Research
Research papers on AI operating systems and AI at work
Research papers on AI operating systems for business, and methods for deciding what an AI team may do and what it is worth, each with its sources and limits.
- 12
- 3
- 168
- 2
Every research paper Topics: AI and governance, AI and business value, AI operating systems.
ResearchThe approval ladder: deciding how much an AI team may do without askingA method for grading each action an AI agent can take by what it reaches and whether it can be undone, setting the human approval each grade needs, and widening an agent's freedom only for reversible internal work and only on evidence from its own approved work. It sets out the statistics behind that evidence, the related frameworks it builds on, and where MatruOS does and does not yet follow the method.ResearchCounting the value an AI team creates: a method that uses a business's own recordsA method for counting what an AI team is worth from a business's own records. It separates three levels of claim (recorded, compared with the business's own baseline, and caused), gives each a test a finance head can apply, measures time saved from the business's own timings including rework, and requires every figure to carry its level, its basis and its limits. It also reports where MatruOS's own figures follow the method and where they do not yet.ResearchBridging the 94% Enterprise AI Value Gap: The Case for a Unified AI Operating System (AI OS)Individual AI assistance can leave a business dependent on disconnected stock checks, credit approvals, and manual order entry. This paper examines how a unified AI operating system could connect those activities through shared records, institutional knowledge, and controlled execution.ResearchInstitutional Memory as Operational Infrastructure: A Framework for Context Persistence, Validity, and Decision Traceability in AI Operating SystemsAn enterprise agent may retrieve an authentic document and still apply an instruction that is expired, superseded, or restricted to another customer. This paper examines how institutional knowledge becomes suitable for operational use in an AI operating system.ResearchTransactional Integrity and Agent Coordination in AI Operating Systems: Architectural Requirements Across CRM, ERP, and CommerceSeveral agents can make individually reasonable decisions that collectively exceed stock, credit, or fulfilment capacity. This paper examines the transaction boundaries required when an AI operating system coordinates customer relationship management, enterprise resource planning, and commerce.ResearchDelegated Authority and Human Accountability in AI Operating Systems: A Framework for Operational GovernanceGiving an agent access to a business application does not establish which decisions it may make or who remains responsible for their consequences. This paper develops a framework for delegated authority in enterprise AI operating systems.ResearchSecurity Boundaries in AI Operating Systems: Prompt Injection, Data Isolation, and the Protection of Execution AuthorityAn enterprise agent must read information from customers, suppliers, documents, and external services while retaining the authority limits established by the business. This creates a security problem when instructions embedded in that material influence tool use or access to protected data.ResearchFailure Propagation and Recovery in Agentic Business Operations: A Framework for Reliable ExecutionA business workflow can remain incomplete after an agent has produced a plausible response or received a successful result from one tool. Failures can occur between internal updates, external actions, and the communication of completion.ResearchThe Economics of Unified AI Operations: Workflow Costs, Coordination Overhead, and Enterprise Value AttributionThe cost of an AI workflow extends beyond model usage to implementation, supervision, exception handling, maintenance, and unresolved work. This paper develops an economic evaluation framework for enterprise AI operating systems using verified business completion as the unit of analysis.ResearchAI Operating System Adoption in Indian MSMEs: Organisational Readiness, Informal Knowledge, and Operational ConstraintsAn AI operating system requires usable records, explicit authority, and the capacity to manage exceptions after deployment. These conditions can differ substantially across Indian micro, small, and medium enterprises, even within one industry.ResearchEnterprise Transition to an AI Operating System: Interoperability, Context Portability, and Operational ContinuityIntroducing an AI operating system into an established enterprise requires decisions about which system owns each record and which service may create a business commitment during transition. This paper develops a migration framework that connects interoperability, data reconciliation, authority transfer, and recovery.ResearchEvaluating Enterprise AI Operating Systems: A Research Protocol for Execution Correctness, Control Effectiveness, and Operational OutcomesAn enterprise AI operating system should be evaluated through the business states it creates, the authority it respects, and the obligations it leaves unresolved. A fluent response or successful demonstration provides limited evidence for those requirements.
Run the whole business on one system Set up for your trade.
Describe your business once. MatruOS sets up your apps, brings in your data and puts Muni and the AI team to work, with your approval on every step that matters. 7-day free trial, no card needed.
See plans and what each includes