Identity and authority
Separate user identity, agent identity and delegated permissions; least-privilege tool scopes; time-bounded tokens.
Proposed educational extension · 4 November 2026
AI agents can combine planning, memory, tools and autonomous task selection with deterministic business-process automation. The resulting systems vary from a single agent with tools to layered, distributed, event-driven and cloud–edge multi-agent networks. No one topology is inherently AGI or ASI: exceptional generalisation and reproducible performance must be established independently.
Source and illustration policy: These are original explanatory schematics informed by architectural categories in industry guidance and the supplied diagram references. They are not copies of vendor figures, deployed systems, safety certifications or evidence that an ASI system exists. Original proposal records remain on their own dated pages.
| Topology | Control flow | Suitable applications | Key failure mode |
|---|---|---|---|
| Single agent and tools | One scoped planner with retrieval, memory and task-specific tools | Low-complexity, bounded workloads | Unnecessary tool permissions; unbounded execution |
| Sequential pipeline | Agent A → Agent B → verifier | Document processing, staged transformations | Error propagation; serial latency |
| Supervisor–worker | Planner → specialist workers → reconciler | Cross-domain evidence review | Central bottleneck; correlated source assumptions |
| Hierarchical teams | Coordinator → domain managers → specialist agents | Many interdependent domains | Complexity, tracing and compounded costs |
| Concurrent fan-out/fan-in | Independent branches → evidence-aware merge | Twelve-direction checks, independent analyses | Conflicting results, data races, high compute |
| Handoff / triage | Classifier or expert → appropriate specialist | Unexpected issues requiring specialist expertise | Routing loops, lost context and wrong delegation |
| Iterative maker–checker | Proposer ⇄ separate evaluator, with capped retries | Research synthesis and document validation | Self-confirmation, costly nonconvergence |
| Peer-to-peer mesh | Authenticated agent-to-agent exchanges | Federated cross-organisation discovery | Unclear ownership, message loops and trust boundaries |
| Event-driven BPA / agent mix | Scheduled/webhook events → deterministic rules → scoped agent | News digest, claims triage and back-office processing | Duplicate events, stale context, unsafe actions |
| Cloud–edge–device | Cloud coordination ↔ edge inference ↔ telemetry | Industrial monitoring and latency-sensitive systems | Network partitions, unsafe actuation and version mismatch |
Choose the simplest architecture that reliably meets the task, rather than multiplying agents by default. Hybrid workflows can switch between deterministic sequential steps, parallel specialist checks and human-reviewed interventions as dependencies and risk require.
Separate user identity, agent identity and delegated permissions; least-privilege tool scopes; time-bounded tokens.
Tool discovery and invocation are interfaces, not permission to act; validate arguments and outputs, enforce policy outside the model.
Typed schemas, versioned contracts, sender authentication, message integrity, replay protection and trace IDs.
Distinguish context-window state, durable task state, cited knowledge, inferred conclusions and update histories.
Idempotency keys, ordered event logs, locks/transactions, deadline propagation, cancellation and bounded queues.
Default read-only operation; explicitly approved writes; simulation before actuation; escalation and abort paths.
Trace agent paths, model versions, retrieved sources, tool invocations, latency, cost and corrections.
Retries with backoff, circuit breakers, timeouts, checkpoints, disaster recovery and manual fallback.
Independent source checks, calibration, causal validity, task-level rubrics and expert adjudication.
Purpose limitation, minimisation, retention controls, meaningful contestability and restricted data access.
Extract graph and text claims with source positions; send independent date, definition, forecast and contradiction checks; reconcile claim IDs and evidence; preserve uncertain or ambiguous relationships. No external writes are required.
Process records through dated rules, accessibility checks, deadlines, procedural standards and evidence registries. Specialist assistants may recommend action; a designated person reviews rights-sensitive decisions, with reasons and challenge routes recorded.
A scheduled workflow collects multiple feeds, de-duplicates records, validates timestamps and sources, summarises distinct claims, retains original links and routes a proposed digest for review. Event retries should be idempotent. A failed agent should not trigger repeated public posting.
Sensors provide telemetry to an edge gateway with deterministic alarm thresholds and network-partition safeguards. Cloud agents support diagnostics and planning. Physical interventions remain under independent operational safety and authorised human control.
Compare a single-agent baseline, tool-enabled agent, multi-agent design without independent assurance, and the corresponding design with an independent verifier. Assess all four on equal-cost and equal-wall-clock budgets, reporting both rather than collapsing them into a single score. Use held-out novel tasks, grounded claims, calibrated uncertainty, human expert adjudication, causal validity, consistency, idempotent execution, latency, token and API spend, permissions, reversibility and error recovery.
Record exact model and tool versions, dataset dates, seeds, timeout policies, retries and approval interventions. Parallel agents sharing one underlying model are not necessarily independent evidential sources. More complexity can worsen performance and costs.
A companion methodology sets out eight assessment dimensions—evidence, causality, probabilistic calibration, robustness, safety, autonomy, efficiency and generalisation—across matched baseline configurations and explicit compute budgets. Read the Advanced AI Evaluation Protocol. These are proposed tests rather than reported results.
Read alongside the Verifiable Advanced Intelligence framework, AI Research Lab, Evidence Verification Across Sectors, Information Science and Feature Request and Contribution Register. This page introduces new educational illustrations; it does not alter the historical records of prior proposals.
Research addition dated 4 November 2026. Educational design study, not a deployed workflow. No third-party diagrams reproduced.