Optimal Dynamics

The Partner Integrations page lists a named integration catalog, described as two-way integrations with TMS and ELD systems: McLeod, Samsara, Trimble, Motive, Maven, Isaac, Geotab, Solera, Manifold, Ditat and Motorcity Systems. The Agentic Automation Layer page states that Atlas connects decision outputs directly to execution systems, automatically accepting freight and adjusting as data arrives, and the Scale release has the agents searching, negotiating, bidding and procuring across load boards, direct shipper channels, email and EDI.

SourceOptimal Dynamics, optimaldynamics.com/platform/partner-integrations, /platform/agentic-automation and Scale release 2026-03-02read 2026-09-08

Atlas commands execution on the engine's decisions, embedding decision intelligence into role-based workflows across freight procurement, load planning and driver management. Procurement agents are launched per lane with target and available counts and run search, negotiate, bid and procure sequences across channels, and the Scale release describes the whole as a single execution layer aggregating opportunities from multiple sources. That is multi-step, multi-agent execution under a named vendor runtime.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automation and Scale release 2026-03-02read 2026-09-08

The agents act from a decision-native foundation that transforms the customer's operational reality (constraints, commitments and preferences) into actionable context: a continuously maintained model of the carrier's whole network, fed by two-way TMS and ELD integrations, with a network digital twin exposed in the Network Simulation product. That gives the agents a persistent, maintained and queryable structure over the customer's operational knowledge. It is an optimization model over structured operational data rather than a document index, and no retrieval mechanism is described.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automation and Scale release 2026-03-02read 2026-09-08

The vendor's Agentic Automation Layer page documents its own review-and-approve surface. Agent-found freight is presented with the loads available to accept and their combined impact on revenue and network balance, behind an "Accept Loads" control. Auto-acceptance is limited to freight that meets specific profit thresholds set as policy, and the procurement-agent dashboard gives operators view, cancel and launch controls over each agent. The approval surface is shown through product screenshots, their captions and one sentence of prose, and no page covers configuring approval policy.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automationread 2026-09-08

The vendor's Security page (updated October 2025) states SOC 2 Type 1 and Type 2 attestation under AICPA standards, scoped to the software services on its platform, with AES-256 encryption at rest, TLS 1.2+ in transit, AWS KMS key management, third-party penetration testing and static analysis. No SSO, SAML, SCIM, role-based access or customer-facing audit control is published.

SourceOptimal Dynamics, optimaldynamics.com/securityread 2026-09-08

The vendor's Agentic Automation Layer page documents two vendor-supplied surfaces. In the Explain pillar, Atlas translates optimization logic into plain-English decision stories, so users see why a load was rejected, a driver was selected or a plan changed. The procurement-agent dashboard shows agent statuses, bids awarded and load-level procurement detail per lane, a vendor-supplied record of what the agents did. Decision stories explain a decision and the dashboard is a live status and performance view; no run trace, log surface, retention period or per-action audit record is documented.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automationread 2026-09-08

Learned network state is carried across decisions by the optimization model itself, the same model that grounds the agents in the customer's network. The vendor's Personalize pillar describes Atlas learning and adapting to behavioral patterns, driver preferences, customer priorities and lane sensitivities, which is learning absorbed into the model rather than agent memory. No agent memory layer with a stated scope and lifetime and no session state is documented.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automation and Scale release 2026-03-02read 2026-09-08

No hosting region, region list, customer environment option or selection surface is documented. The vendor's Security page names Amazon Web Services as the hosting provider, with KMS, GuardDuty and Inspector, and the privacy policy (March 2025) covers the website and says nothing about product hosting or residency.

SourceOptimal Dynamics, optimaldynamics.com/security and /privacyread 2026-09-08

Atlas ships specialized agents out of the box, and procurement agents are a named unit the customer launches per lane from the Network Manager, with view, cancel and launch controls. That is a single agent type instantiated per lane, not a catalog of distinct agents a buyer selects among. The four solutions (Network Simulation, Bid Analysis, Network Management, Dispatch Management) are product modules and a pricing structure, not a pack.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automationread 2026-09-08

Continuous market and network signals bring work to the agents across named inbound channels. The Scale release describes an always-on autonomous agentic workforce operating continuously on the customer's behalf across load boards, direct shipper channels, email and EDI, aggregating opportunities from multiple sources, with the engine proactively identifying where capacity is constrained and imbalance is developing. The Agentic Automation Layer page has Atlas making real-time adjustments as new data arrives. No separate schedule or webhook surface is documented.

SourceOptimal Dynamics, Scale release 2026-03-02 and optimaldynamics.com/platform/agentic-automationread 2026-09-08

No customer or admin model choice, routing control or named provider is published. The agents run on a proprietary decision optimization engine, and whatever language model produces the plain-English decision stories is undisclosed.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automationread 2026-09-08

The vendor's Partner Integrations page states that the platform offers a robust API for custom integrations alongside its prebuilt TMS and ELD connectors. No developer subdomain, docs page, API reference or SDK is linked anywhere on the site, and no scope or access path is stated, so the API is asserted rather than documented. The agents reaching load boards, broker APIs and EDI is outbound action.

SourceOptimal Dynamics, optimaldynamics.com/platform/partner-integrationsread 2026-09-08

What the vendor documents is the Deploy pillar on its Agentic Automation Layer page: specialized agents automate environment configuration, validation and iterations across the setup, testing and rollout of new models. It is illustrated by a validation check confirming that the last known hours-of-service data is valid, with a response time and a note that the counter-evidence query returned no results. That check confirms the change installed correctly, not what the agent did.

Validation covers configuration and data integrity, environment readiness and input validity in the rollout path; nothing documented scores, compares, simulates or holds out agent behavior, or tunes the agents' own decisions against realized outcomes after deployment. The optimization engine's continuous evaluation of freight opportunities measures the customer's network rather than the agent.

SourceOptimal Dynamics, optimaldynamics.com/platform/agentic-automationread 2026-09-08

The agents reach load boards, brokers and shippers through APIs, EDI and email, and TMS and ELD systems through two-way integrations. No browser, desktop or remote computer control is documented.

SourceOptimal Dynamics, optimaldynamics.com/platform/partner-integrations and Scale release 2026-03-02read 2026-09-08