Why The AI Economy And Your Company Needs Orchestrators And Builders

By Dean DeBiase

We are living through the most expensive tech adoption cycle in corporate history.

Gartner forecasts total global AI spending will hit $2.52 trillion by 2026, a 44% year-over-year increase. But, while every vendor has a new AI pitch, and hyperscalers and neoclouds report record revenues—not every boardroom has a realistic roadmap. In fact, inside most enterprises, who are touting their AI transformation story, a more alarming reality is playing out.

AI is not failing. It is stalling.

IDC projects enterprises will invest $632B in AI solutions by 2028. Yet despite that historic commitment, the vast majority of enterprises have yet to move AI beyond experimentation and into production at any meaningful scale. According to Gartner, 65% of CMOs expect AI to dramatically change their role within two years, yet only 32% say significant changes have been made to their skill set or operating model. The same pattern holds across functions and industries: investment is accelerating while execution lags. This lag is causing global leaders to ask more pointed questions like: Are we pouring trillions into a funnel with a leak at the bottom?

I don’t think the problem is just about the technology. The models are remarkable. The hyperscalers have built infrastructure that would have seemed like science fiction a decade ago. The problem is the gap that lies between what companies are buying and what they are actually able to deploy. I have spent years rebooting organizations and sectors, from online games, ecomm, and social media to cloud, mobile, AI, manufacturing and logistics companies—and the patterns I’ve experienced are repeatably consistent. The hype arrives first, as our industry makes up new brandable names for the new tech phenomena. The hard work of implementation comes later. And the companies that figure out implementation, not just procurement, are the ones that pull away from the field.

AI Infatuation, Experimentation, And Scale Phases

As I have discussed with top AI leaders on The Reboot Chronicles Show, we have migrated from the Infatuation phase where people rushed to try novel new gimmicky applications (like those hilarious animations your uncle or coworker sent you) into the Experimentation phase, where most enterprises may be overspending with returns that have not yet moved the needle. For insights into this, ask how much your company is spending on tokens, or check out you latest Anthropic or Open AI invoices. They are larger than you think.

We have yet to enter the next phase of Scale, with mass adoption and impact. As enterprises journey through Scale, on their way toward a Final Frontier phase, there will be a new era of prosperity that will ripple through global economies as the true ROI of tech-intelligence is unleashed across all sectors.

I recently met with Balazs Fejes, President and CEO of EPAM Systems, one of the world’s largest pure-play engineering and technology transformation firms. His perspective was refreshingly direct. In a market full of AI advisors, strategists, and platform sellers, Fejes had a simple answer when I asked him how EPAM thinks about its role: “We are builders. That’s all we are.”

That sentence says more about where enterprise AI value actually gets created than any analyst report I have read this year.

Who’s Actually Doing the Real Work: The Execution League

There is no shortage of self appointed AI advisors and consultants—with promises masquerading as builders now. A natural phenomenon during Adoption phases, is companies rushing in with clever ads—even airport signage—telling us we need to “partner” with this consultant or that platform to turn AI chaos into order. This fuels experimentation costs—but what’s scarce is execution. Some of my colleagues in Texas, call this the all hat—no cattle phase.

Fejes draws a hard line around what real implementation requires. No maintenance layers. No generic outsourcing. Just building and transforming systems that actually run. EPAM’s model seems to be built on a simple but increasingly rare premise: specialization compounds. Decades of pure engineering focus means they are not learning how to wire AI into enterprise complexity on their clients’ dime. Take that, Big 3 MBB. Companies in their Execution League have been operating inside that complexity for decades.

Balazs Fejes: Leading with Discipline in an Age of Disruption

Execution Leagues know the real challenge is not generating code. It is integrating AI into the systems enterprises already run: ERP platforms, CRM stacks, data infrastructure, and customer experience layers that were never designed to talk to each other, let alone to a large language model. I see this from the largest corporations on the planet to mid-market companies, like tech-enabled manufacturer, PBC Linear, where I am Chairman. The hyperscalers and neoclouds have built the data center foundations. But this layer, where systems connect and decisions get made, is where value is either realized or lost. And it is where companies like EPAM have chosen, deliberately, to plant flags on top of growth opportunity mountains.

Gold Rush Vendors Attack Every Tech Cycle

The race to own the AI implementation space is crowded and getting more so by the quarter. You have seen their ads and promotions. Every major technology services firm has repositioned itself around AI, and the catch-up strategies vary widely.

Accenture has pursued an aggressive acquisition strategy, raising its fiscal 2026 M&A target to $5 billion, betting that scale and breadth will win enterprise AI contracts. Cognizant has built a framework around what it calls the “AI velocity gap,” leaning on a partner ecosystem spanning Nvidia, Microsoft, and Google to bridge the divide between AI investment and actual returns. Capgemini is pushing further still into physical AI and human-robot collaboration, banking on the convergence of AI, data, and engineering maturity to open new markets. Smart way to play the long-game.

Each reflects a firm adapting an existing model to meet the AI moment, through acquisition, partnerships, or category expansion. What seems to set EPAM apart is a different kind of bet: that in an AI economy, decades of pure engineering depth, not assembled or acquired but built, will be what enterprises need most when the hard implementation work begins. As Fejes put it, “you cannot vibe-code an enterprise application.” Some will argue that no acquisition budget or partner ecosystem changes that fundamental truth. Perhaps.

The Growth Shift Leaders Miss: Optimizing Build vs. Buy vs. Borrow

When I am scaling large enterprises for global capabilities and growth, I start with what should the company do next: Build organically, Buy through acquisitions, and Borrow with dynamic partnerships with an immediate mandate to move the needle. Though many C-Suite leaders gravitate to one or two, my guidance is all the above.

For years, enterprise technology strategy revolved around a familiar calculus: buy a standardized platform, customize at the margins, and call it transformation. SAP for your ERP. Salesforce for your CRM. A handful of SaaS tools for everything else. AI is quietly breaking (some say destroying) that old model.

When the cost of generating a line of code drops dramatically, the economics of custom development shift with it. Bespoke solutions that were once cost-prohibitive become not just viable but preferable. Speed favors those who can build. Differentiation comes from what you create, not what you install off the shelf. This is proving out across large and mid-market enterprises, like PBC Linear, who created a proprietary industry-shaping configuration and automation system on Infor rails, that shrank complex order, supply chain, and manufacturing lead-times from weeks to hours—accelerating ecommerce revenue, while reducing company-wide headcount by 15%.

This shift is playing out on the marketing side just as visibly as on the engineering side. Empathy Lab, built specifically to help CMOs navigate AI implementation, is seeing it in real time. Their Growth OS framework replaces the sprawling, disconnected mar-tech stack most enterprises have accumulated with a unified architecture connecting media, content, audiences, and commerce into one decision system. What is happening in enterprise engineering is now happening in growth. The build era is back, and this time it has the economics to prove it.

CMO Reality Check: Expectations Accelerating Faster Than Execution

If you only read AI headlines, you would think ROI is already solved. It’s not.

The real conversation is not about cost savings. It is about growth. Most enterprises are not having that conversation yet because they are still stuck trying to get their underlying systems ready for AI in the first place. Fejes diagnosis is blunt on this point: “Garbage in, garbage out.” Many C-Suite leaders are learning that legacy infrastructure, fragmented data, incomplete cloud adoption, and years of accumulated tech debt do not disappear because a company bought a new AI platform. They become the ceiling that caps every initiative before it can scale.

This is not a new story. I have experienced the same pattern with cloud, ERP, and CRM. Each technology went through a infatuation and experimentation hype cycle before the hard work of integration produced real value. What is different now is the hyper-speed at which expectations are outrunning execution. In the marketing stack, expectations are also accelerating faster than execution—and the gap is uncomfortably measurable. CMOs are struggling to keep up with the rapid tech changes, while their teams try to stay current with a growing list of expectations around improving personal knowledge and marketing performance.

What Scale Phases Actually Look Like

The companies pulling ahead are not experimenting more. They are moving beyond that phase and implementing differently.

The pattern I’ve noticed, echoed by Fejes, is consistent across every enterprise that has gotten it right: start with real experimentation, build internal capability rather than just training, and embed AI across functions rather than in silos. He ran this experiment on his organization before advising a single client. Eighteen months of internal AI experimentation between 2023 and 2024, followed by embedding AI across every function including legal, HR, finance, compliance, and talent acquisition. As he put it: “There are no courses because the paradigms shift every six months.” This is not something you outsource to a partner or solve with a certification program. It is something you build, continuously, from the inside out.

On the growth side, Empathy translates that same discipline into a front-end execution model for CMOs. Their approach replaces isolated tools with connected systems, automation with orchestration, and pilots with production. The proof is in the results. Together with Mars, they used AI-driven synthetic audiences to cut research costs, accelerate innovation, and democratize concept testing across global teams.

I was previously the CEO of top tech research firm TNS Media, which was acquired by WPP, so this ROI peaked my interest. Yasmeen Cohen, Global Product Strategy Lead at Mars, was direct about what that meant in practice: “We have 75% accuracy in terms of the response rate from the human’s vs the synthetics, which gives us enough certainty to start expanding this capability and use it as an augmentation of our traditional research.”

Reckitt took a similar path, deploying 10 AI marketing tools within a single enterprise-grade solution and empowering more than 1,000 marketers across 15 markets. As Bastien Parizot, SVP IT and Digital at Reckitt, described it: “AI doesn’t scale through isolated use cases. To create real impact, each application has to operate as part of an integrated whole.”

What Orchestration Leaders Should Do Now

This is where the gap starts to close, but only for the leaders willing to make the shift.

For CEOs and CTOs I meet with, the first execution acceleration move is rebooting the metrics and reward systems. Stop measuring AI by spend or the number of pilots running. Start measuring what is actually in production and what revenue it is generating. The firms winning the implementation race are not the ones with the most tools. They are the ones who treated AI as infrastructure from the beginning and invested in the engineering depth to match.

For CMOs, the choice is starker than most want to admit. Automate or orchestrate. Automators chase efficiency, accumulate disconnected tools, and gradually cede control of their brand and demand generation to external agents they do not own. Orchestrators build connected systems where pricing, media, creative, and commerce reinforce each other and compound over time.

Fejes is direct about where the real prize is: “Cost saving has limited benefits. What you should be concentrating on is how can you drive better revenue streams going forward. How can you actually increase the revenues of the organization. That’s the really difficult part, because that is the actual hard part. It’s easy to take the costs out. It’s much more difficult to figure out how can I grow the company faster. And that’s the real discussion.”

AI is not slowing down. But for most companies, progress still is. The difference is not access to technology. It is the ability to implement it. The organizations that figure this out, with the right builders and orchestrators, the right systems, and the right operating model, will not just adopt AI. They will quickly depart the experimentation phase and choose paths that actually get scale and value from it. And in this market, that distinction is everything.