The AI-Centric Future Needs a Data-Centric Foundation

The AI-Centric Future Needs a Data-Centric Foundation: Latttice Delivers It
The Missing Foundation: Shift #0
When I first read McKinsey's latest article "The AI-centric Imperative: Navigating the Next Software Frontier", one thing was immediately clear: this isn't just a description of Silicon Valley start-ups, it's a blueprint for how every enterprise will need to operate. Their seven shifts capture a world where software is built faster, smaller teams deliver more, AI accelerates everything, and context flows seamlessly across systems.
But as someone who has spent years inside enterprise projects, I couldn't stop thinking one thing: None of these shifts are possible unless the organization has trusted, governed, business-owned data.
And that's exactly what most enterprises don't have. That missing piece is what I call Shift #0.
The Reality
Most enterprises lack the data foundation needed for AI transformation
The Solution
Shift #0 provides the business-owned, governed foundation
Shift #0: A Business-Owned, Governed, AI-Ready Data Foundation
McKinsey outlines seven shifts for software vendors to become AI-centric, but if we apply these same shifts to enterprises, one thing becomes obvious: none of the seven can work without a foundational transformation first.
This is Shift #0, the essential prerequisite that must be in place before any AI strategy, automation initiative, or data modernisation programme can succeed.
What Shift #0 Requires:
Business Ownership
The business must own the data, not just consume it
Data Products
Data must exist as governed, reusable products with clear accountability and ownership.
Embedded Governance
Governance must be built-in, not bolted on through committees, contracts, or checklists.
Unified Context
Context and semantics must be consistent across systems and domains.
Automatic Lineage
Lineage must be generated automatically, not manually curated.
Instant Access
Access must be immediate for authorised users and systems, not dependent on engineering queues.
Without Shift #0, Everything Fails
AI amplifies broken data instead of insights
Autonomous workflows collapse under poor quality
Costs explode without proper governance
Trust evaporates across the organisation
Dashboards become meaningless noise
Transformation programmes fail spectacularly
This isn't theoretical. It's what I've seen repeatedly across organizations, including one transformation I recently reviewed that is destined to fail unless it changes course immediately.
A Real Example: A Transformation Doomed Before It Begins
A major enterprise announced its "AI-ready transformation." The intention was sincere, but the reality was flawed from the start. This case study illustrates exactly why Shift #0 is non-negotiable for any organization pursuing AI-centric operations.
1
Platform Selection
A full platform stack was chosen before speaking to the business teams who would use it
2
Missing Foundation
No semantic understanding, governance model, or success criteria were established
3
Excluded Stakeholders
No involvement from domain teams who own the business context
4
Cost Blindness
No consideration of compute cost or price-performant approaches
5
No Data Strategy
No data foundation or ownership model in place
Critical Warning: If they don't course-correct, if they don't implement Shift #0 and apply price-performant compute (PPC), the programme will fail. The pattern is predictable and has been seen countless times before.
The Predictable Failure Pattern
Timeline Collapse
Timelines blow out as teams struggle with data quality and access issues
Budget Overruns
Budgets sink as compute costs spiral and rework becomes necessary
Trust Erosion
Dashboards appear but aren't trusted by business users
AI Stagnation
AI projects stall due to poor data quality and governance
Leadership Crisis
Leadership loses confidence in the transformation programme
This is the same pattern highlighted by research across the industry. Most data initiatives fail according to Gartner. Only 17% of organizations have a real data product strategy according to BARC Germany. Enterprises are drowning in data they cannot use, as reported by MIT Technology Review.
Shift #0 isn't optional. It's the foundation everything else depends on.
AI Fails When Data Fails, The Evidence Is Everywhere
Across industries, the highest-profile AI failures all share a single root cause: broken data. These aren't small start-ups or experimental projects, these are major technology companies and enterprises with vast resources. Yet they all stumbled on the same fundamental issue.
IBM Watson Health
Collapsed because clinical data was inconsistent and incompatible across healthcare systems
Zillow Offers
Lost $420M in one quarter because its pricing model misread market signals from poor data
Amazon Recruiting Engine
Was shut down after it absorbed historical human bias from skewed hiring data
Google Flu Trends
Overestimated flu activity for 100 out of 108 weeks due to flawed data interpretation
The Universal Pattern
Different industries. Different use cases. Different technologies. Yet the same failure pattern emerges every single time.
AI only works when the data works.
Data Mesh: Brilliant Theory, Broken in Practice
Originally, Data Mesh introduced a powerful idea: the business should own its data. This was revolutionary thinking that promised to transform how organizations manage and leverage their data assets.
But without AI or zero-code tools, business teams simply couldn't build or govern data products themselves. The vision was sound, but the technology to enable it didn't exist yet.
The Engineering Takeover
And so Data Mesh was taken over by engineering teams who implemented it the only way they knew how, through traditional technical approaches that defeated the original purpose.
Pipelines
Complex data pipelines requiring specialist knowledge
Contracts
Rigid data contracts that added friction
Committees
Endless governance committees and bureaucracy
Technical Complexity
YAML, semantic layers, and federated governance overhead
The Disappointing Result
Business further removed from their data
Poor adoption across organizations
Huge consulting costs with limited returns
Heavy engineering overhead and bottlenecks
Data contracts that added friction instead of trust
Even more alarming is that the original Data Mesh founders and early advocates are now building “Data Mesh technologies” using the same engineering-first mindset that caused Data Mesh to fail the first time.
This is exactly the vendor problem McKinsey describes: new technology built on old thinking.
It ignores what the business is actually asking for, simplicity, speed, ownership, and instead recreates the same complexity under a new label. 
The result?
New tools, same old outcome: business teams are still locked out.
Our Interpretation: Data Mesh Re-Imagined for the AI Era
A Business Empowerment Model
At Data Tiles, we never saw Data Mesh as an engineering architecture. We saw it as a business empowerment model, and AI finally makes that possible.
AI + Zero-Code = The Original Vision, Finally Realised
What Latttice Enables:
Zero-Code Creation
Domain teams can create data products with zero code, removing technical barriers entirely
Automatic Governance
Governance is embedded automatically into every data product from creation
Generated Lineage
Lineage is generated automatically, providing complete transparency
AI-Unified Semantics
Semantics are unified through AI, eliminating manual mapping
Instant Data Fusion
Structured and unstructured data fuse instantly without complex integration
Immediate Access
Access is immediate for authorised users across the organisation
Natural Language
Natural language becomes the interface, democratising data access
This is what Data Mesh always intended to be,  but could never become until AI made it possible. The vision was always correct; we simply needed the technology to catch up with the ambition.
Traditional Mesh vs Latttice Mesh
The difference between traditional Data Mesh implementations and Latttice's AI-era approach represents a fundamental shift in how organizations can empower their teams and leverage their data assets.
Traditional Data Mesh
Engineer-led implementation
Pipeline heavy architecture
Data contracts and bureaucracy
YAML and technical interfaces
Business as consumer only
Batch processing and brittle
Hard to scale effectively
Latttice AI-Era Data Mesh
Business-led empowerment
Zero pipelines required
Governance built-in automatically
Natural language interfaces
Business as creator and owner
Real-time and governed
Scales at the pace of conversation
Latttice is Data Mesh re-invented for AI-centric organizations. And this is where McKinsey's seven shifts become real and achievable for enterprises of any size.
How Latttice Enables All Seven of McKinsey's AI-Centric Shifts
With Shift #0 in place through Latttice, organizations can finally achieve all seven of McKinsey's AI-centric shifts. Each shift builds upon the foundation of business-owned, governed, AI-ready data.
01
AI-Centric Product Development
Business teams create governed data products in minutes, no engineers required. Domain experts become data product creators.
02
AI-Driven Business Models
Data products become operational assets and monetisable services that generate new revenue streams.
03
AI-Accelerated Go-To-Market
Real-time, trusted data powers personalisation, retention, and customer intelligence at scale.
04
AI-Augmented Development
Developers no longer waste time finding or preparing data, AI handles the context automatically.
05
AI-Enabled Operations
AI workflows finally work because the underlying data is governed and trusted from the start.
06
AI-Ready Infrastructure with PPC
Price-Performant Compute reduces AI and analytics costs by 50–80%, making AI economically viable.
07
AI-Fluent Talent Models
Domain teams own their data; engineers support rather than gatekeep, transforming organizational dynamics.
The Foundation Makes Everything Possible
All seven shifts depend on Shift #0. Without it, they remain aspirational concepts rather than operational realities.
Latttice delivers Shift #0, so the seven shifts can finally work in enterprise environments.
The Future Belongs to Those Who Own Their Data
McKinsey captured the future of AI-centric software with remarkable clarity. But, for enterprises to become AI-centric, what my experience confirms, is that none of it is achievable without a business-owned, governed, AI-ready data foundation.
That is what Latttice was built to deliver. After years of watching enterprises struggle with data initiatives that promised transformation but delivered frustration, we created a platform that finally makes the original Data Mesh vision possible.
Shift #0 is not a concept. It's the foundation the AI era demands.
50-80%
Cost Reduction
Reduction in AI and analytics costs through price-performant compute
17%
Current Maturity
Organizations with a real data product strategy today (BARC Germany)
100%
Foundation Required
Of AI-centric shifts depend on Shift #0 being in place first
And for the first time, thanks to AI and zero-code capabilities, business teams can finally own their data destiny. The technology has caught up with the vision, and organizations that embrace this shift will define the next era of enterprise software.
Join a Data Conversation,
Jessie Moelzer.
References
McKinsey & Company (2024). "The AI-centric Imperative: Navigating the Next Software Frontier"
Gartner Research (2023–2025). Failure Rates in Enterprise Data Initiatives.
BARC Germany (2024). Data Products and Organisational Maturity Study.
MIT Technology Review (2024). The Paradox of Data Abundance and Inaccessibility.
IBM Watson Health Case Study (2017–2021).
Zillow Offers Financial Reports (2021).
Amazon AI Recruiting System Report (2018).
Google Flu Trends Retrospective Analysis (2015).