AI-assisted human-law collaboration protocol system

Abstract

This system provides an AI-powered inter-corporate collaborative protocol that, when multiple corporations use AI for project processing, outsourcing, resource sharing, or outcome acceptance, comprehensively verifies the delegation conditions, approval conditions, resource constraints, and compensation conditions of each corporation, prevents execution when conditions do not match, and enables collaborative processing only when conditions are met. [Solution] The AI-powered inter-corporate collaborative protocol system comprises a collaborative proposal acquisition unit 151, a delegation condition matching unit 152, an agreement generation unit 153, a collaborative execution control unit 154, and a collaborative audit record unit 155. The agreement generation unit generates multiple agreement candidates with candidate_ids, and uses the agreement data corresponding to the selected selected_candidate_id for collaborative execution. The delegation condition matching unit checks the differences in conditions for each corporation in advance and controls whether or not execution is possible. The collaborative audit record unit maintains the entire process as an audit record and enables tamper detection through hash chaining.

Description

Technical Field

[0001] The present invention relates to a technology for supporting corporate operations using artificial intelligence (AI). In particular, when AI agents or AI decision-making modules belonging to multiple corporations perform collaborative processing related to case processing, outsourcing, resource financing, consideration adjustment, and result acceptance, they collate the pre-set commission conditions, approval conditions, resource constraints, and consideration conditions for each corporation, and according to the results, generate a plurality of agreement candidates with `candidate_id`, and based on the agreement data corresponding to the selected `selected_candidate_id`, it relates to an AI-assisted inter-corporate cooperation protocol system that branches to execution, suspension, conditional rollback, or corrective escalation.

Background Art

[0002] In recent years, AI has been used as an auxiliary or semi-autonomous means for proposal generation, judgment assistance, data shaping, external API calls, etc. in various corporate operations such as contract review, financial management, registration, internal control, procurement, audit, logistics, and customer response.

[0003] Also, within a single corporation, a corporate operation technology has been proposed that combines a machine-readable articles of incorporation, pre-set approval policies, an AI execution engine, an audit node, an information sharing infrastructure, and an audit log module to allow automatic execution under certain conditions while maintaining auditability and transparency.

[0004] Furthermore, a configuration in which multiple corporations share an information sharing infrastructure and share contract condition databases, audit nodes, external APIs, or audit trails is also assumed. However, most of the conventional ones focus mainly on information sharing itself or audit records themselves, and the framework for treating "what, under what conditions, who, and how far automatic execution is allowed" as a machine-readable condition collation among multiple corporations is not sufficient.

[0005] In particular, when handling cases between corporations, the scope of delegable tasks, monetary limits, approval requirements, deliverable acceptance conditions, external transmission permissions, resource usage limits, return conditions, and division of responsibility may differ from one corporation to another. Therefore, simply having AIs exchange requests and responses could lead to the erroneous execution of tasks due to mismatched conditions.

[0006] For example, if company A automatically outsources part of a contract review to company B, and company B then uses company C's AI services, then the terms of delegation from company A, the approval policies from company B, the resource constraints from company C, and the compensation terms between each company all become issues simultaneously. If the AI ​​performs these tasks autonomously without prior verification of these factors, problems such as delegation beyond its authority, excessive resource consumption, generation of unacceptable deliverables, insufficient evidence, and unclear responsibility may arise.

[0007] Therefore, for AI collaborative processing across multiple legal entities, there is a need for technology that can machine-readably compare the differences in conditions for each legal entity, generate multiple agreement candidates with `candidate_id` based on the comparison results, determine the `selected_candidate_id` from among them, and then branch to execution, postponement, rejection, or corrective escalation. [Prior art documents] [Patent Documents]

[0008] [Patent Document 1] Japanese Patent Application No. 2026-005527 "AI-assisted corporate management support system"

[0009] The company discloses a corporate management system equipped with machine-readable articles of incorporation, an AI execution engine, audit nodes, and an information sharing platform. While it shows the overall architecture of corporate management, it does not disclose an inter-corporate collaborative protocol system that integrates the functions of machine-readable delegation terms verification, normalization of resource constraints, consideration conditions, and outcome acceptance conditions, consensus generation, collaborative execution control, and collaborative audit records. [Patent Document 2] Japanese Patent Application No. 2026-007023 "AI-assisted corporate management support system"

[0010] The system discloses data item definitions such as AI_registry_id, candidate_id, proposal_id, and policy_version, as well as version control of approval policies, task generation, and audit trail recording.

[0011] While it provides a basic structure for AI agent management, it does not describe inter-company protocols for normalizing delegation conditions, resource constraints, compensation conditions, and outcome acceptance conditions across multiple legal entities, generating and evaluating agreement candidates, and controlling collaborative execution. [Patent Document 3] US11474854B2

[0012] A patent relating to a collaborative system between multiple entities. While it provides information sharing and notification functions, it does not mention an inter-company protocol integrating a machine-readable delegation condition matching unit, agreement generation unit, and collaborative execution control unit. [Non-patent literature]

[0013] [Non-Patent Document 1] A. Norta et al., "Designing a Smart-Contract Application Layer for Transacting Decentralized Autonomous Organizations", Springer, 2016

[0014] https: / / link.springer.com / chapter / 10.1007 / 978-981-10-5427-3_61

[0015] Design of a smart contract application layer for decentralized autonomous organizations. While DAO foundational technologies are discussed, specific implementations for machine-readable condition matching, normalization of resources, compensation, and outcomes, consensus generation, and cooperative execution control are not shown.

Summary of the Invention

Problems to be Solved by the Invention

[0016] The present invention has been made to solve the above problems, and when performing case processing, business outsourcing, resource financing, or result acceptance using AI among multiple legal persons, it comprehensively collates the commission conditions, approval conditions, resource constraints, and consideration conditions of each legal person, suppresses execution when the conditions do not match, and aims to provide a mechanism that can establish collaborative processing only when the conditions are met.

[0017] The first problem of the present invention is to make the governance conditions distributed among multiple legal persons machine-readable and collatable on a case-by-case basis.

[0018] The second problem of the present invention is to enable automatic or semi-automatic processing branches such as execution, suspension, conditional return, re-execution request, and correction escalation according to the collation result.

[0019] The third problem of the present invention is to maintain an audit record that can be verified at a later date and can detect tampering for the entire process of inter-legal-person collaborative processing.

Means for Solving the Problems

[0020] An AI-assisted inter-legal-person collaborative protocol system according to one aspect of the present invention is an information processing system in which AI agents or AI decision-making modules belonging to multiple legal persons share and execute at least a part of case processing, and includes a collaborative proposal acquisition unit, a commission condition collation unit, a consensus generation unit, a collaborative execution control unit, and a collaborative audit record unit. The consensus generation unit generates a plurality of consensus candidates with `candidate_id`, and uses the consensus data corresponding to the selected `selected_candidate_id` among the plurality of consensus candidates for collaborative execution.

[0021] The aforementioned collaborative proposal acquisition unit can acquire collaborative proposals that include project information, request details, target resources, deliverable conditions, desired delivery date, desired compensation conditions, or requested time, corresponding to a collaborative request from one corporation to another corporation.

[0022] The delegation condition matching unit retrieves at least a portion of the delegation conditions, approval conditions, resource constraints, and consideration conditions pre-set for each of the multiple legal entities, associating them with the `entity_id` and `policy_version` corresponding to each legal entity, and can determine whether or not it can be executed or the conditions for execution by comparing them with the project information, requirements, target resources, or deliverable conditions included in the collaborative proposal.

[0023] Based on the determination result, the agreement generation unit generates multiple agreement candidates with `candidate_id`, each including at least one of the following: responsible company, target work, work unit, resource allocation conditions, compensation conditions, delivery date conditions, result acceptance conditions, return conditions, and recipient of correction notifications. The agreement generation unit can then determine the agreement data corresponding to the selected `selected_candidate_id`.

[0024] The cooperative execution control unit may, based on the agreement data or determination result corresponding to the `selected_candidate_id`, select at least one of the following actions for the processing corresponding to the cooperative proposal: execute, hold, reject under conditions, request re-execution, request manual approval, or escalate for correction.

[0025] The aforementioned collaborative audit record unit stores collaborative proposals, matching results, agreement candidates, selected agreement data, execution results, remand results, re-execution requests, or corrective escalation results as audit records, and may, if necessary, add and retain `candidate_id`, `selected_candidate_id`, `hash_value`, signature value, or prior record reference information.

[0026] The aforementioned delegation conditions may include, for example, a corporate identifier, a role identifier, the type of work that can be delegated, whether sub-delegation is permitted, whether approval is required, a monetary limit, a resource usage limit, whether deliverables can be taken off-site, external transmission conditions, SLA conditions, or return conditions.

[0027] The aforementioned resource constraints may include CPU time, GPU time, token consumption, API usage, storage usage, human review slots, budget limits, deadlines, or priorities.

[0028] The aforementioned compensation conditions may include at least one of the following: fixed fee, usage-based fee, emergency response surcharge, reduction for failure to meet SLA, reduction for re-execution, final amount after approval, final amount after acceptance.

[0029] The aforementioned conditions for accepting deliverables may include at least one of the following: fulfillment of mandatory requirements, non-applicability of prohibited items, completion of review, meeting or exceeding a specified quality score, attachment of specified supporting documents, conformity to specified format, or delivery within the deadline.

[0030] The aforementioned system can normalize and use for matching different ledgers, approval policies, contract terms data, or audit ledgers for each of the multiple legal entities. For normalization, version information, item correspondence information, schema correspondence information, value conversion rules, currency conversion rules, or unit conversion rules may be used, and the candidate record side may maintain the correspondence between `candidate_id`, `selected_candidate_id`, and `hash_value`.

[0031] The aforementioned system may operate on an information sharing platform for multiple corporations, and this information sharing platform may be implemented as a centralized database, a cloud database, a distributed database, a blockchain platform, or a combination thereof.

[0032] The aforementioned system may adopt a configuration in which each legal entity has its own AI execution engine and audit node, while only registering proposal data, verification results, agreement data, and audit records necessary for inter-legal cooperation on the shared platform.

[0033] The aforementioned corrective escalation may include notification to administrator terminals, supervisor terminals, audit terminals, legal terminals, or external expert agency terminals, trail freezing, re-approval requests, manual review transfers, or requests for termination of external contracts. [Effects of the Invention]

[0034] According to the present invention, in AI collaborative processing spanning multiple corporations, the ability to execute a process can be controlled after prior verification of the differences in conditions among each corporation, thereby preventing unauthorized execution, execution with inconsistent conditions, excessive resource consumption, or the generation of unacceptable results.

[0035] Furthermore, according to the present invention, since the protocol includes not only execution but also suspension, conditional remand, re-execution request, or corrective escalation, it is possible to improve practical usability, auditability, and clarification of responsibilities in collaboration with multiple legal entities.

[0036] Furthermore, according to the present invention, the entire process of inter-company collaborative processing is retained as an audit record and can be tampered with by hash chaining or the like, thereby enhancing the post-dispute verifiability, explainability, and effectiveness of internal controls in the event of a dispute. [Brief explanation of the drawing]

[0037] [Figure 1] This is a block diagram showing the overall configuration of an AI-assisted human-cooperative protocol system according to one embodiment of the present invention.

[0038] [Figure 2] This is a functional block diagram showing the functional coordination between the Cooperative Proposal Acquisition Unit, the Delegation Condition Verification Unit, the Agreement Generation Unit, the Cooperative Execution Control Unit, and the Cooperative Audit Recording Unit.

[0039] [Figure 3] This is a process flow diagram for normalizing and comparing the delegation conditions ledger, approval policy ledger, resource ledger, and audit ledger for multiple legal entities.

[0040] [Figure 4] This is a decision flowchart showing the branching paths for a collaborative proposal: execution, postponement, rejection with conditions, request for re-execution, or escalation for correction.

[0041] [Figure 5] This is a sequence diagram illustrating an example of the verification and return process for deliverables.

[0042] [Figure 6] This figure shows the process of generating corrective data with added responsibility demarcation information and notifying the supervisor's terminal.

[0043] [Figure 7] This diagram shows how audit records with hash chains are stored.

[0044] [Figure 8] This is a use case diagram illustrating an embodiment of the present invention in which corporations A, B, and C share the processing of contract review tasks. [Modes for carrying out the invention]

[0045] 1. Overall structure

[0046] As shown in Figure 1, the AI-assisted inter-corporate cooperation protocol system 100 of this embodiment includes corporate system A 110, corporate system B 120, corporate system C 130, and an information sharing platform 140. Each corporate system 110, 120, and 130 may have an AI execution engine, an audit node, a proposal generation module, an evaluation module, an execution module, and an audit log module, respectively.

[0047] The information sharing platform 140 may include a collaborative proposal registry 141, a delegation conditions ledger 142, an approval policy ledger 143, a resource ledger 144, an agreement data ledger 145, an audit record ledger 146, and a normalization rule storage unit 147, and the agreement data ledger 145 and the audit record ledger 146 may hold at least a portion of `candidate_id`, `selected_candidate_id`, and `hash_value`.

[0048] The collaborative proposal registry 141 may register the following: the task identifier `task_id`, the entity ID as the requester, the entity ID as a candidate recipient, the request details, the target resources, the deadline, the desired compensation conditions, the conditions for accepting deliverables, and the evidence document reference `evidence_ref`.

[0049] The delegation conditions ledger 142 may store the types of work that can be delegated for each corporation, whether or not sub-delegation is permitted, whether or not approval is required, the maximum amount, the maximum resource usage, whether or not external transmission is permitted, the conditions for retaining deliverables, and the conditions for returning the work to the delegation office.

[0050] The approval policy ledger 143 may store policy_id, policy_version, set_by_decision_body_ref, effective_from, effective_to, allowed_scope, and allow_conditions.

[0051] Resource ledger 144 may store CPU usage limits, GPU usage limits, API rate limits, human review limits, task processing queue lengths, and budget balances for each corporation.

[0052] The agreement data ledger 145 may record agreement_id, mandate_id, `task_id`, `candidate_id`, `selected_candidate_id`, `entity_id` (managed as a base concept including role-derived representations), resource_allocation_rule, consideration_rule, acceptance_condition, rollback_condition, `policy_version`, and `hash_value`.

[0053] The audit logbook 146 may store the proposal acquisition time, matching results, candidate content for each `candidate_id`, the set of conditions selected by `selected_candidate_id`, the applied `policy_version`, the branch result of execution or rollback, notification recipient identification information, and `hash_value`.

[0054] 2. Collaborative proposal acquisition department

[0055] The Collaborative Proposal Acquisition Unit 151 receives collaboration requests from one corporation to another and structures these collaboration requests as collaborative proposals. Collaborative proposals may include the business category, target documents, desired deadline, required quality, desired budget, confidentiality classification, whether or not external transmission is permitted, and the format of the deliverable submission.

[0056] The collaborative proposal acquisition unit 151 may, if necessary, extract the type of work, monetary conditions, or conditions for accepting results from the proposal text using natural language processing and normalize them into a standard schema.

[0057] 3. Delegation Condition Verification Unit

[0058] The delegation conditions matching unit 152 obtains the delegation conditions, approval conditions, resource constraints, and consideration conditions corresponding to the requesting corporation, the candidate recipient corporation, and, if necessary, the candidate sub-delegator corporation.

[0059] The delegation condition matching unit 152, if the ledger item names or data types differ for each of the multiple legal entities, can refer to the item correspondence table in the normalization rule storage unit 147 and convert, for example, "gpu_quota" and "accelerator_limit", and "amount_cap" and "max_fee" to a common schema.

[0060] The delegation conditions matching unit 152 matches the types of work, amounts, deadlines, requirements for external transmission, deliverable conditions, etc. included in the collaborative proposal with the conditions of each corporation and determines at least one of the following: feasible, conditionally feasible, requires approval, on hold, or not feasible.

[0061] For example, if company A makes a collaborative proposal to company B for "contract review," "within 48 hours," and "budget limit of 200,000 yen," and company B's terms of delegation are "contract review can be accepted," "automatic approval is possible if the amount is 200,000 yen or less," and "documents that are prohibited from being sent externally must be manually checked," then a decision will be made that takes into account the confidentiality classification of the document in question and whether or not it needs to be sent.

[0062] Furthermore, when Corporation B uses Corporation C's GPU analysis service, the available GPU slots, usage rates, emergency response surcharge conditions, and API call limits in Corporation C's resource ledger are checked.

[0063] 4. Consensus generation part

[0064] The agreement generation unit 153 generates multiple agreement candidates for establishing cooperative processing based on the determination result of the delegation condition matching unit 152, and stores each agreement candidate with a `candidate_id`.

[0065] Each proposed agreement may include the responsible entity, work assignments, resource allocation conditions, compensation conditions, deadline, rework conditions, return conditions, outcome acceptance conditions, responsibility demarcation information, notification recipients, associated `entity_id`, associated `evidence_ref`, and optionally `policy_version`. As a concrete example of proposed data, the following structured data may be generated: ```json { "candidate_id": "cand-D-2026-001-A", "task_id": "task-nda-2026-031", "entity_id": ["corp-A", "corp-B", "corp-C"], "timestamp": "2026-03-25T09:15:00Z", "policy_version": ["v2.1", "v1.8", "v1.5"], "work_allocation": { "corp-B": { "task": "text_review_analysis", "resource_limit": {"cpu_hours": 2.5, "token_count": 50000}, "compensation_amount": 120000, "compensation_currency": "JPY" }, "corp-C": { "task": "clause_extraction", "resource_limit": {"gpu_minutes": 90}, "compensation_amount": 40000, "compensation_currency": "JPY", "sla_penalty": {"condition": "sla_miss", "reduction_rate": 0.20} } }, "deadline": "2026-03-27T09:15:00Z", "acceptance_condition": { "mandatory_items": ["Extraction of prohibited clauses", "Clause comparison table"], "quality_threshold": 0.95, "evidence_required": true }, "rollback_condition": { "resource_overrun": true, "sla_breach": true, "approval_revoked": true }, "evidence_ref": ["ev-nda-031-proposal", "ev-nda-031-conditions"], "reason_code": ["auto_approved", "resource_available"], "hash_value": "e7a3c9f1b2d8e4a6c3f9b1d2e8a4c6f3" } ```

[0066] For example, a first candidate where company B is responsible for analyzing the review text and company C is responsible only for extracting clauses may be assigned the `candidate_id`: "cand-D-2026-001-A`, while other candidates such as a candidate processed solely by company B (`candidate_id`: "cand-D-2026-001-B`) or a candidate reserved by company C (`candidate_id`: "cand-D-2026-001-C`) may be generated. Subsequently, a candidate with the following compensation terms is selected as `selected_candidate_id`: "cand-D-2026-001-A": "120,000 yen to B, 40,000 yen to C, with a 20% reduction in C's compensation if the SLA is not met," and the resource allocation condition is "C's GPU usage is limited to a maximum of 90 minutes." The corresponding agreement data can then be finalized with `timestamp`: "2026-03-25T10:22:00Z."

[0067] The agreement generation unit 153 may, if there are condition mismatches, not finalize the complete agreement data, but generate a list of mismatches, proposed corrections, and a candidate for rejection with an identifier such as `candidate_id`: "cand-D-2026-001-reject-X", and may leave the `selected_candidate_id` corresponding to the selected rejection candidate in the audit record along with `timestamp`, `reason_code`: "condition_mismatch", and `evidence_ref`.

[0068] 5. Cooperative Execution Control Unit

[0069] The collaborative execution control unit 154 calls the AI ​​execution engine or external API of each legal entity according to the agreed data and executes collaborative processing. At the start of execution, the following may be recorded: `selected_candidate_id`, `task_id`, the `entity_id` of the executing entity, `timestamp`, the applied `policy_version`, and `evidence_ref`.

[0070] The cooperative execution control unit 154 stops the execution start and branches to either hold or return the condition if the condition mismatch has not been resolved. This branching decision may be accompanied by `reason_code`: "unresolved_mismatch" and the associated `evidence_ref`.

[0071] The cooperative execution control unit 154 may initiate a re-execution request, a manual approval request, or a corrective escalation if resource overload, SLA failure, artifact format mismatch, approval expiration, or external API failure is detected during processing. For example, if the GPU usage time is expected to exceed the 90-minute limit, the following structured corrective request data will be generated: ```json { "escalation_id": "esc-D-2026-031-001", "task_id": "task-nda-2026-031", "selected_candidate_id": "cand-D-2026-001-A", "entity_id": "corp-C", "timestamp": "2026-03-25T12:48:00Z", "issue_type": "resource_overrun", "resource_type": "gpu_minutes", "allocated_limit": 90, "estimated_usage": 120, "estimated_overrun_amount": 30, "policy_version": "v1.5", "reason_code": ["resource_exceeded", "sla_at_risk"], "evidence_ref": ["ev-gpu-log-031", "ev-realtime-monitor-031"], "fallback_plan_id": "fallback-D-031-pause", "notification_targets": ["admin-corp-A", "supervisor-corp-C"], "hash_value": "a2c9f8e1d3b7a4c6f2e8b1d9c3a7e4f6" } ```

[0072] The cooperative execution control unit 154 checks the status of acceptance conditions for the deliverable or processing result and may branch to accept, return due to insufficient conditions, request re-execution, reduce price, or request additional evidence. The acceptance decision record includes `task_id`, `selected_candidate_id`, decision result, `timestamp`, and the decision criteria `evidence_ref` and `hash_value`.

[0073] 6. Cooperative Audit Records Department

[0074] The collaborative audit record unit 155 saves audit records of the following events: when a collaborative proposal is obtained, when conditions are matched, when an agreement candidate with `candidate_id` is generated, when a candidate is selected using `selected_candidate_id`, when execution starts, when execution ends, when a rejection occurs, when a re-execution is requested, and when a corrective escalation occurs. Each event record is assigned a `timestamp` in ISO 8601 format.

[0075] Each audit record may include proposal_id, agreement_id, `candidate_id`, `selected_candidate_id`, `policy_version`, `entity_id`, action_type, decision_result, `reason_code`, `timestamp`, `evidence_ref`, and `hash_value`. For example, a candidate selection record has the following structure: ```json { "audit_record_id": "audit-D-031-005", "task_id": "task-nda-2026-031", "event_type": "candidate_selection", "candidate_id": ["cand-D-2026-001-A", "cand-D-2026-001-B", "cand-D-2026-001-C"], "selected_candidate_id": "cand-D-2026-001-A", "selection_timestamp": "2026-03-25T10:22:00Z", "selection_criteria": { "cost_efficiency": 0.85, "resource_availability": 0.92, "compliance_score": 0.98 }, "policy_version": ["v2.1", "v1.8", "v1.5"], "entity_id": ["corp-A", "corp-B", "corp-C"], "reason_code": ["optimal_cost", "resource_available", "compliance_met"], "evidence_ref": ["ev-selection-criteria-031", "ev-resource-check-031"], "hash_value": "d3f9c8e2a1b7c4f6e8d2a9b3c7e1f4a6", "prev_hash": "c2e8f1a9d3b6c4e7f2a8d1b9c3e6f4a5" } ```

[0076] The cooperative audit record unit 155 stores the `hash_value` corresponding to each record in association with the `hash_value` of the previous record (prev_hash), and may form a chain structure that allows for tamper detection by associating the `hash_value` with `candidate_id` or `selected_candidate_id` as needed. SHA-256 may be used as the hash algorithm.

[0077] Furthermore, audit records may be kept searchable from administrator terminals, audit terminals, or supervisor terminals, and may be reorganized and displayed on a case-by-case basis, by legal entity, by agreement_id, by `candidate_id`, by `selected_candidate_id`, or by responsibility boundary.

[0078] 7. Corrective escalation

[0079] The Cooperative Execution Control Unit 154 or the Cooperative Audit Record Unit 155 generates corrective data with at least one of the following attached: responsibility demarcation information, `candidate_id`, `selected_candidate_id`, or `hash_value`, if it detects a mismatch in conditions, mismatch in deliverables, violation of delegation conditions, over-allocation of resources, failure of external collaboration, or insufficient evidence.

[0080] The aforementioned responsibility demarcation information may include the requesting entity, the implementing entity, the sub-delegated entity, the time of the anomaly, the classification of the cause of the anomaly, the applied policy version `policy_version`, the applicable resource conditions, the applicable compensation conditions, whether re-execution is possible, whether temporary suspension is necessary, and the related `evidence_ref`. Specific corrective data example: ```json { "escalation_id": "esc-D-2026-031-002", "task_id": "task-nda-2026-031", "selected_candidate_id": "cand-D-2026-001-A", "timestamp": "2026-03-25T14:15:00Z", "issue_classification": "acceptance_condition_breach", "responsible_entity_id": "corp-B", "delegating_entity_id": "corp-A", "sub_delegated_entity_id": "corp-C", "policy_version": ["v2.1", "v1.8"], "resource_allocation": {"cpu_hours": 2.5, "gpu_minutes": 90}, "compensation_condition": {"corp-B": 120000, "corp-C": 40000, "currency": "JPY"}, "missing_items": ["Clause comparison table"], "re_execution_allowed": true, "temporary_suspension_required": false, "reason_code": ["mandatory_item_missing", "quality_threshold_not_met"], "evidence_ref": ["ev-acceptance-check-031", "ev-deliverable-log-031"], "fallback_plan_id": "fallback-D-031-rework", "notification_targets": ["admin-corp-A", "supervisor-corp-B", "audit-team"], "hash_value": "f8e2d3a9c1b7e4f6a2d8c3e9b1f4a6c7", "prev_hash": "d3f9c8e2a1b7c4f6e8d2a9b3c7e1f4a6" } ```

[0081] Corrective data may be sent to administrator terminals, supervisor terminals, audit terminals, legal terminals, or external expert agency terminals, and, if necessary, trail freezes, additional approvals, contract suspensions, payment withholds, or manual review transfers may be performed.

[0082] 8. Ledger integration and version difference absorption

[0083] If multiple legal entities employ different ledger structures, the normalization rule storage unit 147 may store item correspondence tables, value conversion rules, currency conversion rules, unit conversion rules, version correspondence tables, or schema conversion rules.

[0084] For example, the "max_external_fee" of corporation A, the "amount_cap" of corporation B, and the "maximum commission amount" of corporation C can be converted into a common field `consideration_cap` to make them comparable. The normalization process record may include the `entity_id` of the source ledger, the field name, the field name of the destination common schema, the conversion time `timestamp`, the applicable rule version `policy_version`, and the conversion evidence `evidence_ref`.

[0085] Furthermore, even if collaborative proposals based on older `policy_version` remain, the applicable version may be selected based on the effective date, and compatibility checks with successor versions may be performed as needed. In this case, the correspondence between the `candidate_id` generated in the old version and the current `selected_candidate_id`, the version transition time `timestamp`, the version difference `evidence_ref`, and the `hash_value` for consistency checks may be retained. [Examples]

[0086] Division of contract review work between corporations

[0087] Company A has accepted a request to review an NDA with a client, but wants to delegate the clause extraction process to Company B and the machine translation review of the foreign law clauses to Company C.

[0088] The collaborative proposal acquisition unit 151 generates a collaborative proposal that includes the project identifier `task_id`: "task-nda-2026-031", the requesting entity `entity_id`: "corp-A", the potential recipients `entity_id`: ["corp-B", "corp-C"], the desired deadline "2026-03-27T09:15:00Z" (48 hours later), the budget limit of 200,000 yen, the outcome acceptance condition "Extraction rate of prohibited clauses of 95% or more", and related `evidence_ref`: ["ev-nda-031-proposal", "ev-nda-031-client-requirements"]. The proposal generation time `timestamp`: "2026-03-25T09:15:00Z" is recorded.

[0089] The delegation conditions matching unit 152 matches the delegation conditions of Corporation B, "Japanese contract review permitted, external transmission permitted, automatic approval permitted for amounts under 200,000 yen" (`entity_id`: "corp-B", `policy_version`: "v1.8", `timestamp`: "2026-03-25T09:16:30Z"), with the delegation conditions of Corporation C, "English clause comparison permitted, GPU within 90 minutes, anonymization of documents containing personal information is mandatory" (`entity_id`: "corp-C", `policy_version`: "v1.5", `timestamp`: "2026-03-25T09:16:45Z"), by associating them with their respective `entity_id` and applicable `policy_version`. The matching record will have `evidence_ref`: ["ev-delegation-condition-corp-B", "ev-delegation-condition-corp-C"] appended to it.

[0090] The verification result indicates that the document to be sent to Corporation C contains personal information (`reason_code`: "pii_detected"), and therefore cannot be executed as is (`reason_code`: "condition_breach"), and `candidate_id`: "cand-D-2026-031-reject-001" is assigned as a candidate for return for anonymization resubmission. The determination time `timestamp`: "2026-03-25T09:18:00Z", the basis for the determination `evidence_ref`: ["ev-pii-scan-031", "ev-corp-C-policy-v1.5"] and the determination record `hash_value`: "b3d8f1c9a2e7b4f6c8d3a1e9b2f7c4a6" are recorded.

[0091] The agreement generation unit 153 does not generate complete agreement data, but instead identifies rejection data for "re-propose after anonymization," "corporate B can implement immediately," and "corporate C rejects conditions," and adds `candidate_id`: "cand-D-2026-031-reject-001", `timestamp`: "2026-03-25T09:19:00Z", `reason_code`: ["pii_detected", "anonymization_required"], `evidence_ref`: ["ev-pii-scan-031"] and `hash_value`: "c4e9f2a8d3b7c5f6e8d2a9b3c7e1f4a6" to the rejection data to be audited.

[0092] After Corporation A re-registers the anonymized document (`task_id`: "task-nda-2026-031-revised", `timestamp`: "2026-03-25T10:20:00Z", `evidence_ref`: ["ev-anonymized-doc-031"]), the delegation condition matching unit 152 determines that it is executable (`reason_code`: "conditions_met"), and the agreement generation unit 153 generates a candidate agreement as follows: Corporation B pays 120,000 yen, Corporation C pays 40,000 yen, the deadline is "2026-03-27T09:15:00Z" (48 hours), and if the SLA is not met, C's compensation will be reduced by 20%: ```json { "candidate_id": "cand-D-2026-031-A-revised", "task_id": "task-nda-2026-031-revised", "entity_id": ["corp-A", "corp-B", "corp-C"], "timestamp": "2026-03-25T10:21:00Z", "policy_version": ["v2.1", "v1.8", "v1.5"], "work_allocation": { "corp-B": {"task": "text_review", "compensation": 120000}, "corp-C": {"task": "clause_extraction", "compensation": 40000, "sla_penalty_rate": 0.20} }, "deadline": "2026-03-27T09:15:00Z", "evidence_ref": ["ev-anonymized-doc-031", "ev-revised-proposal-031"], "hash_value": "d5f8c1a9e2b7d4f6a8c3e9b1d2f7a4c6" } ``` The selected candidate is then confirmed with `selected_candidate_id`: "cand-D-2026-031-A-revised", selection time `timestamp`: "2026-03-25T10:22:00Z", and selection reason `reason_code`: ["optimal_allocation", "compliance_verified"].

[0093] The Cooperative Execution Control Unit 154 starts the AI ​​module of each corporation according to the agreed data corresponding to the `selected_candidate_id`: "cand-D-2026-031-A-revised" (`timestamp`: "2026-03-25T10:25:00Z", `evidence_ref`: ["ev-execution-start-031"]), and since the deliverable acceptance conditions have been met (`timestamp`: "2026-03-26T20:10:00Z", `reason_code`: ["acceptance_criteria_met", "quality_threshold_achieved"], `evidence_ref`: ["ev-deliverable-031", "ev-quality-check-031"]), completes the case, and sets the `hash_value`: corresponding to the completion record. Add "e6a9f2c8d1b7e4f6a2d8c3e9b1f4a7c6" to the audit record. [Examples]

[0094] Rework due to discrepancies in deliverables

[0095] If the review results prepared by Corporation B do not include a mandatory clause comparison table, the Cooperative Execution Control Unit 154 detects a mismatch in the outcome acceptance conditions (`timestamp`: "2026-03-26T18:30:00Z", `reason_code`: ["mandatory_item_missing"], `evidence_ref`: ["ev-acceptance-check-031", "ev-deliverable-incomplete-031"]), stops the acceptance process, and generates a return request.

[0096] The resubmission request includes the missing item "Clause Comparison Table", the resubmission deadline "2026-03-26T22:00:00Z", the resubmission destination `entity_id`: "corp-B", the reduction condition "15% reduction", the request for additional evidence `evidence_ref`: ["ev-comparison-table-template"], and the resubmission time `timestamp`: "2026-03-26T18:32:00Z", `reason_code`: ["deliverable_incomplete", "re_submission_required"], and `hash_value`: "f7b2d8c3a9e1f4b6d2a8c7e3b9f1a4c6". This ensures that the resubmission request is based on conditions linked to the `candidate_id`, rather than a vague "redo". [Examples]

[0097] Corrective escalation in the event of resource surplus

[0098] If, despite Corporation C accepting the assignment under the condition of a 90-minute GPU usage limit, 120 minutes of usage is expected during execution, the Cooperative Execution Control Unit 154 detects the resource overallocation (`timestamp`: "2026-03-25T12:48:00Z", `reason_code`: ["resource_overrun_predicted"], `evidence_ref`: ["ev-gpu-monitor-031", "ev-usage-projection-031"]) and temporarily suspends processing.

[0099] The collaborative audit record section 155 contains the estimated time of excess "2026-03-25T13:08:00Z", the applied resource condition "gpu_minutes: 90", the responsible entity identifier `entity_id`: "corp-C", `candidate_id`: "cand-D-2026-031-A-revised", `selected_candidate_id`: "cand-D-2026-031-A-revised", the difference value "estimated_overrun: 30 minutes", the applied `policy_version`: "v1.5", `reason_code`: ["resource_limit_exceeded", "sla_at_risk"], `evidence_ref`: ["ev-gpu-log-031-realtime"] and `hash_value`: Correction data containing "g8c2f9a1d3b7e4c6f8a2d9b3c7e1f4a6" is generated and notified to the administrator terminal `entity_id`: "admin-corp-A" and the supervisor terminal `entity_id`: "supervisor-corp-C" at `timestamp`: "2026-03-25T12:50:00Z".

[0100] If the administrator does not grant additional approval (`timestamp`: "2026-03-25T13:05:00Z", `reason_code`: ["additional_approval_denied"], `evidence_ref`: ["ev-admin-decision-031"]), the cooperative processing will be terminated and payment for the processing portion will be suspended based on the consideration terms (`payment_status`: "suspended", `suspended_amount`: 40000, `currency`: "JPY", `timestamp`: "2026-03-25T13:06:00Z", `hash_value`: "h9d3f8b2c1a7e4f6a8d2c9e3b1f4a7c6")). [Industrial applicability]

[0101] This invention is applicable to various industrial fields where AI-assisted task sharing is performed among multiple legal entities, including legal and contract review, accounting and tax support, registration and notification, purchasing and procurement, BPO, audit assistance, group company management, franchise headquarters management, SPV management, joint research management, supply chain coordination, and more.

[0102] In particular, it is well-suited for inter-company collaborative work that requires a high level of condition matching and audit traceability, such as coordination of review processes between financial institutions, quality control across supply chains in manufacturing, data management for multi-center clinical trials in pharmaceutical development, and process coordination among numerous subcontractors in the construction industry.

[0103] Furthermore, the AI-assisted inter-corporate cooperation protocol system according to the present invention can also be applied to collaborative operations between public institutions, such as administrative cooperation between governments and local authorities, data sharing between international organizations, and credit transfer management between educational institutions. [Explanation of symbols]

[0104] 100 AI-assisted inter-governmental collaborative protocol system 110 Corporate System A 120 Corporate B System 130 Corporate C System 140 Information sharing infrastructure 141 Collaborative Proposal Registry 142 Delegation Conditions Register 143 Approval Policy Ledger 144 Resource Ledger 145 Agreement Data Ledger 146 Audit Records Register 147 Normalization rule storage section 151 Collaborative proposal acquisition department 152 Delegation Condition Verification Unit 153 Consensus Generation Department 154 Cooperative Execution Control Unit 155 Cooperative Audit Records Department 201 Corporate System A (Representation on a Collaborative Platform) 201a AI Execution Engine A 201b Collaborative Interface A 202 Corporate System B (Representation on a Collaborative Platform) 202a AI Execution Engine B 202b Collaborative Interface B 203 Corporate System C (Representation on a Collaborative Platform) 203a AI execution engine C 203b Collaborative Interface C 210. Shared Platform (Collaborative Platform) 211 Collaborative Proposal Acquisition Department (on the platform) 212 Delegation Condition Verification Unit (on the platform) 212-N Normalization Engine 213 Consensus Generation Unit (on the platform) 214 Cooperative Execution Control Unit (on the platform) 214-E Execution Control Engine 214-M Execution Monitoring Department 214-A Outcome Acceptance and Judgment Department 214-R Re-processing Unit 214-T Anomaly Detection Trigger 214-RD Demarcation of Responsibility Determination Department 214-CG Correction Data Generation Unit 214-CA Corrective Action Selection Section 214-ES Corrective Escalation Department 215 Cooperative Audit Records Department (on the platform) 220 Shared ledger group 220-R Normalization Rules DB 221 Collaborative Proposal Registry (Shared Ledger) 222 Delegation conditions ledger (shared ledger) 223 Approval Policy Ledger (Shared Ledger) 224 Resource ledger (shared ledger) 225 Agreement Data Ledger (Shared Ledger) 226 Audit Records Ledger (Shared Ledger) 226-1~15 Audit Record Table Fields

Claims

1. An information processing system in which AI agents or AI decision-making modules belonging to multiple legal entities share and execute at least a portion of the processing related to case handling, outsourcing, resource sharing, compensation adjustment, or outcome acceptance, A collaborative proposal acquisition unit that acquires collaborative proposals, including project information, request details, target resources, or deliverable conditions, in response to collaborative requests from one corporation to another corporation. A delegation condition matching unit obtains at least one of the pre-set delegation conditions, approval conditions, resource constraints, or compensation conditions for each of the aforementioned multiple corporations, and compares them with at least one of the project information, requirements, target resources, or deliverable conditions included in the collaborative proposal to determine whether the collaborative proposal can be implemented or what the implementation conditions are. Based on the determination result by the delegation condition matching unit, the agreement generation unit generates agreement data that includes at least one of the following: the responsible corporation, the target work, resource allocation conditions, compensation conditions, and outcome acceptance conditions. A cooperative execution control unit that, based on the agreement data or the determination result, selects at least one of the following for the cooperative proposal: execute, hold, reject under conditions, or escalate to corrective action. A cooperative audit record unit stores the aforementioned cooperative proposal, the verification results by the delegated conditions verification unit, the agreed data, and the selection results by the cooperative execution control unit as audit records. Equipped with, The cooperative execution control unit, when it determines that execution is impossible or that there is a mismatch in conditions by the delegated condition matching unit, stops the processing related to the cooperative proposal or branches to processing related to condition rejection or corrective escalation. An AI-powered human-cooperative protocol system characterized by the following features.

2. In the AI-based human collaborative protocol system described in claim 1, The aforementioned delegation conditions, approval conditions, resource constraints, or compensation conditions include at least one of the following: corporate identifier, role identifier, type of delegable work, approval requirement, monetary limit, resource usage limit, outcome acceptance conditions, and return conditions. An AI-powered human-cooperative protocol system characterized by the following features.

3. In the AI-based human-cooperative protocol system according to claim 1 or 2, The agreement generation unit generates the agreement data, including resource allocation conditions or compensation conditions for each responsible corporation, based on at least one of the resource constraints, processing load, budget limits, compensation conditions, or requested time for each of the multiple corporations. An AI-powered human-cooperative protocol system characterized by the following features.

4. In the AI ​​human cooperation protocol system according to any one of claims 1 to 3, The aforementioned cooperative execution control unit determines whether the output or processing result generated as an execution result is acceptable or not based on the output acceptance conditions, and if there is a discrepancy, it branches to a process related to a return request or a re-execution request. An AI-powered human-cooperative protocol system characterized by the following features.

5. In the AI ​​human cooperation protocol system according to any one of claims 1 to 4, The delegation condition matching unit applies different matching conditions or determination conditions depending on at least one of the following related to the collaborative proposal: the type of project, the type of work to be performed, the amount range, whether or not external transmission is required, or the risk classification. An AI-powered human-cooperative protocol system characterized by the following features.

6. In the AI ​​human cooperation protocol system according to any one of claims 1 to 5, The aforementioned cooperative execution control unit, upon detecting at least one of the following: mismatch in conditions, mismatch in deliverables, violation of delegation conditions, over-allocation of resources, or failure of external system integration, generates corrective data with responsibility demarcation information added and performs corrective escalation to at least one of the administrator terminal, supervisor terminal, or legal review terminal. An AI-powered human-cooperative protocol system characterized by the following features.

7. In the AI ​​human cooperation protocol system according to any one of claims 1 to 6, The aforementioned cooperative audit record unit generates a hash value for each audit record and stores the hash value in association with the hash value corresponding to the record preceding the audit record. An AI-powered human-cooperative protocol system characterized by the following features.

8. In the AI ​​human cooperation protocol system according to any one of claims 1 to 7, The delegation conditions matching unit or the agreement generation unit uses information obtained from delegation conditions ledgers, approval policy ledgers, resource ledgers, or audit ledgers managed for each of multiple legal entities, after normalizing it based on at least one of version information, item correspondence information, or schema correspondence information, for matching or agreement generation. An AI-powered human-cooperative protocol system characterized by the following features.

Images