A Practical Guide for Businesses
In April 2026, two timely pieces cut through the AI hype cycle. Ethan Mollick, writing in The Economist, warned that “the IT department [is] where AI goes to die.” His core argument: AI is a profoundly odd, risky, and powerful technology, a next-word predictor that unexpectedly writes code, gives strategic advice, or shows empathy, and companies are killing its potential by trying to “de-weird” it. Traditional IT processes, risk-averse governance, standardized KPIs, and legacy systems force AI into the mold of conventional enterprise software, stifling experimentation and emergent value.
At the same time, the Financial Times published “Investors are betting on AI chaos. History suggests otherwise.” Author Richard Waters noted that markets are pricing in revolutionary disruption – new winners, old losers – but past technology revolutions (PCs, internet, cloud) show savvy incumbents often muddle through, adapt, and even thrive. The real story is rarely the total chaos investors crave.
Together, these perspectives paint a clear picture of the most likely future of AI in business: not a dystopian job apocalypse or unicorn-disrupting chaos, but a pragmatic, evolutionary integration. AI will augment human work, reshape workflows, and deliver real value, but only for organizations that treat it as the strange, probabilistic tool it is while building the right foundations. Incumbents with strong data capabilities will have the edge.
Why AI “Dies” in Traditional IT, and Why That’s the Wrong Approach
Mollick’s essay resonates because it diagnoses a widespread problem we see daily in enterprise deployments. AI isn’t deterministic software with predictable outputs. It’s generative, context-dependent, and often surprising. When companies hand it to IT teams focused on security, compliance, uptime, and cost control, the natural response is to:
- Wrap it in rigid approval processes
- Demand ROI projections before pilots
- Force it into existing tech stacks without workflow redesign
- Prioritize “safe” use cases over creative experimentation
The result? Pilots that never scale. According to recent analyses (including Deloitte’s 2026 State of AI in the Enterprise), while worker access to AI tools has surged, the number of organizations moving projects into full production remains modest. Barriers like data quality, skills gaps, infrastructure readiness, and risk management continue to stall progress.
HBR has similarly observed that many companies report “widespread AI usage but disappointing returns,” with adoption stalling at the integration stage. The problem isn’t execution – it’s treating AI like a new CRM module instead of a fundamentally new way of working.
History Shows Incumbents Can Win If They Adapt Smartly
The FT piece offers reassurance: AI won’t necessarily destroy every incumbent. Past waves of technology (from electricity to the internet) initially sparked predictions of massive disruption, yet established players who invested in complementary capabilities—new skills, processes, and organizational structures—came out stronger.
In 2026, the winners won’t be the pure-play AI startups alone. They will be enterprises that:
- Combine their domain expertise and proprietary data with AI’s capabilities
- Redesign workflows around human-AI collaboration (what some call “co-intelligence”)
- Scale from pilots to enterprise-wide agentic systems under proper guardrails
PwC’s 2026 AI Business Predictions and similar reports emphasize a “disciplined march to value”: top-down enterprise strategies, measurable business outcomes, and governance that doesn’t kill experimentation.
The Most Likely Future: Pragmatic, Data-Driven, and Agentic
By late 2026 and into 2027, we expect the following trajectory:
- From pilots to production at scale — Organizations doubling the share of AI projects in production, driven by agentic AI (autonomous agents that execute multi-step workflows).
- J-curve productivity — Initial flat or negative returns as companies rewire processes, followed by steep gains once complementary innovations (new roles, data pipelines, decision protocols) are in place.
- Governance catching up — Mature frameworks for agentic AI, data quality, and responsible use becoming table stakes. Shadow AI will decline as secure, enterprise-grade platforms mature.
- Incumbents leveraging data moats — Companies with clean, governed data and domain expertise will outperform pure AI-native players in regulated or complex industries.
This future is neither utopian revolution nor failure – it’s an evolutionary transformation, provided organizations avoid the “IT department trap.”
Optimal Use of AI in Businesses: Five Practical Principles
Drawing from Mollick, historical lessons, and 2026 enterprise reports (Deloitte, McKinsey, PwC), here’s how forward-thinking companies are winning:
- Embrace the weirdness—experiment deliberately
Give teams space to discover unexpected uses. Mollick advocates leadership that encourages crowdsourced experimentation and “labs” to scale promising ideas. Treat AI like a creative collaborator, not just an automation tool. - Build on rock-solid data foundations
Data quality and governance remain the #1 barrier cited across reports. Without trustworthy data pipelines, AI outputs are unreliable. This is where specialized partners excel – unifying siloed data, implementing real-time pipelines, and ensuring privacy/compliance. - Redesign workflows and roles around human-AI co-intelligence
Don’t automate jobs – augment them. Successful organizations are re-architecting processes so humans focus on judgment, creativity, and relationships while AI handles analysis, drafting, and routine execution. - Deploy secure, governed agentic AI
Autonomous agents are the next frontier, but they require bounded orchestration, threat modeling, and compliance-by-design. Enterprises need platforms that support multi-agent systems without introducing new risks. - Measure what matters – and iterate
Move beyond vanity metrics. Track business outcomes (revenue impact, cost savings, customer satisfaction) and accept that ROI may follow a J-curve.
How Datarella Helps Businesses Navigate This Future
At Datarella, we’ve spent years helping organizations move beyond AI hype and pilot purgatory. Our expertise in AI agent development and security, full-stack application modernization, Web3-enabled decentralized solutions, and privacy-preserving data architectures directly addresses the challenges outlined above.
Whether you need:
- Secure, production-ready autonomous agents
- Data platforms that make AI reliable and compliant
- Integration of AI into legacy systems without the usual friction
- Or decentralized approaches that enhance trust and data integrity
We combine deep technical capability with practical business understanding to help you safely and scalably embrace AI’s weirdness.
The future of AI in business isn’t about replacing your IT department or betting everything on chaos. It’s about evolving how your organization learns, decides, and creates value—by treating AI as the strange, powerful tool it is, while building the data, governance, and cultural foundations it demands.
Ready to move from pilots to production without letting AI “die in IT”? Let’s talk. Contact Datarella to explore how we can help you capture the real, pragmatic upside of AI in 2026 and beyond.