Go back to Market Research
EdgeVerve develops enterprise applications within Infosys. Its supply chain product, TradeEdge, connects manufacturers, distributors, retailers, and logistics partners to exchange and reconcile commercial data. It supports downstream sales visibility, demand sensing, order allocation, product traceability, and procurement analytics. The current website presents these applications as Infosys Topaz AI Next Supply Chain. 1
Overall supply chain score: 4.0/10
| Dimension | Score |
|---|---|
| Supply chain depth | 4.0/10 |
| Decision and optimization substance | 3.6/10 |
| Product and architecture integrity | 4.4/10 |
| Technical transparency | 4.0/10 |
| Vendor seriousness | 4.0/10 |
EdgeVerve’s strongest evidence concerns getting information across organizational boundaries and processing operational transactions. These activities can produce substantial gains: faster recalls, fewer order touches, and shorter document-processing cycles. The evidence is much weaker for selecting economically superior replenishment, allocation, or production decisions under uncertainty.
The distinction matters because the marketing moves readily from data availability to forecasting, optimization, and autonomy. Those transitions require additional machinery. A consolidated inventory position does not specify which customer should receive scarce stock. Yesterday’s sales do not determine an economical order arriving several weeks later. The public material leaves the forecasting methods, treatment of uncertain replenishment dates, and monetary decision criteria largely unexplained.
EdgeVerve versus Lokad
TradeEdge addresses the exchange and interpretation of information between trading partners. Lokad’s inventory optimization service addresses the quantities and priorities that should follow from that information. There is overlap around replenishment and allocation, but the respective centers of gravity differ.
Consider a distributor deciding whether to ship its remaining stock to a nearby retailer or retain it for another customer. Availability and proximity can identify a feasible shipment. Choosing economically requires the margin, delivery costs, contractual consequences, likelihood of alternative demand, and distribution of the next replenishment date. EdgeVerve documents the first kind of allocation criteria; it does not publish a corresponding calculation of the latter trade-offs. 4
Lokad organizes recurring decisions around expected financial contribution, using probabilistic projections and maintained decision logic. Its Supply Chain Scientists are responsible for that logic as the business changes. The approach described in Introduction to Supply Chain treats uncertainty and economic priorities as the substance of planning.
A partner network can supply data to such a calculation. Its usefulness should be assessed separately from the quality of the resulting decisions. Faster transmission of an uneconomical order merely executes the mistake sooner.
What TradeEdge computes
Demand sensing: fresher observations, unexplained forecasts
The demand-sensing page describes store-SKU daily sales arriving the following day, validation within two hours, and suggested orders using inventory, lead time, working capital, and promotions. It reports a 20% improvement in forecast accuracy when using sales rather than shipment information, across approximately 70% of a customer’s global sales. The metric, forecast horizon, evaluation period, and treatment of unavailable products are not specified. 2
This is a relative improvement claim, not a claim of 20% or 120% absolute accuracy. Its economic meaning remains unresolved. A retailer can improve tomorrow’s prediction while leaving decisions governed by next month’s uncertain deliveries almost untouched. An apparent sales shortfall may also represent unavailable inventory rather than low demand.
SAP’s own partner listing describes the Infosys demand-sensing solution as a SAP Business Technology Platform extension that converts SAP IBP weekly forecasts into daily forecasts and incorporates additional signals. 3 Daily granularity can support execution; it does not establish a new forecasting method or a better replenishment policy. The decisive test is whether the additional information changes actions at their actual commitment horizon and improves their financial consequences.
No inspected technical source provides calibrated probability distributions for demand and lead times, their dependencies, or the process by which those distributions select orders. Mentioning lead time as an input is materially different from modeling its uncertainty. The term demand sensing obscures this distinction when it stands in for a disclosed model.
Order orchestration: allocation rules, not a disclosed economic optimizer
The order-orchestration page identifies stock availability and proximity as allocation criteria and supports splitting an order among fulfillment partners. It advertises responses within two seconds and an 8–10% fill-rate improvement. 4
These are concrete execution functions. Their limitations follow from the objective: nearby stock is not necessarily the cheapest stock to consume. A shipment can preserve today’s fill rate while increasing tomorrow’s shortage cost; splitting can improve apparent service while multiplying transport and handling costs. Evaluating these choices requires consequences across orders and time, not merely a faster response to each request.
The published material supplies neither a mathematical decision objective nor an experiment comparing the allocation policy with a realistic alternative. There is insufficient evidence to attribute a particular solver, stochastic optimization method, or network-wide economic calculation to TradeEdge.
Integration and traceability are more substantiated
The integration product documents EDI and API connections, canonical data models, prebuilt connectors, and configurable workflow and exception processing. 5 Standardizing message formats and identifiers can reduce repeated integration work. The remaining difficulty is semantic: whose stock is available to promise, which timestamps are trustworthy, and whether a return, cancellation, or late correction changes an earlier decision.
Mars provides the strongest named operational example. EdgeVerve’s 2023 case reports a reduction in traceability and hold-and-release time from two days to two hours. The application combines Market Connect ingestion, parent-child stock relationships, compliance reporting, and integrated hold/release. Reusable mappings and automated regression tests supported deployment; the first proof of concept took 12 weeks. A named Mars executive confirms the integration improvement. 7
The result concerns identifying and containing affected products. Its value can be substantial even without a forecasting breakthrough. The case does not supply the deployment budget, recurring operating cost, or realized reduction in recall losses, so a full financial return cannot be reconstructed.
Customer outcomes and labor productivity
The 2024 G-Cloud service definition includes an anonymous consumer-goods case reporting an 80% reduction in manual order touches, alongside lower idle inventory and improved fill rate. Another example is explicitly a ten-week proof of concept, with a forecast-accuracy improvement of up to 15%. These have different evidential weight: a pilot forecast metric does not establish a sustained operating result. 6
An April 2026 Infosys account describes a shipping company using AI Next for bookings, bills of lading, and freight auditing and invoicing. It reports process automation rising from 10% to 70%, straight-through processing from zero to 70%, and freight-audit/invoice turnaround falling from 24 hours to 30 minutes. The company is unnamed. The application coordinated more than 90 interfaces across over 20 systems, digitized over 8,000 customer policies, and incorporated procedures for 116 countries. 8
Those figures describe substantial workflow engineering. They do not measure autonomous inventory optimization. Nor does the report quantify net labor after implementation, exception handling, policy maintenance, and platform operation. Turnaround time can fall dramatically while considerable specialist effort remains elsewhere.
For an agent-era productivity assessment, the relevant unit is the entire recurring workload. The economic gain is the labor and errors removed after subtracting the work needed to keep the automation correct. A shorter queue, more transactions per employee, and a smaller team are different outcomes; the published cases do not consistently distinguish them.
Configuration maintenance and ownership cost
TradeEdge’s 2024 service definition includes an 8×5 managed-services team for distributor onboarding, issue resolution, and follow-up to obtain data on time. It describes configurable data-cleansing rules and regular reporting into client planning and analytics systems. 6 This identifies work that survives deployment. Moving it to a supplier changes the invoice and staffing arrangement; it does not eliminate the activity.
For a network product, configuration accumulates at each boundary: item mappings, partner identities, message variants, validation rules, exception ownership, and contractual definitions. A distributor changing its ERP can invalidate an otherwise stable connection. A new promotion can require more than adding a data field if its commercial meaning differs across channels.
The shipping implementation makes the scale of policy maintenance particularly visible. Every retained business rule needs an owner, a test, and a retirement condition. Otherwise automated processing preserves obsolete decisions more efficiently than the previous manual process.
This is the practical issue addressed by Lokad’s Holimization: the economic problem itself must remain open to revision. Faster execution cannot compensate for rules that encode yesterday’s operating assumptions. EdgeVerve’s public cases describe the initial transformation more clearly than the ongoing cost of revising that transformation.
Technology history and delivery
| Period | Public claim or development | What the record establishes |
|---|---|---|
| 2008–2015 | TradeEdge began with a customer request for downstream sales and inventory visibility; EdgeVerve subsequently became an Infosys subsidiary. | The product originated in a concrete integration problem. 9 |
| 2017 | Infosys presented Nia as an AI platform capturing organizational knowledge and transforming processes, with data, ontology, and automation components. | The pre-LLM AI narrative combined identifiable technologies with much broader claims of organizational intelligence. 10 |
| July 2017 | Infosys told investors it was integrating its blockchain network with TradeEdge to create monetization opportunities. | A specific technological promise was made. The inspected current TradeEdge material does not identify the delivered blockchain component or attributable customer economics. 11 |
| 2019 | Finacle and R3 reported a trade-finance blockchain trial involving 18 banking groups. | This establishes activity in banking, a distinct application from TradeEdge distribution planning. 12 |
| 2022–2023 | TradeEdge articles grouped AI, machine learning, blockchain, automation, and digital twins as supply chain transformation technologies. | The articles supplied neither comparative algorithm evaluations nor predictive validation of the proposed twins. 13 14 |
| 2025–2026 | EdgeVerve added an agent-governance narrative; Infosys subsequently subsumed the former EdgeVerve platforms under Topaz AI Next. | The branding changed, while concrete automation cases continued. A common brand does not establish a common decision model. 15 17 |
The historical weakness is the distance between the announced technology and the supplied proof. Blockchain was presented as an imminent addition to an existing network. Years later, the accessible TradeEdge evidence still explains integration, data harmonization, and execution benefits much more clearly than blockchain’s contribution.
The digital-twin promotion has a similar problem. A predictive twin makes claims about a large collection of future states. Its credibility depends on validation across those states and on the decisions it improves. Presenting the twin as a transformation technology without discussing forecast error leaves the central technical problem unexplained.
Early AI claims require the same discipline. Ontologies, predictive models, and workflow automation were real software techniques before 2023. They did not justify the implication that enterprise knowledge and reasoning had been comprehensively automated. The later agent narrative should be assessed through deployed responsibilities and measured outcomes, not treated as retrospective validation of the earlier vocabulary.
Bill of software materials
The most specific primary-source stack description is the supplier’s G-Cloud 14 listing. Its dated requirements identify the following components. 16
| Component | Documented scope |
|---|---|
| PostgreSQL 12 and PgBouncer 1.5.5 or 1.12 | TradeEdge database and connection-pooling requirements |
| .NET Framework 3.5/4.5 and IIS 7 | Microsoft runtime and web-server requirements |
| Java 1.8.191/192 and WinSCP | Additional runtime and transfer dependencies |
| Excel 2016 | Client-side spreadsheet dependency |
This is a 2024 procurement disclosure. It does not identify the versions operating in November 2026. Its age makes a current component inventory, supported-version policy, and upgrade history material omissions.
The mixture of runtimes entails distinct patching, testing, and deployment responsibilities. The listing does not map these components to application functions or explain who maintains compatibility during upgrades. Those responsibilities determine how much integration and testing work customers absorb.
The architecture’s least visible boundary is between the established supply chain applications and the new agent platform. A shared interface can conceal separate data models and operating teams. Public diagrams and release material should explain how business rules, access rights, and changes propagate across that boundary.
Security and agent authority
The G-Cloud listing specifies SAML/OpenID Connect, CyberArk, AES-256 encryption, and TLS 1.2. Its certification section also refers to AssistEdge inside the TradeEdge listing, making product scope inconsistent. 16 Actual identity and encryption mechanisms matter; a list of certifications cannot resolve the ambiguity about which product implements which control.
EdgeVerve’s 2025 governance paper recommends bounded tool access, audit trails, approvals, sandbox environments, reversible actions, and emergency overrides. These are recommendations for enterprises, not a deployment specification demonstrating that every AI Next agent receives those restrictions. 17
The current agent product describes orchestration and lifecycle management, with centralized visibility and human oversight. 18 It leaves important execution details undisclosed: the permissions available to each tool call, credential isolation, outbound network restrictions, and the treatment of instructions embedded in documents received from trading partners.
These boundaries are central to a product that reads external documents and can act on enterprise systems. A malicious instruction inside an invoice must remain data; it must not become authority to change a supplier account or export records. Human approval of selected outputs does not replace isolation of the execution environment.
Operational autonomy should also be measured at the right level. Requiring a person to approve exceptional financial commitments can be sensible. Requiring routine review of every order would preserve the labor bottleneck the automation is supposed to remove. The public material does not provide a supply chain production measure of unattended decisions together with their error costs and supervision workload.
Economic accountability
The public G-Cloud package includes a generic master-services agreement marked as a draft for discussion. Its fixed-price deliverable warranty concerns specification conformity for 30 days, with repair or reperformance as remedies. Other provisions disclaim assurance that business requirements will be met and exclude categories including lost profit and anticipated savings. These are proposed contract terms, not evidence of the agreement governing a named customer. 19
The pricing document offers time-and-materials work, where the customer finalizes scope and manages deliverables, and fixed-price work whose scope changes can change the price. It also explicitly offers business-outcome or gainshare pricing tied to increased profit or reduced cost. 20
Gainshare is a meaningful alternative to billing entirely for activity. The public cases do not show which customers selected it, how the baseline was defined, or how maintenance costs and adverse outcomes entered the calculation. A fee based on tickets resolved or transactions processed remains payment for output volume, not necessarily for economic improvement.
The accountability gap is therefore specific: published commercial options can share value, but the demonstrated customer results are not accompanied by enough measurement or contractual detail to establish who bears the risk when the transformation is technically delivered and economically disappointing.
Supply chain score
Supply chain depth
| Criterion | Score |
|---|---|
| Economic framing | 4/10 |
| Decision end-state | 6/10 |
| Conceptual sharpness on supply chain | 4/10 |
| Freedom from obsolete doctrinal centerpieces | 3/10 |
| Robustness against KPI theater | 3/10 |
Average: 4.0/10
Economic framing — 4/10. Inventory, sales, service, and processing costs appear in the commercial argument. The connection from these outcomes to a single monetary decision criterion is missing. Integration savings and inventory returns need separate accounting. 1 2 4 6
Decision end-state — 6/10. Order allocation and hold/release produce operational actions. Demand-sensing outputs and analytics are less explicit about the final replenishment decisions and the responsibility for executing them. 2 4 7
Conceptual sharpness on supply chain — 4/10. Partner-data fragmentation is understood concretely. Uncertainty, competing claims on inventory, and the difference between observed sales and underlying demand receive much less technical treatment. 2 5 6
Freedom from obsolete doctrinal centerpieces — 3/10. The emphasis remains on visibility, faster forecasts, and connected workflows. The public design does not establish a replacement for target-driven planning with economically ranked decisions under uncertainty. 2 13
Robustness against KPI theater — 3/10. Traceability time and processing time have intelligible operational meanings. Forecast-improvement and fill-rate figures lack enough evaluation detail to establish financial value or comparability. 2 4 7 8
Decision and optimization substance
| Criterion | Score |
|---|---|
| Probabilistic modeling depth | 2/10 |
| Distinctive optimization or ML substance | 3/10 |
| Real-world constraint handling | 5/10 |
| Decision automation and labor productivity | 5/10 |
| Resilience under real operational complexity | 3/10 |
Average: 3.6/10
Probabilistic modeling depth — 2/10. Neither joint demand and lead-time distributions nor calibration methods are disclosed. Additional signals and more frequent updates do not fill that gap. 2 3
Distinctive optimization or ML substance — 3/10. Data harmonization and document processing have identifiable applications. The evidence does not expose a distinctive supply chain optimization algorithm, objective, or reproducible evaluation. 2 5 8
Real-world constraint handling — 5/10. Partner availability, order splitting, traceability relationships, and local processing rules are represented. Their coexistence in workflows does not establish an economically coherent treatment of shared resources. 4 7 8
Decision automation and labor productivity — 5/10. The cases report material reductions in transaction handling. Net recurring labor, including supplier-operated services and configuration upkeep, is insufficiently measured. 6 7 8
Resilience under real operational complexity — 3/10. The software handles heterogeneous systems and documents. Public evidence is weaker on selecting decisions during correlated shortages and on reformulating and retesting the rules when an operational failure exposes a mistaken business assumption. Workflow execution alone does not establish either capability. 4 6 7 8
Product and architecture integrity
| Criterion | Score |
|---|---|
| Architectural coherence | 4/10 |
| System-boundary clarity | 6/10 |
| Security seriousness | 4/10 |
| Software parsimony and ownership cost | 3/10 |
| Compatibility with programmatic and agent-assisted operations | 5/10 |
Average: 4.4/10
Architectural coherence — 4/10. Canonical integration models and reusable mappings give the network product a coherent purpose. The relationship between that architecture and the newer cross-product agent layer remains thinly documented. 5 6 18
System-boundary clarity — 6/10. TradeEdge’s role between trading partners and existing enterprise systems is reasonably clear. The boundary between supplying a better forecast and owning the resulting inventory decision is less clear. 2 5
Security seriousness — 4/10. Published identity, credential, and encryption mechanisms are more substantial than compliance badges. Agent-specific confinement and the product scope of several assurances remain unresolved. 16 17 18
Software parsimony and ownership cost — 3/10. The dated component inventory and explicit managed-services model expose continuing operational responsibilities. No public lifecycle cost reconciles hosting, upgrades, mappings, and rule maintenance with claimed automation savings. 6 16 20
Compatibility with programmatic and agent-assisted operations — 5/10. APIs, data exchange, and orchestration support automated integration. Public evidence does not yet establish safely unattended supply chain decisions at scale. 5 18
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 | 5/10 |
| Evidence density behind technical claims | 3/10 |
Average: 4.0/10
Public technical documentation — 5/10. Procurement documents expose more operational detail than the marketing site. Current algorithm specifications, component versions, and agent permission models remain sparse. 6 16 18
Inspectability without vendor mediation — 3/10. A reader can inspect product descriptions and case studies, but cannot reproduce the forecast or allocation claims from released code, datasets, or evaluation protocols. 2 3 4
Portability and lock-in visibility — 4/10. Standard interfaces reduce transport-level dependence. The G-Cloud exit provision leaves data transfer to mutually agreed processes; preservation of executable mappings and business rules is not specified. 16
Implementation-method transparency — 5/10. The cases identify integration, validation, configuration, testing, and deployment work. The recurring effort needed after rollout is much less measurable. 6 7 8
Evidence density behind technical claims — 3/10. Named operational testimony supports some integration results. Broad AI, optimization, and forecasting claims exceed the published method and measurement detail. 2 7 18
Vendor seriousness
| Criterion | Score |
|---|---|
| Technical seriousness of public communication | 4/10 |
| Technology roadmap integrity | 4/10 |
| Conceptual sharpness | 4/10 |
| Economic accountability | 4/10 |
| Defensibility in an agentic-software world | 4/10 |
Average: 4.0/10
Technical seriousness of public communication — 4/10. Concrete integration work is repeatedly surrounded by broader intelligence claims. The resulting ambiguity makes it difficult to distinguish a new computational capability from a new presentation of an established workflow. 1 13 18
Technology roadmap integrity — 4/10. Traceability and processing applications show delivery over time. The TradeEdge blockchain promise and later digital-twin promotion lack comparably traceable technical and economic follow-through. 7 10 11 13 14
Conceptual sharpness — 4/10. Data access, forecasting, optimization, and autonomy are frequently treated as successive benefits of the same platform. Each requires separate evidence, especially where observations must become financial commitments. 2 4 18
Economic accountability — 4/10. The published gainshare option can align incentives. The available customer cases do not establish the applicable baseline, full recurring costs, or allocation of downside risk. 7 8 19 20
Defensibility in an agentic-software world — 4/10. Established partner connections and their maintained semantics can retain value. Generic document processing and workflow orchestration face stronger substitution pressure; durable differentiation requires measured decision quality and lower total labor. 5 6 8 18
Sources
Sources were checked on November 4, 2026. Customer outcomes below are attributed to the publishing vendor unless otherwise stated. The G-Cloud 14 documents describe the 2024 offering.
[1] TradeEdge / Infosys Topaz AI Next Supply Chain
TradeEdge / Infosys Topaz AI Next Supply Chain, Undated. Current product scope and branding.
Demand Sensing, Undated. Data frequency, suggested-order inputs, and the undefined relative forecast-improvement claim.
[3] Infosys demand sensing solution: SAP partner listing
Infosys demand sensing solution: SAP partner listing, Undated. SAP’s description of the IBP extension and weekly-to-daily forecasting.
Order Orchestration, Undated. Documented allocation criteria and response/fill-rate claims.
Integration Solution, Undated. Interfaces, canonical models, and integration functions.
[6] TradeEdge G-Cloud 14 service definition
TradeEdge G-Cloud 14 service definition, 2024-05-03. Implementation, managed services, and case examples; publication date appears inside the PDF.
[7] Mars Slashes Product Traceability Time
Mars Slashes Product Traceability Time, 2023. Named Mars traceability application and customer testimony.
[8] How a unified AI platform can boost turnaround times for shipping companies
How a unified AI platform can boost turnaround times for shipping companies, 2026-04-16. Anonymous shipping implementation, operational scope, and reported automation outcomes.
[9] EdgeVerve Annual Report 2014–15
EdgeVerve Annual Report 2014–15, 2015. Product origins and corporate formation.
Infosys Form 20-F 2017, 2017. Contemporaneous description of Nia’s AI, knowledge, and automation components.
[11] Infosys Q1 FY2018 earnings call
Infosys Q1 FY2018 earnings call, 2017-07-14. Contemporaneous announcement of blockchain integration with TradeEdge.
[12] Finacle and R3 conclude blockchain trade-finance trial
Finacle and R3 conclude blockchain trade-finance trial, 2019-09-23. Issuer release concerning the separate banking application.
[13] The role of modern technologies in shaping the future of supply chains
The role of modern technologies in shaping the future of supply chains, 2022-12-09. Pre-2023 technology narrative for TradeEdge.
[14] 5 Supply Chain Trends in 2023
5 Supply Chain Trends in 2023, 2023-10-12. Digital-twin and automation promotion.
[15] Infosys ESG Report 2025–26
Infosys ESG Report 2025–26, 2026. Disclosure of the transition to Topaz AI Next branding.
[16] TradeEdge Platform: G-Cloud 14 listing
TradeEdge Platform: G-Cloud 14 listing, 2024. Historical requirements, controls, service boundaries, and exit provisions.
[17] The Platform Imperative: Balancing Autonomy, Governance and Innovation
The Platform Imperative: Balancing Autonomy, Governance and Innovation, 2025. Agent-governance recommendations, distinct from an installed-control specification.
[18] Infosys Topaz AI Next Agentic AI
Infosys Topaz AI Next Agentic AI, Undated. Current agent-platform scope and lifecycle claims.
[19] G-Cloud 14 attached draft master services agreement
G-Cloud 14 attached draft master services agreement, 2024-05-02. Generic draft agreement published with the procurement offer; sections 13–14 concern warranties and liability.
[20] G-Cloud 14 pricing document
G-Cloud 14 pricing document, 2024-05-03. Commercial models, including time-and-materials, fixed-price, and gainshare.