GenAI & Agentic AI: Insights from Industry Reports on Enterprise Adoption

Every major technological shift comes with a familiar pattern: a few companies leap ahead, while others lag behind, drowning in complexity and inertia. GenAI is no different.

Reports from Google, BCG, PWC, and WEF paint a clear picture of what is working and what we can expect in the days to come. Here are the major insights from these reports:

✅ Leverage GenAI effectively: Deploy AI internally, redesign core business functions, and launch AI-driven products/services.
✅ Three critical enablers: Data, domain expertise, and design.
✅ Biggest opportunities: Data and search.
✅ Shift focus: Drive top-line growth instead of obsessing over short-term productivity gains and ROI.
✅ Win with depth: Solve highly specific, niche problems with domain expertise.

But here’s the challenge: enterprises don’t move at the speed of technology—they move at the speed of their own internal processes, people, and decision-making.

I read through these reports and picked the best insights from these industry reports. Let’s get into it.

# Actionable Insights from Google

Of all the reports I read, Google’s report stood out. Maybe it’s because it wasn’t written by analysts in an ivory tower—instead, they interviewed both investors and operators, and even put their names on it. In other words, there’s real skin in the game.

If you read only one report, read the one from Google.

Some quotes:

The timeline for widespread enterprise adoption of Al will be slower than people think. A lot of last-mile issues need to be solved that aren't obvious until you're deep into them. - Dylan Fox, Founder and CEO, AssemblyAl

This comes from my own experience—both in building personal GenAI applications and leading GenAI initiatives at the enterprise level. In theory, developing GenAI seems like a cakewalk. But once you step into the real world, you uncover all sorts of messy, practical challenges. Most glowing reports? They’re usually written by those who haven’t actually built anything.

The fastest ROl in Al is in agents, but the biggest opportunity is in enterprise search. - Edo Liberty, Founder and CEO, Pinecone

Data and search have long been major bottlenecks in enterprises—whether it’s CRM, ERP, or e-commerce, they all struggle with search limitations. With LLM-powered analytics and search, we might finally break through these barriers. Expect to see a wave of new search interfaces—recommendations, personalization, and even conversational data exploration to surface the right insights when they matter most.

What are the advice for founders (and equally for CXOs):

# Focus on top-line growth

Arvind Jain:

Al is not just about efficiency. The bigger opportunity is about increasing your topline-doing things and building products you were never able to do before.

Most CXOs see GenAI as a productivity booster—freeing up top talent to focus on new products and services. In reality, that rarely happens. The people deeply embedded in existing workflows aren’t always the best at envisioning the future. Innovation often comes from those who operate at intersections, think unconventionally, and aren’t bound by how things have always been done.

As I discussed in How to Deliver Value in the Digital Age, new technology should drive new business models. GenAI will be no exception.

# Riches are in niches

Harrison Chase:

Vertical, narrow-focused agents basically replace human workflows, and the best way to build them is to think about how a human would do something and then build a combination of code and prompts to replicate that process.

Specific is terrific and it seems it is true in building agentic AI too.

Dylan Fox also said something similar:

Instead of aiming for general-purpose AI, focus on solving specific niche problems with a lot of depth.

# Don't build for the middle

Sarah Guo & Mike Vernal

Be aware of the different layers of the AI stack (foundation models, middleware, dev tools, and applications) and consider where your company fits. Be cautious about building a company in the middle layer, because it can be under pressure if the foundation models evolve rapidly.

Get closer to the customer (through UX and design) or focus on the powerhouse driving everything else (data).

That’s why I emphasize three key focus areas for companies:

• Data – Not just any data, but diverse, behavioral data that reveals what customers actually do—not just what they say.
• Design – How users interact with your product and the experience it creates.
• Domain – Deep expertise in a specific industry, like e-commerce, that gives you an edge.

# Customers are looking for solutions, not prompts

Matthieu Rouif:

Design your user experience in a way that helps people accomplish their goals without having to use prompts.

As Clayton Christensen put it, “Customers want a 4-inch hole, not a 4-inch drill.”

Most GenAI solutions today are just prompt-wrappers—take Cursor and Windsurf, for example, which are essentially VS Code with huge system-prompts under the hood. The real opportunity? Build hundreds of these specialized prompt-wrappers in your domain, leveraging strong design and industry expertise.

Then, let data guide you—see what resonates with customers and double down on what works.

# Companies don't move at the speed of technology— They move at the speed of organization

This line from Deloitte hit hard

No matter how quickly the technology advances or how hard the companies producing GenAl technology push organizational change in an enterprise can only happen so fast.

This has always been true for any major technology shift, but with the breakneck pace of GenAI, it feels even more pronounced. Running a business smoothly requires a completely different skill set than experimenting with the latest tech. Finding the right talent to explore new capabilities—and, more importantly, creating a culture that actually absorbs those changes—is something I hear enterprises struggle with constantly.

The more I speak with those pioneering GenAI use cases in enterprises (and the more survey reports I skim over coffee), the more I see the same theme emerging—one that Deloitte’s report underscores:

GenAI adoption is driven by internal demand, with early adopters seeking to use the tools to meet their specific needs

Translation? No matter how many GenAI vendors knock on your door with "game-changing" solutions, if your teams aren’t actively asking for it, adoption will be sluggish.

Deloitte defines agents as:

AI agents are software systems that can complete complex tasks and meet objectives with little or no human intervention. They are called “agents” because they have the agency to act independently, planning and executing actions to achieve a specified goal.

(Basically, experienced employees who never get grumpy, don’t take coffee breaks, and—best of all—never ask for a raise.)

Three types of GenAI solutions are emerging:

• Agentic solutions – Autonomous AI systems (think: a self-driven analyst).
• Co-pilots – AI assistants that enhance human decision-making and productivity (think: that 10x productive employee without the ego)
• API playgrounds – Infrastructure for companies to build their own AI-powered tools (think: a Lego set, but for AI).

The question isn’t whether GenAI will change the enterprise—it’s whether enterprises can change fast enough to keep up.

# Agentic Powered Automation

Corporates have always embraced technology to automate operations. Until now, most automation has been built on rigid, rules-based "if-this-then-that" logic. Every rule required a meticulously documented process that could be codified. The problem? The moment the process encounters a scenario that wasn’t pre-programmed, it falls apart faster than a budget projection at year-end.

Agentic automation, powered by LLMs, flips this model. Instead of defining every step, you describe the desired outcome, and LLMs orchestrate specialized agents to handle the tasks dynamically—kind of like hiring an intern who figures things out without needing a 200-page SOP.

PWC defines these as:

APA marries the capabilities of intelligent automation and the GenAI-powered agents. Within this framework, the agents act as the brain of the system, understanding user requests and planning a set of actions to fulfil the request.

# Use cases

The report lists many use cases inluding:

• Finance & accounting
• Supply chain
• Sales & marketing
• IT

A few use cases stood out:

• SEO automation – Analyze website content for keywords and performance, then recommend improvements to meta tags, backlinks, and structure.
• Automated search visibility reports – Track rankings, generate reports, and suggest SEO strategies to drive organic traffic.
• AI-powered customer support – Generate query responses for FAQs, trigger automated workflows, or escalate unresolved requests to a human.

The report calls for transparency and explainability. Given that we're dealing with probabilistic models, we may need to redefine both ‘transparency’ and ‘explainability’ to mean ‘best guesses wrapped in confidence", much like the reports these consultants write.

# Three value plays to maximize AI potential

BGC:

• Deploy Al in everyday tasks to realize 10% to 20% productivity potential
• Reshape critical functions for 30% to 50% enhancement in efficiency and effectiveness
• Invent new products and services to build long-term competitive advantage

The pattern is clear: Use GenAI to boost internal productivity first—then leverage those gains to innovate and create new products and services.

# Cross-functional AI adoption

WEF

BMW introduced a platform with multiple genAI agents across its sales, supply chain and marketing functions to accelerate the conversion of data into real-time insights. The platform intelligently chooses a data source specific to the function and then pulls information corresponding to the user’s prompt. This faster transformation of enterprise data into actionable knowledge has improved productivity across both the firm’s corporate functions and on its showroom floors by 30-40%.

Technology has always broken down silos, and GenAI—especially agentic AI—will take this further by redefining roles and responsibilities. Unlike humans, it works seamlessly across functions without ego.

As the saying goes, AI won’t replace you—but a generalist who works with GenAI without ego will. So, don’t cling too tightly to a rigid job description. Instead, learn how different functions operate and find ways to help everyone work smarter.

Oh, and one more thing—get comfortable with data. In the GenAI era, not understanding your company’s underlying data isn’t just a weakness; it’s career suicide.