Governed Memory: A Production Architecture for Multi-Agent Workflows

Hamed Taheri

cs.AIMar 18, 2026

Governed Memory: A Production Architecture for Multi-Agent Workflows

Authors: Hamed Taheri

Abstract

Enterprise AI deploys dozens of autonomous agent nodes across workflows, each acting on the same entities with no shared memory and no common governance. We identify five structural challenges arising from this memory governance gap: memory silos across agent workflows; governance fragmentation across teams and tools; unstructured memories unusable by downstream systems; redundant context delivery in autonomous multi-step executions; and silent quality degradation without feedback loops. We present Governed Memory, a shared memory and governance layer addressing this gap through four mechanisms: a dual memory model combining open-set atomic facts with schema-enforced typed properties; tiered governance routing with progressive context delivery; reflection-bounded retrieval with entity-scoped isolation; and a closed-loop schema lifecycle with AI-assisted authoring and automated per-property refinement. We validate each mechanism through controlled experiments (N=250, five content types): 99.6% fact recall with complementary dual-modality coverage; 92% governance routing precision; 50% token reduction from progressive delivery; zero cross-entity leakage across 500 adversarial queries; 100% adversarial governance compliance; and output quality saturation at approximately seven governed memories per entity. On the LoCoMo benchmark, the architecture achieves 74.8% overall accuracy, confirming that governance and schema enforcement impose no retrieval quality penalty. The system is in production at Personize.ai.

Figures & tables

Figure 1: Governed Memory four-layer architecture. Agent nodes interact through a shared, organization-scoped API surface. Layers are independently configurable.

Figure 2: Dual extraction pipeline. A single LLM call produces both open-set facts and schema-enforced typed properties, followed by quality gates and write-side deduplication.

Figure 3: Governance routing with tiered modes (fast/full) and progressive context delivery. The session layer tracks delivered variables to inject only delta content on each autonomous step.

Figure 4: Reflection-bounded retrieval with entity-scoped isolation. The optional reflection loop generates targeted follow-up queries within bounded rounds; entity isolation is enforced by CRM key pre-filtering.

Dataset	Samples	Ground Truth Elements
Primary corpus	250 samples, 5 types	8–12 facts, 5–8 props/sample
Multi-source entity	5 sources	40 unique facts, 8 cross-source dupes
Entity isolation profiles	100 entities	High overlap with unique markers
Recall query sets	500 queries	Expected topics & min source counts
Governance variable pairs	5 pairs	Known targets for 15 tasks
Conflict pairs	30 pairs, 15 cats	Stale + fresh claims with known dates

Table 1: Experimental dataset summary.

Content Type	Samples	Fact Recall
Call notes	50	100%
Documents	50	100%
Emails	50	100%
Transcripts	50	100%
Chats	50	98%
Overall	250	99.6%

Table 2: Extraction quality by content type (250 samples).

Category	% of Total
Captured by both modalities	34%
Open-set only (long-tail insights)	38%
Schema-enforced only (typed values)	12%
Missed by both	16%
Combined recall	82.8%

Table 3: Coverage distribution across memory modalities (20 samples).

Condition	Avg Compl.	Avg Results	Avg Latency
No reflection (baseline)	37.1%	15.0	9.4s
API-managed, 1 round	40.4%	22.5	10.4s
Manual multi-hop, 1 round	61.2%	20.8	6.5s
Manual multi-hop, 2 rounds	62.8%	21.9	10.0s

Table 4: Reflection ablation on 10 hard multi-hop queries (4 conditions).

Metric	Definition	Result
Conflict detection (primary)	Fresh info surfaced in answer	83.3%
Full stale suppression (strict)	Answer contains only fresh keywords	33.3%
Incorrect stale	Answer reflects only the outdated claim	3.3%

Table 5: Semantic conflict resolution (30 pairs).

Condition	Avg Score /100	$\Delta$ vs. Baseline
A: No memory	79.5	,
B: Raw memory	85.2	+5.7
C: Open-set + governance	86.4	+6.9
D: Full governed memory	85.9	+6.4

Table 6: End-to-end evaluation (10 prospects, 3 runs each, sales rubric /100).

Category	Accuracy	vs. Human Baseline
Single-hop	78.7%	$-$ 16.4pp
Multi-hop	51.7%	$-$ 34.1pp
Temporal	64.6%	$-$ 28.0pp
Open-ended	83.6%	+8.2pp
Overall	74.8%	$-$ 13.1pp

Table 7: LoCoMo benchmark results.

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

Tier	Category	Entity Types	Detection
1	Secrets	API keys, private keys, passwords	Pattern matching
2	Financial PII	Credit cards, IBAN	Regex + Luhn
3	Identity PII	Social Security Numbers	Regex + validation
4	Contact PII	Emails, phones, IPs	Format matching

Appendix

Table 12

Density	Avg Recalled	Score /100	Mem Use /30
Sparse (0 memories)	0	69.3	19.0
Minimal (3)	3	86.0	28.0
Light (7)	7	88.0	28.5
Moderate (12)	12	84.4	26.0
Rich (20)	20	85.2	27.0
Full (30)	30	88.3	29.5

Appendix

Table 13

Step	Task	Without	With	Savings
1	Cold outreach	9,312	9,312	0%
2	Follow-up w/ pricing	6,802	959	85.9%
3	Support escalation	2,873	2,873	0%
4	Troubleshooting	2,698	1,358	49.7%
5	Closing proposal	9,482	982	89.6%
Total		31,167	15,484	50.3%

Appendix

Table 14

Score Pattern	Interpretation	Indicated Action
Low Context Util., high Completeness	Agent succeeded despite routing issues	Improve governance metadata
High Context Util., low Completeness	Appropriate context but insufficient memory	Improve memory coverage
Low Personalization, high Accuracy	Entity memories sparse or not recalled	Check density; review recall
Low across all criteria	Model or prompt issue	Review model & system prompt
High variance within criterion	Schema-data alignment issue	Refine low-scoring types

Appendix

Table 8: Recurring diagnostic patterns from evaluation records.

Preset	Criteria (weight)
Default	Accuracy (25), Relevance (25), Completeness (25), Context Util. (25)
Sales	Personalization (30), Value Prop. (25), CTA (20), Tone (25)
Support	Problem Understanding (25), Solution Accuracy (30), Clarity (25), Empathy (20)
Research	Thoroughness (30), Source Quality (25), Analysis (25), Organization (20)

Appendix

Table 16

Agent Memory Governance

See all 17 papers

AkasicMEM: Governed Enterprise Memory for Agents
Sep 22, 2026Jeongmin Bae, Yongjae Kim, Kyoung Hur +2Agent Memory Agent Memory Governance
GraphAI Seoul, Republic of Korea
Governed Shared Memory for Multi-Agent LLM Systems
Jun 23, 2026Yanki Margalit, Nurit Cohen-Inger, Erni Avram +2Multi-Agent LLM Systems Data Provenance
Computer Science and Information, Ben-Gurion University of the Negev
GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents
Jun 17, 2026Zhe Ren, Yibo Yang, Yimeng Chen +7LLM Agent Memory Agent Memory Governance
School of Artificial Intelligence, Jilin University · 2Shanghai Jiao Tong University · 3King Abdullah University of Science and Technology (KAUST) +2

Sep 22, 2026J/K move · Enter open · S save

AkasicMEM: Governed Enterprise Memory for Agents

Jeongmin Bae, Yongjae Kim, Kyoung Hur, Donghyoung Han+1

GraphAI Seoul, Republic of Korea

Agent Memory Agent Memory Governance Information Flow Control Agent Memory Management

Open paper AskPDF arXiv

Agent memory enables enterprise agents to retain knowledge acquired during work and reuse it across tasks and agents, turning execution experience into persistent organizational knowledge. Realizing this potential requires both source--memory integration, through which enterprise sources and accumulated memory can be utilized together, and memory governance, through which shared memory remains subject to organizational policies throughout its lifecycle. These requirements interact when information from enterprise sources persists in memory. As this information is repeatedly derived and reused under changing principals and policies, source restrictions may be bypassed, resulting in information leakage. Preventing such leakage requires authorization continuity, under which source restrictions remain effective throughout source-to-memory and memory-to-memory derivation and reuse. Existing approaches address these concerns individually, but do not treat source--memory integration, memory governance, and authorization continuity as combined core design targets across the memory lifecycle. We define Governed Enterprise Memory as agent memory designed around this combined scope and present AkasicMEM as its realization. AkasicMEM realizes authorization continuity through transitive lineage, policy composition during memory formation, and policy re-evaluation during retrieval. It is built on GraphAI's AkasicDB, a unified vector--graph--relational database whose storage and execution substrate enables the underlying operations of these mechanisms to be jointly optimized and executed.

Abstract

Agent Memory Governance

Explore similar work

AkasicMEM: Governed Enterprise Memory for Agents

Governed Shared Memory for Multi-Agent LLM Systems

GateMem: Benchmarking Memory Governance in Multi-Principal Shared-Memory Agents