Best Practices

Unlocking Enterprise AI: Strategies, Agentic AI, and the Path to ROI

A comprehensive analysis based on Economist Impact research involving 715 technical executives and 385 data and AI technologists across the enterprise landscape.

KData Content Team KData Content Team
August 12, 2025
15 min read

Executive Summary

This article is a summary of an Economist Impact report, commissioned by Databricks, and combines a global survey of 715 technical executives and 385 data and artificial intelligence (AI) technologists who work across the fields of data engineering, data science and enterprise architecture.

Download Full Report

Access the complete Economist Impact research

Download Full Report
1

The AI Moment Has Arrived

AI adoption is no longer experimental—it's mainstream. The survey shows 85% of organizations already use generative AI (GenAI) in at least one business function, and 97% of enterprises over $10B in revenue are engaged. Yet, only one in five leaders believe their current investments are sufficient, signaling underinvestment risks.

85%

Organizations already using GenAI in business functions

97%

Large enterprises ($10B+ revenue) are AI-engaged

Action for Leaders:

  • Act with urgency—waiting risks falling behind.
  • Avoid "pilot purgatory": move from experimentation to production quickly, but with clear guardrails.
  • Differentiate by data—AI models are increasingly commoditized; proprietary data and business context are the real competitive edge.
2

Focus First on Productivity, Then on Revenue

AI's immediate benefits show up in productivity and efficiency—coding assistants, workflow automation, customer support, marketing content. Flo Health, for example, boosted data engineering efficiency by 24%, and Repsol saw productivity improvements of up to 30% among coders.

But leaders see long-term value in business model innovation, market positioning, and ESG alignment. Revenue growth hasn't been the main driver yet—but it will become decisive as adoption matures.

Flo Health

24%

Boost in data engineering efficiency

Repsol

30%

Productivity improvement among coders

Action for Leaders:

  • Track dual horizons: capture quick productivity gains while building toward revenue and business model transformation.
  • Measure ROI beyond cost savings—include innovation capacity, talent attraction, and customer experience.
  • Set realistic timeframes—3+ years to meaningful returns, particularly in regulated sectors like healthcare.
3

Adopt a Risk-Graded, Test-Bed Strategy

Executives recommend methodical, intentional pilots. Providence, for example, created governance and technical environments for controlled experiments. TD Bank restricted its GenAI chatbot to internal staff, protecting customers from unpredictable outputs.

Action for Leaders:

  • Prioritize internal, low-risk pilots first, then scale customer-facing applications.
  • Use sandboxes—HP runs 75 private sandboxes to test GenAI use cases securely.
  • Establish centralized oversight to avoid "AI chaos" from uncontrolled pilots.
  • Fail fast but learn deliberately—most pilots (80%+) won't reach production, so design intake and evaluation processes that funnel the best ideas forward.
4

Build the Infrastructure Foundation

(The Plumbing Problem)

Only 22% of organizations say their architecture can fully support AI workloads today. Data silos, latency, and fragmented systems remain the biggest barriers. Data engineers report spending much of their time just fixing data pipelines.

22%

Organizations with AI-ready architecture

47%

Architects cite real-time processing as top gap

Action for Leaders:

  • Upgrade to unified data architectures (lakehouse, mesh, or hybrid), balancing speed, security, and governance.
  • Focus on real-time streaming—47% of architects cited real-time processing as their top gap.
  • Consolidate tools—overlapping pipelines and duplicate systems slow progress and increase risk.
  • Balance cloud choices—use multi-cloud for flexibility, but weigh compliance, cost, and data ownership. Sensitive workloads often need private cloud.
5

Treat Data as the Core Asset

Data is the differentiator. Two-thirds of organizations see greatest potential in integrating GenAI with proprietary data, yet almost half of data scientists still use LLMs without retrieval-augmented generation (RAG).

2/3

Organizations see potential in proprietary data + GenAI

~50%

Data scientists use LLMs without RAG

90%

Enterprise data is unstructured

Action for Leaders:

  • Invest in data governance and quality first—without reliable data, AI fails.
  • Exploit unstructured data—90% of enterprise data is unstructured, representing untapped value (diagnostic images, customer service logs, IoT data).
  • Apply a "data intelligence flywheel"—use AI to create insights, then feed back into improved products, services, and personalization.
  • Treat compliance and privacy as design principles, not afterthoughts.
6

From Rational Exuberance to Disciplined ROI

The report highlights a gap between $1T+ projected AI spend and near-term revenues. Executives warn against hype but agree that underinvesting is riskier than overinvesting.

$1T+

Projected AI spend vs. near-term revenue gap

Action for Leaders:

  • Tie AI projects to business outcomes—GM requires every dollar of compute to map to measurable return.
  • Adopt ROI frameworks—KPIs should include revenue per worker, error reduction, customer retention, and top-line growth, not just cost savings.
  • Practice strategic patience—returns may take years; don't kill promising initiatives prematurely.
  • Balance "table stakes" vs. "strategic bets"—optimize core workflows (contact centers, coding) while exploring breakthrough opportunities (new business models, external AI products).

Ready to Transform Your Enterprise with AI?

Partner with KData's experts to implement strategic AI initiatives that drive measurable ROI and competitive advantage.

Agentic AI Strategy Data Engineering & Analytics AI Governance & Ethics Enterprise AI Training
Get Expert Guidance
7

Democratize AI Across the Enterprise

Executives stress that AI is an augmentation tool, not just automation. Leaders are preparing for a future where natural language is the default interface to data. Nearly 60% of enterprise architects expect non-technical staff to interact with data solely via natural language within three years.

60%

Enterprise architects expect natural language as default data interface within 3 years

Action for Leaders:

  • Invest in AI assistants and self-service tools that empower employees across functions.
  • Balance governance and enablement—establish Centers of Excellence to coordinate policies while driving adoption.
  • Train and reskill staff to interpret AI outputs critically; judgment and domain expertise remain irreplaceable.
  • Experiment with agentic AI—task-specific agents that plan and act can extend beyond Q&A into decision-making and workflow execution.
8

Governance as the Growth Enabler

Governance is often seen as a brake, but the report reframes it as an accelerator. Half of data engineers say governance takes more time than anything else, yet weak governance risks compliance breaches and reputational harm.

Governance Paradox

50% of data engineers spend more time on governance than core work

Action for Leaders:

  • Embed security and governance into design, not as afterthoughts.
  • Adopt human-in-the-loop safeguards for high-risk use cases (finance, healthcare, defense).
  • Use AI responsibly in customer-facing products—test rigorously to avoid hallucinations or harmful outputs.
  • Anticipate regulation—privacy, safety, and consumer protection laws are evolving fast; leaders must stay ahead.
9

Global Lessons and Sectoral Insights

Financial Services

Lead in scaling GenAI

45%

Expanding production

Healthcare & Public

Lag due to regulatory hurdles

High Potential

Untapped data value

Regional AI Strategies

India

Aggressive scaling with open-source models

Japan

Prefers custom proprietary models

United States

Experimenting with riskier, innovative use cases

Action for Leaders:

  • Tailor AI strategy to sectoral realities—match risk appetite, regulatory context, and data maturity.
  • Learn from global peers—scaling approaches vary, but disciplined pilots and custom data integration are universal success factors.
10

The Leadership Mandate

"There is no such thing as a technology strategy. There's only a business strategy that technology supports."

— Executive Survey Participant

The central message: AI leadership is business leadership. There is no "AI strategy" separate from business strategy.

Action for Leaders:

  • Own the AI narrative at the C-level—make AI a core element of business transformation.
  • Shape culture and change management—value creation requires rethinking workflows, roles, and governance.
  • Champion responsible adoption—balance speed with trust, democratization with control.
  • Bet on data intelligence as the differentiator—models will commoditize, but proprietary data and business integration will endure.
Final Takeaway

From Hype to Hard Value

Enterprise AI is moving from hype to hard value. The winners will be those who master these strategic imperatives:

Modernize Infrastructure

Handle AI scale with unified data architectures

Exploit Data Assets

Proprietary and unstructured data as unique differentiator

Balance Horizons

Productivity gains with long-term revenue growth

Disciplined Pilots

Risk-graded pilots that scale securely

Democratize Access

AI access while enforcing strong governance

Strategic Leadership

AI as core business transformation element

The Transformation Imperative

In short, AI success requires as much organizational and cultural transformation as it does technical innovation. Leaders must align investments, infrastructure, governance, and workforce strategies to turn AI into a true competitive advantage.

Stay Connected with KData

Follow us on LinkedIn to get the latest insights on data engineering, Databricks, Snowflake, AI strategies, and cloud best practices. Join our professional community of data experts.

KData Company

Data Engineering & AI Experts

Join thousands of data professionals