Go back to Market Research
Aera Technology sells an enterprise decision-automation platform, Aera Decision Cloud, that connects to business systems, prepares data, runs decision logic, and writes actions back to operational applications. Its packaged Skills cover forecasting, inventory, order fulfillment, logistics, procurement, and other business functions. Customers can adapt these Skills or create their own using models, rules, code, and agents. (1, 6)
Overall supply chain score: 4.5/10
| Dimension | Score |
|---|---|
| Supply chain depth | 4.6/10 |
| Decision and optimization substance | 4.6/10 |
| Product and architecture integrity | 5.4/10 |
| Technical transparency | 4.4/10 |
| Vendor seriousness | 3.6/10 |
Aera’s substantial contribution is the machinery around a recurring decision: data preparation, model execution, policy checks, write-back, and a record of what happened. Its current product also documents versioned decision flows and the ability to observe a policy’s effect before enforcing it. Those capabilities can support improvements to reusable logic. The unresolved problem lies inside that logic: public descriptions do not establish how Aera chooses economically preferable actions under joint demand and supply uncertainty. Its supply-chain catalog still centers important calculations on service levels, safety stocks, and reorder points. (2, 3, 23, 30)
The historical promises make this gap more consequential. Aera advertised an autonomous enterprise in 2017, then cognitive automation, decision intelligence, and agentic AI. Current disclosures show a configurable engineering platform and substantial implementation and model-support work. Its standard agreement also assigns customers responsibility for validating generative-AI outputs and the risks of relying on them. Automation still depends on customers and implementers establishing the economic logic, maintaining it, and accepting responsibility for its use. (9, 14, 17, 20)
Aera versus Lokad
Aera organizes work around a configurable decision process: bring together data, choose engines, compose a flow, apply policies, and execute the result. Its platform can combine forecasts, scenarios, mathematical optimization, business rules, and agent reasoning. Aera’s engineering challenge is to keep those components and the customer’s decision policies coherent. (3, 8)
Lokad’s inventory optimization documentation describes a narrower operating model: numerical recipes produce recurring supply-chain decisions from probabilistic forecasts and customer-specific economics, with Supply Chain Scientists maintaining the logic. Demand, lead times, and other material uncertainties are inputs to deciding which actions offer the best expected economic return. The broader perspective appears in Introduction to Supply Chain and Holimization.
Consider transferring seasonal stock between two warehouses. Aera can identify an imbalance, propose a transfer, apply approval rules, and write the action into an execution system. The economic question is whether the receiving site’s possible additional sales outweigh freight, handling, and the opportunity to sell the same stock at the original site before the season ends. That requires interacting forecasts and costs. Matching both locations to target stock levels does not answer it. (23)
Lokad makes that calculation the central deliverable. Aera supplies a general platform in which a customer or implementer must establish the relevant decision logic. Both approaches require maintained data meanings, explicit commercial assumptions, and engineering work. The practical comparison is the quality of the resulting transfers and purchases, together with the labor needed to keep producing and improving them.
Aera’s version control, rollback, and policy observation address the cost of changing reusable logic. Public documentation still leaves the economic tests for accepting those changes to the implementation. (3, 30)
How Aera produces supply-chain decisions
Data, models, policies, and execution
The current architecture starts with Data Cache, which ingests selected operational data, performs transformations, and applies access controls. Decision Engines provide forecasting, machine learning, analytics, scenario modeling, optimization, agents, and visual flows. Agentic Ambient Intelligence can assemble resources for a situation, while Decision Governance checks policies and records the outcome. Write-back applies the chosen action to source systems. (1, 2, 7)
Aera’s December 2023 forecasting article gives a concrete example. Forecast changes can trigger purchase orders, transfers, or safety-stock adjustments. Customers configure review thresholds, such as the percentage change in a forecast or the value of a proposed purchase order. This establishes a route from a numerical update to an executable action. It also exposes the operating assumptions: thresholds determine when the system proceeds and when a planner intervenes. (27)
The current Skills catalog includes stock transfers, replenishment, fulfillment selection, and dynamic inventory parameters. Its Dynamic Norm Setting explicitly derives safety stocks and reorder points from service objectives. Dynamic Lead Time uses lane, carrier, and route performance to update lead-time estimates in planning systems. These functions address operational variables; they do not disclose a common stochastic economic calculation across them. (23)
A frequently refreshed estimate can still miss the decisive risk. An item with two possible receipt dates may warrant a different purchase from an item reliably arriving on their average date. Likewise, several Skills acting on the same stock or supplier capacity need a common valuation and coordinated constraints. Shared data and sequential execution do not establish that coordination.
Optimization components and missing model fidelity
Aera now identifies a built-in optimizer and optional use of CPLEX, NVIDIA cuOpt, and other tools under customers’ own licenses. Its examples include distribution, truck loading, and container utilization. Solver connectivity gives an implementer computational options, but the product page does not disclose the actual supply-chain formulations, probability models, or the interactions retained when problems become expensive to solve. (3)
The burden sits in the formulation. Order minimums, supplier price breaks, scarce shared capacity, substitution, and uncertain arrivals interact. Linear formulations simplify these relationships; retaining discrete decisions and expanding them across possible futures can make computation prohibitive. A fast solution to a reduced deterministic problem can still recommend the wrong action. Neither a solver’s reputation nor the speed of a component benchmark resolves that defect.
The same issue affects improvements after deployment. When an expensive mistake reveals an omitted interaction, the useful capability is to incorporate it into the recurring decision calculation. If implementation instead adds another exception rule or splits the problem into locally optimized pieces, the operating burden grows while the economic inconsistency survives. Aera’s public material does not show a supply-chain benchmark that tests this process.
Forecasting and learning need economic validation
Auto-Forecast combines statistical and ML methods, automated selection, external inputs, and a shared deep-learning model across products and aggregation levels. The public description does not give an item-location evaluation with a specified horizon, error definition, stockout treatment, and frozen comparison. Its separate demand-sensing descriptions mainly promise quicker responses to fresh signals. They supply no distinct mathematical method or controlled economic test. (3, 6)
Aera’s 2022 cognitive-automation patent application is more informative about learning. It describes storing recommendations, user responses, context, and subsequent observations, then ranking recommendations and using those rankings for further learning. One example compares projected demand with later sales; the disclosure also describes comparing an accepted user action with the outcome of the proposed alternative. (22)
This raises a specific measurement problem. Sales after a decision do not directly reveal what would have happened under an unexecuted alternative, particularly when stock availability, price, or allocation changed. Learning whether users accept recommendations is also different from learning which decisions make money. Aera’s current Decision Ledger improves traceability, but the public material does not establish a reliable counterfactual economic evaluation. (22, 30)
Technology history and roadmap integrity
The product’s roots are visible well before the Aera brand. A March 2012 partnership announcement describes FusionOps supplying packaged cloud dashboards, SAP integration, and self-service analytics through KloudData. Rapid deployment and reduced IT work were already central selling points. (26)
In June 2017, Aera launched from the FusionOps technology with a much larger promise: a cognitive operating system autonomously orchestrating enterprise operations. November’s Cognitive Skills announcement extended that claim to demand forecasting and manufacturing, alongside voice interaction on mobile devices. These were broad pre-LLM autonomy claims. The releases did not demonstrate the general understanding and adaptive reasoning implied by that language. (17, 18)
The 2020 Cognitive Operating System announcement continued the autonomous-operation message. In May 2021, Aera announced integration with Microsoft Azure Digital Twins, describing a real-time representation of the enterprise and improved recommendations. The announcement explains the data-modeling connection, but provides no validation of forecasts of future system states or the resulting economic decisions. A digital representation of current operations cannot, by itself, supply that predictive capability. (28, 25)
March 2022 brought the Decision Cloud name. June added identifiable engineering features: Jupyter-based Python and R integration, AutoML, model monitoring and versioning, graph exploration, and a recommendation-confidence framework. Those are substantive platform developments. They also reveal how much of the capability depended on building and maintaining application-specific models, five years after the autonomous-enterprise launch. (19, 5)
In November 2024 and June 2025, Aera introduced agentic workflows, natural-language Skill authoring, and broader user-facing interaction. Its August 2026 announcement added a decision ledger, policy tooling, conversational construction of Skills, and A2A/MCP interoperability. The current pages describe selective use of LLMs alongside deterministic computations and configurable levels of autonomy. (20, 21, 24)
The history shows sustained product development, but repeated claims of enterprise-wide autonomy have run ahead of the published demonstrations. The next release’s vocabulary cannot validate the earlier promise. What remains missing is a longitudinal account of which decision classes became economically reliable, how much maintenance they required, and how later technology changed those results.
Software materials and engineering costs
Employer-authored vacancies identify several parts of the engineering effort. A CALC-engine role seeks low-latency Java expertise, including threading, allocation, garbage collection, JVM behavior, and language or compiler design. A backend role explicitly describes Java services and APIs. An infrastructure-management vacancy covers a production platform team with Kubernetes and container experience. The standard agreement identifies Microsoft Azure as its then-current backup hosting provider. (13, 15, 12, 9)
| Component or boundary | Public evidence | Ownership implication |
|---|---|---|
| Calculation and backend services | Java/JVM calculation-engine work and Java service development | Aera maintains specialized runtime and performance engineering. |
| Cloud operations | Infrastructure/SRE team, container-platform expertise, Azure backup provision | Availability and upgrades require a continuing operations function. |
| Customer decision logic | Rules, multidimensional models, ML, optional external solvers, and agent flows | Each configured process needs compatible assumptions, testing, and change control. |
| Source-system integration | Data transformations, crawlers, write-back, and customer-installed components | Both data meanings and execution permissions must remain correct as systems change. |
Aera publishes no component-level hosting cost. The visible source of technical weight is the number of computational and operational boundaries: source systems, harmonized data, several engines, optional customer models and solvers, external agents, and write-back. Aera documents elastic ML compute and selective LLM use, which can reduce idle infrastructure and token spending. Those savings do not eliminate the cost of testing the whole decision process. (2, 3, 7)
Configuration maintenance and productivity
The current Dynamic Engagement page claims business users can build and adapt Skills without specialized technical resources. Aera’s own Sustaining Engineer vacancy describes the continuing work more concretely: diagnose customer production issues in decision models, transformations, recommendations, and business logic; validate fixes; and coordinate with engineering and modeling teams. Its Senior Engagement Principal coordinates customer delivery, specialists, consultants, and system integrators under statements of work. (8, 14, 16)
Natural-language authoring can reduce the time needed to express a change. It does not establish that the change uses the right financial assumptions, respects data semantics, or improves outcomes outside the examples used to create it. Versioned flows, branching, merging, rollback, and Policy Studio’s ability to observe proposed rules against live decisions are therefore valuable capabilities. They support disciplined maintenance of reusable logic. (3, 30)
MCP provides another connection to external tools and assistants. That is inexpensive interoperability, with no independent claim to supply-chain intelligence. If its main result is to let users ask more questions and approve more recommendations, the manual planning bottleneck survives in a conversational interface. Aera also documents unattended execution and agent-assisted Skill construction; those functions should be evaluated separately through the decisions automated and the total work removed. (8, 24)
Customer evidence provides narrower measurements. Aera’s August 2024 account quotes BAT reporting a 90% reduction in time spent on one analysis activity. Its April 2025 pharmaceutical article attributes $7 million in two-year savings to an unnamed company’s data management, lead-time tracking, process alignment, and software-cost changes, with a further $50 million presented as an expectation. These accounts describe useful opportunities, but do not measure the complete labor and ownership cost of an autonomous supply-chain operation. (29, 31)
Licensing also enters that calculation. The standard agreement permits usage limits on developers, virtual machines, Skills, and other quantities, and its user definition includes authorized bots. Agent deployment therefore does not imply unlimited automation at a fixed subscription cost. Customer-created Skills remain customer property under the agreement, while their operation depends on the Aera platform. (9)
Security boundaries and accountability
Aera documents default-deny network policies, MFA for remote access, logical data segregation, and row-level access controls before data reaches reasoning. These controls address specific routes to unauthorized access. The public descriptions do not explain isolation of customer-supplied code, dependency installation, resource limits, or protection of secrets from that code. (10, 2)
This matters because Aera introduced Python and R through Notebook in 2022, while its public Cortex description still encourages externally built models using live enterprise data. The published security document explicitly excludes Aera Developer from its scope, leaving the execution boundary around customer logic inadequately explained. (5, 4, 10)
The standard agreement specifies weekly backups with daily differentials and thirty-day retention. It does not explain whether those copies remain recoverable after compromise of powerful production or administration credentials. For agents, current pages describe permission checks and configured credentials for external tools; they do not establish how untrusted documents or tool responses are prevented from redirecting an authorized action. (9, 8)
Economic responsibility is more explicit. Sections 4.4 and 8.4 of the public agreement require customer validation of generative-AI outputs and assign the associated reliance risk to the customer, while disclaiming liability for resulting decisions or actions. These provisions concern GenAI outputs. They put a material qualification on claims that an agent can take over business decisions. A deployment must still identify who performs the required validation and how its labor is counted. (9)
Aera’s FAQ cites millions of executed decisions, and its engagement role assigns staff responsibility for delivered value. Neither establishes a contractually measured net return. The public service warranty concerns conformity with documentation, with correction or limited refund remedies; it does not guarantee profitable replenishment or allocation. The score reflects that gap between decision volume, marketing benefits, and accepted economic responsibility. (11, 16, 9)
Supply chain score
Each criterion has equal weight. Dimension scores average their five criteria; the overall score is 113/25, rounded to 4.5/10.
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 4/10 |
| Decision end-state | 7/10 |
| Conceptual sharpness on supply chain | 5/10 |
| Freedom from obsolete doctrinal centerpieces | 3/10 |
| Robustness against KPI theater | 4/10 |
Average: 4.6/10.
Economic framing — 4/10. Recommendations can display cost, revenue, and inventory consequences. Public descriptions do not establish a common monetary objective across actions competing for stock, cash, or capacity. The customer must supply much of that decision logic. (3, 8)
Decision end-state — 7/10. Write-back supports purchase orders, transfers, and other operational changes, with configurable autonomous execution. Economic usefulness depends on the calculation and policy gates preceding that action. (7, 27)
Conceptual sharpness on supply chain — 5/10. Aera identifies practical problems involving aging stock, fulfillment, transport, and lead-time updates. Its general decision framework leaves the underlying economic and uncertainty models insufficiently specified. (6, 31)
Freedom from obsolete doctrinal centerpieces — 3/10. Dynamic norm setting preserves service-level objectives and inventory buffers, while the forecasting workflow can automate safety-stock changes. More frequent recalculation does not remedy the weaknesses of these policy targets. (6, 27)
Robustness against KPI theater — 4/10. Decision records connect recommendations with actions and observed results. The public material does not establish that recorded impacts are causal net benefits, and decision counts or user acceptance cannot substitute for that evaluation. (11, 30)
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 3/10 |
| Distinctive optimization or ML substance | 4/10 |
| Real-world constraint handling | 5/10 |
| Decision automation and labor productivity | 6/10 |
| Resilience under real operational complexity | 5/10 |
Average: 4.6/10.
Probabilistic modeling depth — 3/10. Forecasting, lead-time estimates, and confidence scores are visible. Joint probability models of decision-relevant futures, calibrated uncertainty, and their economic use remain undocumented. Confidence in a recommendation is not a distribution of future demand and receipts. (3, 5)
Distinctive optimization or ML substance — 4/10. AutoML, model lifecycle tools, and a disclosed feedback architecture provide identifiable mechanisms. Optional external solvers and general references to learning algorithms do not demonstrate a distinctive supply-chain decision method. (3, 5, 22)
Real-world constraint handling — 5/10. Rules, optimization, and execution flows can impose operational restrictions. Public descriptions do not expose how economic interactions and uncertain arrivals survive the simplifications needed to solve large coupled problems. (3, 27)
Decision automation and labor productivity — 6/10. Write-back and unattended Skills can remove recurring transaction work. Conversational authoring, versioned flows, and policy observation also address maintenance productivity. Public customer measurements cover narrower activities than total operating labor. (3, 7, 29, 30)
Resilience under real operational complexity — 5/10. Decision records and model-support responsibilities provide a route from faulty recommendations to diagnosis. Versioned flows, rollback, and observation of proposed rules before enforcement support controlled revision. Economic validation remains unresolved: public evidence does not trace an omitted cost or interaction through a revised formulation, replay, and demonstrated improvement in economic outcomes. (14, 3, 30)
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 6/10 |
| System-boundary clarity | 6/10 |
| Security seriousness | 4/10 |
| Software parsimony and ownership cost | 4/10 |
| Compatibility with programmatic and agent-assisted operations | 7/10 |
Average: 5.4/10.
Architectural coherence — 6/10. Data, computational engines, policy enforcement, execution, and decision records have identifiable roles. The customer-specific coordination of several engines and external models still creates substantial integration and validation work. (1, 2, 3)
System-boundary clarity — 6/10. Data ingestion, source-system write-back, incoming user permissions, and externally authenticated tools are described. Detailed public transaction schemas, failure semantics, and execution guarantees are much thinner. (2, 7, 8)
Security seriousness — 4/10. Default-deny networking, remote-access MFA, and data-access enforcement are specific commitments. Customer-code containment and destructive-access recovery are not comparably explained, and the public security scope excludes Aera Developer. (10, 2, 5)
Software parsimony and ownership cost — 4/10. Specialist calculation engineering, cloud operations, model support, and configurable integrations all carry continuing costs. Elastic compute and selective LLM use help with resource consumption, but no complete ownership-cost evidence accompanies the rapid-implementation claims. (12, 13, 14, 3)
Compatibility with programmatic and agent-assisted operations — 7/10. Reusable flows, model integration, automated actions, and agent-assisted authoring offer several routes beyond manual screen operation. MCP connectivity alone adds no intelligence; the valuable mechanisms are executable logic and controlled revision. (3, 8, 24)
Technical transparency
| Criterion | Score |
|---|---|
| Public technical documentation | 5/10 |
| Inspectability without vendor mediation | 3/10 |
| Portability and lock-in visibility | 4/10 |
| Implementation-method transparency | 6/10 |
| Evidence density behind technical claims | 4/10 |
Average: 4.4/10.
Public technical documentation — 5/10. Current component pages explain more than a generic architecture diagram, and the patent offers a specific learning example. Mathematical formulations, calibration tests, and detailed developer references remain largely outside the public record. (3, 22, 30)
Inspectability without vendor mediation — 3/10. A reader can inspect product descriptions, contractual terms, and selected design disclosures. Reproducing an operational Skill or independently testing its numerical decisions requires further access. (4, 9, 22)
Portability and lock-in visibility — 4/10. The agreement preserves ownership of customer-developed Skills, and external models can be incorporated. Ownership of logic does not provide a documented way to execute the entire decision process independently of Aera. (9, 4)
Implementation-method transparency — 6/10. Role descriptions expose requirements gathering, delivery coordination, model debugging, impact analysis, and deployment validation. This gives a reasonably concrete picture of the specialist work beneath self-service claims. (14, 16)
Evidence density behind technical claims — 4/10. The release history and current engine descriptions identify real mechanisms. Claims about continuously improving decisions and enterprise-wide autonomy remain much broader than the published evaluations. (5, 7, 17)
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 3/10 |
| Technology roadmap integrity | 3/10 |
| Conceptual sharpness | 4/10 |
| Economic accountability | 3/10 |
| Defensibility in an agentic-software world | 5/10 |
Average: 3.6/10.
Technical seriousness of public communication — 3/10. Aera repeatedly attributes broad understanding, autonomy, and improvement to its platform without corresponding operational demonstrations. Concrete engine and governance descriptions are more useful than this anthropomorphic framing. (7, 17, 18)
Technology roadmap integrity — 3/10. Product development is visible from analytics through model tooling to agents. Yet the autonomous-enterprise promise predates the later reasoning and authoring capabilities by years, without a measured account of the earlier claim’s achieved scope. (26, 17, 20, 21)
Conceptual sharpness — 4/10. Separating data, decisions, execution, and feedback is useful. Treating heterogeneous engines, user confidence, and recorded outcomes as sufficient foundations for better decisions leaves the economic and statistical questions unresolved. (1, 5, 30)
Economic accountability — 3/10. Delivery roles and customer stories focus on value, but the public agreement supplies no net-return commitment. Its GenAI provisions explicitly leave validation and reliance risk with customers. (16, 31, 9)
Defensibility in an agentic-software world — 5/10. Executable flows, governed changes, and operational integrations provide a useful foundation. Long-term value depends on economically correct decisions with low maintenance effort; another conversational entry point or automation counter does not establish either. (3, 8, 14)
Sources
Sources were accessed on November 4, 2026. Historical releases, current product pages, contract terms, and employment descriptions are identified separately below.
Aera technology overview. Current architecture overview, redirected from the older Decision Cloud URL: data, engines, execution, governance, and selective LLM use.
Data Cache. Current data ingestion, transformation, write-back, lineage, and access-control description.
Decision Engines. Current seven-engine description: optional external solvers, automated modeling, versioned flows, branching, merging, and rollback.
Aera Cortex. Public Cortex description: externally trained models and direct integration with live datasets.
[5] June 2022 AI/MLOps and Notebook release
June 2022 AI/MLOps and Notebook release. June 6, 2022 release introducing Jupyter-based Python/R integration, AutoML, monitoring and versioning, Graph Explorer, and confidence scoring.
Aera Skills. Current packaged Skills catalog across business functions, including supply-chain inventory policies.
[7] Agentic Ambient Intelligence
Agentic Ambient Intelligence. Current autonomy settings, action write-back, decision memory, feedback, and LLM configuration.
Dynamic Engagement. Current user interaction, Skill authoring, A 2 A/MCP connectivity, and permission boundaries.
[9] Aera Master Agreement v0126
Aera Master Agreement v0126. Public master-agreement template, version 0126. Relevant provisions: GenAI validation 4.4 and risk 8.4; backups 6.3; customer-developed Skills 7.1; service warranty 8.2; usage limits 12.26 and users/bots 12.27. Individual signed terms may differ.
[10] Aera Security and Privacy Documentation
Aera Security and Privacy Documentation. Public security document: network controls, MFA, segregation, recovery provisions, shared responsibilities, and express exclusion of Aera Developer from scope.
[11] Decision Intelligence FAQ
Decision Intelligence FAQ. Current product FAQ, including claimed decision volumes and implementation approach.
[12] Infrastructure Platform Engineering manager vacancy
Infrastructure Platform Engineering manager vacancy. Employer-authored production-platform management vacancy; distinguishes responsibilities from candidate experience preferences.
[13] CALC engine senior engineer vacancy
CALC engine senior engineer vacancy. Employer-authored calculation-engine vacancy describing low-latency Java/JVM and language/compiler expertise.
[14] Modeling Support sustaining engineer vacancy
Modeling Support sustaining engineer vacancy. Employer-authored production-model support vacancy: recommendations, transformations, root-cause analysis, impact analysis, and fix validation.
[15] Backend senior engineer vacancy
Backend senior engineer vacancy. Employer-authored backend vacancy: Java services, APIs, and release engineering.
[16] Senior Engagement Principal vacancy
Senior Engagement Principal vacancy. Employer-authored delivery-lead vacancy: implementation, customer value, statements of work, and coordination with consultants and integrators.
[17] Aera launch: Self-Driving Enterprise
Aera launch: Self-Driving Enterprise. June 20, 2017 vendor-issued launch announcement, linking FusionOps technology with the autonomous-enterprise proposition.
[18] First Cognitive Skills release
First Cognitive Skills release. November 2, 2017 vendor-issued announcement covering demand/manufacturing Skills, AI, and mobile voice interaction.
[19] Aera Decision Cloud debut
Aera Decision Cloud debut. March 16, 2022 introduction of Decision Cloud and the associated development environment.
[20] 2024 Agentic AI, Workspaces, and Control Room release
2024 Agentic AI, Workspaces, and Control Room release. November 7, 2024 release of agentic workflows, Workspaces, and Control Room.
[21] 2025 people-centric agentic AI release
2025 people-centric agentic AI release. June 11, 2025 release describing natural-language Skill construction and data onboarding.
[22] Cognitive automation platform patent application
Cognitive automation platform patent application. US 20220067109 A 1, published March 3, 2022, attributed to Aera Technology; describes stored context, user responses, outcome comparisons, and learning. A design disclosure does not establish deployed performance.
[23] Supply-chain Skills catalog
Supply-chain Skills catalog. Current supply-chain catalog: norm setting, safety stocks, order execution, and dynamic transport lead-time estimates.
[24] August 2026 decision intelligence innovations
August 2026 decision intelligence innovations. August 31, 2026 announcement of decision records, policies, conversational authoring, and A 2 A/MCP interoperability.
[25] Microsoft Azure Digital Twins announcement
Microsoft Azure Digital Twins announcement. May 25, 2021 announced Azure Digital Twins integration; emphasizes current-state representation and automated decisions.
[26] KloudData and FusionOps Insight partnership
KloudData and FusionOps Insight partnership. March 17, 2012 partner-issued release, with FusionOps founder commentary, describing packaged SAP analytics and self-service dashboards.
[27] Demand forecasting and decision automation
Demand forecasting and decision automation. December 21, 2023 vendor article explaining forecast-triggered actions and manual-review thresholds.
[28] 2020 Cognitive Operating System release
2020 Cognitive Operating System release. 2020 general-availability announcement. The page header says September 24; its dateline says February 20. The review uses the year only.
[29] 2024 customer conference accounts
2024 customer conference accounts. August 2, 2024 vendor account quoting Unilever and BAT executives; BAT’s 90% figure concerns one analysis activity.
Decision Governance. Current decision records and Policy Studio, including observation of rule effects before activation.
[31] Pharmaceutical inventory operations
Pharmaceutical inventory operations. April 21, 2025 vendor article reporting an unnamed pharmaceutical deployment; distinguishes claimed two-year savings from expected additional value.