Full diagnostic. AI Radar 5.0 sample: Midea Group, a corporate radar. Illustrative inference from public sources, AI Radar working paper v5.0 (Sawhney, November 2026), Section 8.2.
A sample Radar from the public record
Public sources as cited in the AI Radar working paper v5.0, Section 8.2; illustrative inference. The case scores the three businesses with the richest public record and rolls them up. Weights come from revenue and an illustrative posture for each business. The public record names programs without their cost, so the Maturity Index is computed by count.
Summary
Midea Group is a diversified manufacturer whose businesses run on different value chains, so one polygon cannot describe the company. The corporate radar scores three businesses and rolls them up. Each business concentrates somewhere different, and because Smart Home Solutions carries four fifths of the weight, the corporate polygon follows its shape. For such a company the parent's decisions are the split of AI investment among the businesses and the layer at which each capability is built. Eleven of the twelve programs in the public record are at Stage 3 or above.
This sample is read from public sources only. The record names programs without their cost, so the Maturity Index is computed by count, and the indices that need spend are not read.
1. The company and its businesses
Midea Group is a diversified manufacturer with 2025 revenue of RMB 458.5 billion. Smart Home Solutions sells home appliances and earned RMB 299.9 billion. Its commercial and industrial businesses include Building Technology (commercial air conditioning, elevators, and building controls, RMB 35.8 billion) and Robotics and Automation (KUKA and Swisslog, RMB 31.0 billion) (Midea Group 2026). The businesses run on different value chains, so one polygon cannot describe the company. The case scores these three businesses, which have the richest public record, and rolls them up.
The parent states four priorities: leadership in technology, direct relationships with users, operations driven by digitization and intelligence, and global reach (Midea Group 2026). The commercial businesses grew faster than the consumer business in 2025, and Robotics and Automation is investing in AI and embodied intelligence as frontier technologies. As an illustrative reading, the case assigns a growth posture to Smart Home Solutions and Building Technology and a transformation posture to Robotics and Automation. Revenue multiplied by the posture multiplier gives weights of 0.80, 0.09, and 0.11.
2. Three business radars and the corporate roll-up
Table 1. Three business radars and the corporate roll-up
| Dimension | Smart Home Solutions | Building Technology | Robotics and Automation | Corporate | Corporate tier |
|---|---|---|---|---|---|
| Marketing | 4.00 | 2.52 | 2.52 | 3.70 | Wedge |
| Sales | 4.00 | 4.00 | 3.30 | 3.92 | Selective |
| Pricing | 2.88 | 3.00 | 2.62 | 2.87 | Drop |
| Discovery | 3.63 | 3.00 | 3.63 | 3.57 | Wedge |
| Development | 4.00 | 4.00 | 4.64 | 4.07 | Selective |
| Embedded Intelligence | 4.31 | 4.64 | 4.64 | 4.38 | Concentrate |
| Supply Chain | 4.22 | 3.30 | 3.00 | 4.00 | Selective |
| Operations | 4.64 | 4.00 | 3.63 | 4.47 | Concentrate |
| Customer Service | 4.00 | 4.64 | 4.00 | 4.06 | Selective |
| People | inherits | inherits | inherits | 3.63 | Wedge |
| Finance | inherits | inherits | inherits | 3.30 | Wedge |
| Risk | inherits | inherits | inherits | 3.00 | Wedge |
| Weight | 0.80 | 0.09 | 0.11 | 1.00 |
The Enabling dimensions are scored once for the company. Smart Home Solutions concentrates on Operations, Embedded Intelligence, and Supply Chain. Building Technology concentrates on Embedded Intelligence and Customer Service, because its stated transformation is from equipment supplier to operator and maintainer of buildings. Robotics and Automation concentrates on Development and Embedded Intelligence. The corporate polygon resembles the Smart Home polygon, because that business carries four fifths of the weight. For such a company the parent's decisions are the split and the layer. Marketing shows why: it is selective for the consumer business and a drop for the other two, and the instruction is given business by business.
3. The program ledger, by layer and stage
Table 2. The program ledger, by layer and stage
| Program | Dimension | Layer | Stage | Mode | Evidence |
|---|---|---|---|---|---|
| Factory agent system: a central scheduler with 14 agents across 38 plant scenarios | Operations | Shared | 5 | Run | Built at one washing machine plant; rolled out to other plants and offered to other manufacturers (Midea Group 2026; KrASIA 2026) |
| Overseas agent factory solution, Thailand | Operations | Shared | 4 | Run | Order lead time down 43 percent and complaint rate down 32 percent at the site (AI Magazine 2026) |
| Cross-border supply chain agent | Supply Chain | Shared | 3 | Assist | Monitors 35 nodes; no outcome reported (AI Magazine 2026) |
| Production-and-sales agents | Supply Chain | Business | 4 | Run | Domestic inventory turnover efficiency up 4.4 percent in the first half of 2026 (Futu News 2026) |
| Quotation-to-shipment planning | Supply Chain | Business | 4 | Run | Efficiency gain above 60 percent reported (Futu News 2026) |
| Home AI agent on the company's own large model | Embedded Intelligence | Business | 4 | Invent | More than 20 million users; 95 percent intent recognition (Midea Group 2026) |
| Membership analysis | Marketing | Business | 3 | Assist | In use; no outcome reported (Midea Group 2026) |
| Service capacity forecasting and dispatch | Customer Service | Business | 3 | Run | Punctuality gain reported without a figure (Midea Group 2026) |
| Overseas voice agents | Customer Service | Business | 3 | Run | Live in 22 countries; 18 percent of interactions automated or by text (Midea Group 2026) |
| AI store design tool | Sales | Business | 3 | Assist | In use in stores (Midea Group 2026) |
| KUKA AI product family | Embedded Intelligence | Business | 3 | Invent | AI vision and algorithms in robot commissioning and maintenance (Midea Group 2026) |
| Humanoid robots in plants | Operations | Corporate | 2 | Run | Built by the group research institute; a plant pilot with a 30 percent changeover target (Interesting Engineering 2025) |
Eleven of the twelve programs are at Stage 3 or above, a Maturity Index of 92 by count, with the same caution as for any public record. Four programs in Operations and Supply Chain, the concentrate dimensions of the largest business, report measured outcomes.
4. The layer reading
The layer reading is the reason for the case. The factory agent system shows replication in practice: a capability built in one plant of one business reached production, moved to the shared layer, and now runs in other plants at home and abroad. The annotation still applies to the next step, because the record shows the system in the plants of Smart Home Solutions and is silent on the other two businesses. The two corporate builds pass the layer tests: the large model and the humanoid robots serve several businesses, and neither is the advantage of one. The service programs raise a question for consolidation: domestic dispatch and the overseas voice agents do the same job in two organizations. And the ledger names ten programs from Smart Home Solutions, one from Robotics and Automation, and none from Building Technology. A count of programs is not a measure of spend. With that caution, the two commercial businesses carry a fifth of the weight and little of the disclosed AI work, and the drift between businesses is the question the Alignment Index would put to the parent.
5. What a public record cannot show
Sample Radars for real companies use public sources only, so they show the recommended allocation and the stage of each program. They do not show the indices that need spend data. Companies rarely disclose AI spend by business, so the split of investment among the businesses, the Investment and Alignment Indices, and the Layer Index in currency cannot be computed from the record. A Radar built on the company's own documents and program list is more precise.