Many manufacturers are asking the wrong AI question:
“Which LLM should we use?”
Whether the answer is GPT, Claude, Gemini, MAI, Llama, or another model, the choice alone will not determine success.
In fact, Microsoft’s recent AI strategy highlights an important reality: the future belongs to AI ecosystems, not individual models. Satya Nadella has emphasized that AI models are only one element of a much larger system that includes memory, context, tools, skills, user interactions, orchestration, and learning environments. Microsoft is increasingly using multiple specialized models optimized for specific tasks rather than relying on a single frontier model.
Manufacturers should take note.
The path to AI maturity is not selecting the “best” model. It is building the digital foundation that allows AI to understand products, execute workflows, learn from outcomes, and eventually create digital representations of business operations.
That foundation begins with PIM, CPQ, and Product Intelligence.
The Industry’s Obsession with Models
Today’s AI conversations often focus on:
- GPT vs. Claude
- OpenAI vs. Anthropic
- Microsoft MAI vs. frontier models
- Open source vs. proprietary platforms
These discussions matter, but they do not solve the challenges most manufacturers face.
Consider a company managing thousands of products, configurable assemblies, engineering constraints, pricing rules, and compliance requirements.
A different AI model will not fix:
- Inconsistent product data
- Missing configuration rules
- Duplicate specifications
- Conflicting part relationships
- Knowledge trapped in spreadsheets and individual experts
These are information and process problems. AI can only work effectively when the underlying knowledge is structured and accessible.
Satya Nadella’s Message: The Model Is Only One Piece
Satya Nadella’s recent discussion of Microsoft’s MAI strategy contains a lesson that extends far beyond Microsoft products.
Microsoft’s direction is not centered on creating one giant AI model that solves everything.
Instead, Microsoft describes an orchestration approach where specialized models, memory, tools, context, skills, workflows, and reinforcement learning environments work together as a complete system. Microsoft specifically states that frontier models from OpenAI and Anthropic will continue to operate alongside MAI models within a larger orchestration framework.
More importantly, Microsoft is training models using real product interactions, customer workflows, and business outcomes rather than relying solely on generic benchmark testing. Specialized models are being optimized for environments such as GitHub Copilot and Excel because real-world context matters more than raw model size.
This has enormous implications for manufacturing.
The future AI architecture inside an enterprise will almost certainly be:
- Multi-model
- Context-driven
- Workflow-aware
- Tool-enabled
- Continuously learning
Exactly the same characteristics required for Digital Twins.
Digital Twins Start with Product Intelligence
Digital Twins are no longer limited to machines and factory equipment.
Organizations are increasingly creating digital representations of:
- Products
- Business processes
- Organizational knowledge and workflows
Before a company can create a Digital Twin of its operations, it must first create a structured representation of its products.
That is where PIM and CPQ play a critical role.
Why PIM Matters
AI cannot learn product expertise if product expertise has not been captured.
A Product Information Management (PIM) platform centralizes:
- Product attributes
- Specifications
- Documentation
- Relationships
- Taxonomies
- Product hierarchies
Without a governed source of product information, AI must rely on disconnected files, spreadsheets, emails, and tribal knowledge.
With PIM, AI gains access to trusted product intelligence that can support search, recommendations, automation, and decision-making.
In many ways, a PIM platform becomes the first step toward a product Digital Twin.
Why CPQ Matters
Product information alone is not enough.
Organizations also need to capture how products move through real business processes.
A Configure Price Quote (CPQ) platform contains:
- Configuration rules
- Compatibility logic
- Engineering constraints
- Pricing policies
- Approval workflows
- Sales processes
While often viewed as a quoting tool, CPQ also captures how decisions are made.
It records how products are configured, priced, approved, and sold. That makes CPQ one of the richest sources of operational intelligence in the enterprise.
The Future Is Specialized AI
Enterprise AI is moving toward specialized systems rather than a single universal model.
Manufacturers will likely deploy AI solutions focused on areas such as:
- Product configuration
- Guided selling
- Engineering validation
- Supply chain management
- Forecasting and analytics
These systems will work together through orchestration platforms and shared business knowledge.
The goal is not one AI system.
The goal is a connected ecosystem built around trusted product and process intelligence.
Every Enterprise Needs a Learning Environment
AI should do more than provide answers. It should improve over time.
To achieve that, organizations need environments where AI can learn from:
- Product selections
- Configuration outcomes
- Quote acceptance rates
- Order histories
- Support interactions
- Product performance data
The more AI is connected to real business outcomes, the more valuable it becomes.
The Roadmap to Enterprise AI
Most successful AI initiatives will follow a progression:
Phase 1: Product Intelligence
Establish PIM as the source of product truth.
Phase 2: Process Intelligence
Capture configuration, pricing, and workflow logic in CPQ.
Phase 3: AI Orchestration
Connect AI to trusted product and process knowledge.
Phase 4: Specialized Agents
Deploy AI systems focused on specific business functions.
Phase 5: Learning Environments
Continuously improve AI using real-world outcomes.
Phase 6: Digital Twins
Create digital representations of products, processes, and eventually the organization itself.
Many organizations are trying to start with AI before building the underlying foundation. That often limits results.
Beyond the LLM
The AI conversation should not start with:
“Which LLM should we choose?”
It should start with:
“What knowledge are we giving AI access to?”
Manufacturers that invest in product and process intelligence will be better positioned to support AI initiatives, Digital Twins, intelligent agents, and automation.
At VISTECH, we believe PIM and CPQ form the foundation of that strategy.
PIM creates a trusted definition of products, including attributes, relationships, documentation, options, and business rules. CPQ captures how those products are configured, priced, approved, and sold.
Together they create the structured knowledge layer that AI requires to deliver meaningful business value.
The question is not which model will win.
The real question is whether your organization is building the product and process intelligence needed for any AI model to succeed.