures. In this sense, successful integration of the intelligence layer goes hand in hand with visionary leadership. Organizations where top executives actively champion AI, invest in foundational capabilities, and communicate a clear vision for AI’s role are far more likely to overcome the inevitable obstacles in this journey.

Intelligence at the Core of Business

After decades of promise, artificial intelligence is now assuming a central role in how businesses operate. Generative AI and autonomous agents are no longer confined to tech demos or innovation labs. They are becoming the core intelligence layer of the enterprise, analogous to how the IT systems of the past formed the transactional backbone. Companies that successfully build this intelligence layer are effectively creating a new corporate brain: one that never sleeps, continuously learns, and coordinates complex activity at scale. Early evidence suggests this can yield step-change improvements. High-performing organizations are using AI to achieve not just incremental cost savings but to drive growth and innovation, whether by launching AI-powered products, personalizing services in real time, or optimizing decisions with a sophistication impossible before. As Accenture’s Derek Rodriguez observed, the businesses that thrive in the AI era are those that treat AI not as a peripheral tool but as a core intelligence layer, one that turns data into decisions and insights into action.

Yet the rise of the intelligence layer is not an overnight revolution. It is a gradual rewiring of the enterprise’s nerves. Most companies are still in transition, moving from a scattering of AI experiments to a truly AI-enabled operating model. The coming years will likely determine the new competitive order, as firms that embrace AI deeply pull ahead of those stuck in pilot purgatory. Importantly, the intelligence layer does not replace human ingenuity or strategic acumen. It augments them. The most advanced enterprises envision a future where humans and AI systems collaborate fluidly, with routine tasks automated and human creativity focused on guiding the business. In this model, AI becomes an engine for a smarter, more agile organization: surfacing opportunities, mitigating risks, and executing repetitive work at digital speed, all under human direction.

In conclusion, generative AI’s new intelligence layer represents a fundamental shift in how value is created in business. It blurs the line between technology and operations, making AI an intrinsic part of processes rather than a separate tool. The potential benefits, from giant leaps in productivity to better decisions and new revenue streams, are immense. Realizing them requires more than technology deployment. It demands leadership vision, robust data foundations, employee enablement, and vigilant governance. The effort is akin to installing a new neural system in a legacy body. For those that succeed, the reward is an organization that thinks and acts with augmented intelligence at its core, effectively a smarter enterprise that can adapt and compete in ways that yesterday’s businesses could not. The intelligence layer is here, and it is poised to become the defining asset of the modern business era.

Sources, References, and Further Reading

• McKinsey & Company – “The state of AI in 2025: Agents, innovation, and transformation” (Nov 5, 2025). McKinsey Global Survey on AI adoption and impact. URL: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
• Menlo Ventures – “2025: The State of Generative AI in the Enterprise” (Dec 9, 2025). Market research by T. Tully et al. on enterprise AI adoption rates, spending, and trends. URL: https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/
• World Economic Forum – “Enterprise AI is at a tipping point, here’s what comes next” (U. Sachdev, Jul 31, 2025). Overview of emerging trends in enterprise AI, including agentic AI and integration needs. URL: https://www.weforum.org/stories/2025/07/enterprise-ai-tipping-point-what-comes-next/
• Gartner (Press Release) – “40% of Enterprise Apps Will Feature AI Agents by 2026” (Aug 26, 2025). Gartner predictions on the penetration of AI assistants and agents in enterprise software, and the evolution of agentic AI. URL: https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025
• Infosys Knowledge Institute – “Creating the live enterprise: The agentic AI imperative” (Oct 16, 2025). Research insight on moving from siloed AI deployments to a unified intelligence layer across enterprise operations. URL: https://www.infosys.com/iki/perspectives/creating-live-enterprise.html
• Alvarez & Marsal – “Where AI Meets Financial Impact” (Insights from CFO Roundtable, Dec 8, 2025). Discussion of AI adoption challenges and opportunities, citing an MIT study on low rates of AI integration and examples of agentic AI-driven efficiency. URL: https://www.alvarezandmarsal.com/thought-leadership/where-ai-meets-financial-impact
• Elastic & Accenture – Elastic Blog “Generative AI is driving the evolution of search within enterprises” (D. Rodriguez interview, 2025). Explains the importance of treating search and knowledge retrieval as a core intelligence layer for AI, with case studies. URL: https://www.elastic.co/blog/evolution-search-enterprises
• Deloitte AI Institute – “Benefits and limitations of Generative AI” (2023). Provides definitions and context for generative AI and its implications for enterprise use cases, data strategy, and the future of work. URL: https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/articles/generative-ai-for-enterprises.html