From Holistic Thinking to an Enterprise AI Operating Model
Think holistically. Deliver precisely.
AI initiatives connect business decisions, people, systems, information and control. Their success depends on how these elements work together.
Kernity’s five dimensions provide a practical starting point for understanding those connections:
- Business impact: connect initiatives to business goals, assess benefits and costs, and establish how results are measured.
- People & culture: develop roles, ways of working and skills so people can take ownership, collaborate and put change into practice.
- Technology: build solutions that fit the business, integrate with existing systems and support reliable operation and further development.
- IT security & compliance: translate requirements into responsibilities, access controls and safeguards throughout design and operation.
- Data: make company data and knowledge usable across teams and systems, with appropriate quality, access and ownership.
Together, these dimensions connect the intended outcome with the conditions needed to achieve it. They help teams identify dependencies early and focus implementation on what matters.
For AI, this creates concrete operating questions. Who owns the business outcome? Which services and information sources are approved? How are outputs evaluated? What may an agent do? Who can suspend a service? What evidence justifies continued investment?
An Enterprise AI Operating Model turns these questions into repeatable responsibilities and decisions.
The model developed here organizes that work into seven capabilities:
| Operating capability | Practical responsibility | Connection to Kernity |
|---|---|---|
| Investment and value steering | Prioritize opportunities, allocate funding, assess full costs and verify benefits | Business impact |
| Workflow ownership and workforce enablement | Define ownership, redesign workflows, develop competence and support adoption | People & culture; Business impact |
| Platform and supplier management | Select services, manage integrations and dependencies, establish support and exit options | Technology |
| Data and knowledge stewardship | Maintain source ownership, quality, provenance, permissions and lifecycle obligations | Data |
| Delivery and service operations | Evaluate, release, monitor, change, recover and retire AI services | Technology, with contributions from all five dimensions |
| Risk governance and assurance | Establish decision rights, assess exposure, manage approvals and verify controls | Across all five dimensions |
| Security and delegated authority | Protect access, enforce action limits and enable intervention | IT security & compliance |
These capabilities make the five dimensions operational for AI. Their connections remain important: a production release depends on technical readiness, business acceptance, competent users, suitable information and effective safeguards.
Governance also spans the whole model. Investment authority, workflow accountability and risk acceptance require coordinated decisions across business and specialist functions.
Leadership sets priorities, boundaries and delegations. Shared functions provide common services and specialist support. Departmental business owners remain accountable for workflows and outcomes. Work spanning departments needs one end-to-end business owner.
Every material AI use follows a common lifecycle: set direction, frame the problem, validate, authorize production, operate, review and retire. Reviews lead to decisions to continue, scale, adapt or stop. Material changes return to the relevant evaluations and approvals.
Five evidence categories support those decisions: value, quality, cost, adoption and risk. Connected records link purpose and ownership to configuration, permissions, approvals, operating results and the next decision.
Implementation follows two profiles, with conditional extensions:
- Minimum viable: essential responsibilities work for the declared scope. Accountable owners, approved services, relevant evaluation, competent users, support and effective intervention are in place. Existing teams and systems can provide this baseline.
- Best-practice target: those responsibilities operate consistently, with measured outcomes, useful shared capabilities and improvements driven by operating experience.
- Advanced and emerging extensions: specialized capabilities address a demonstrated need. Examples include multi-model routing, continuous evaluation and agentic workflows. Advanced describes additional sophistication; emerging describes unresolved uncertainty. Extensions can support either profile.
Required safeguards follow exposure throughout. An agent executing consequential actions needs enforceable authority limits and tested intervention even within a minimum implementation.
Kernity’s five dimensions provide a fundamental mental model for framing organizational problems: examine the intended impact, the people involved, the enabling technology, the required safeguards and the information foundation. The AI operating model then supplies the specific responsibilities, decisions and evidence needed to act.
Think holistically. Deliver precisely. becomes a working method: understand the connections, establish accountable operation and improve through evidence.