---
title: "Inditex and AI: coordinating the business, reshaping the experience"
description: "A dossier on Inditex's AI strategy: decisions, capabilities, applications, risks and scenarios, with public sources and clearly identified independent analysis."
lang: en
date: 2026-11-08
date_modified: 2026-11-08
---

# Inditex and AI: coordinating the business, reshaping the experience

**Our interpretation.** Our hypothesis is that AI could strengthen two valuable capabilities at Inditex: coordinating decisions across an integrated commercial system and presenting its offering visually in new ways. The latter already has concrete public examples. The former is supported by indications of technical capability and intent, but still needs public evidence of operational use and impact. The available evidence does not allow us to calculate a standalone return on AI or describe an autonomous fashion company. An alternative explanation remains open: what we are seeing may primarily be defensive efficiency and wider internal adoption than the company describes.

*Research cutoff and access date: 6 November 2026. Labels distinguish reported facts, company statements, announced targets, our interpretation and our own scenarios.*

## A commercial machine that must make decisions quickly

**Documented fact.** Inditex ended FY2025, which closed on 31 January 2026, with €39,864 million in sales, approximately €10,700 million of which came from online, and 5,460 stores. Online sales form part of the total. They are not additional. Sales reached €19,755 million in the first half of the following financial year. These are consolidated business figures, not results attributable to a particular technology. [Inditex, FY2025 results, “Strong sales growth” and “Very strong execution”](https://www.inditex.com/itxcomweb/se/en/press/news-detail/b870d5ec-6b7e-491d-b38e-340cd69036df/fy2025-results) · [CNMV, H1 2026 financial report, income statement, p. 3](https://www.cnmv.es/webservices/verdocumento/ver?e=x2nr8h8c2xhhP723fHLUSFzWKkWgDybUdOsLCgb8yTD1xVx9TYwe6nqlQTLZolJy).

{{ include_html("snippets/analisis-de-empresas/inditex/escala.html", anim_shell="on") }}

**Our interpretation.** Size can magnify a small improvement, but it also multiplies the cost of a mistake. A commercial decision connects the choice, production, availability and presentation of a garment with customer responses across different places and channels. A poor forecast can leave products unsold, and a late replenishment can lose a sale that store traffic had already brought within reach. Economic value depends on the whole chain, not just on whether a screen looks smarter.

**Company statement.** Inditex describes an integrated model with proximity sourcing during the season and centralised logistics. FY2025 sales grew 3.2% in euros and 7.0% at constant currency. The difference illustrates the importance of international exposure, although those two figures alone do not identify a single cause for the gap. [Inditex, FY2025 results, “Strategic initiatives” and “Strong sales growth”](https://www.inditex.com/itxcomweb/se/en/press/news-detail/b870d5ec-6b7e-491d-b38e-340cd69036df/fy2025-results).

**Our interpretation.** This starting point shapes the thesis. For a company whose system already attempts to connect supply and demand, AI can complement existing decisions without inventing its rapid-response model. Up-to-date data is not the same as an accurate forecast. An accurate forecast does not create the physical capacity to move a garment. The potential advantage arises when information, commercial judgment and execution reinforce one another. Attributing all that agility to a statistical model would erase the infrastructure and work that make a recommendation actionable.

## Where the money goes, and what we do not know about returns

**Documented fact.** Reported FY2025 capital expenditure was €2,712 million, including extraordinary investment. The company had allocated €900 million a year in FY2024 and FY2025 to an extraordinary logistics expansion programme. That allocation sits within the reported investment context: it should not automatically be added to total capex or treated as AI spending. [Inditex, FY2025 results, cash table and “Outlook”](https://www.inditex.com/itxcomweb/se/en/press/news-detail/b870d5ec-6b7e-491d-b38e-340cd69036df/fy2025-results).

**Announced target.** For FY2026, the company estimated approximately €2,300 million in ordinary capex and close to €200 million in extraordinary investment in corporate facilities. [Inditex, H1 2026 results, “Outlook”](https://www.inditex.com/itxcomweb/km/en/press/news-detail/a2c9b82b-1e27-4582-a7c9-746bc88b744e/interim-half-year-2026-results).

**Documented fact.** For H1, Inditex reported €1,307 million in capex, while its half-year report recorded €1,541 million in asset additions, principally in stores, logistics and corporate assets. These are different measures: accounting additions do not automatically equal cash capex. [Inditex, H1 2026 results, cash table and “Outlook”](https://www.inditex.com/itxcomweb/km/en/press/news-detail/a2c9b82b-1e27-4582-a7c9-746bc88b744e/interim-half-year-2026-results) · [CNMV, H1 2026 financial report, note 5, p. 11](https://www.cnmv.es/webservices/verdocumento/ver?e=x2nr8h8c2xhhP723fHLUSFzWKkWgDybUdOsLCgb8yTD1xVx9TYwe6nqlQTLZolJy).

{{ include_html("snippets/analisis-de-empresas/inditex/inversion.html", anim_shell="on") }}

**Our interpretation.** These decisions show resources committed to general infrastructure, but not how much Inditex buys in AI computing, licences, staff or services. Nor do we know recurring expenditure, investment executed exclusively for AI, or net savings after integrating, supervising and maintaining systems. A large technology investment figure is not an AI budget. An improvement in sales or margin is not its return. Measuring that return would require a baseline, a comparison group and full costs, as well as identifying which intervention changed each outcome.

**Our interpretation.** Capital remains a constraint even if the technology works. A system that detects demand without the capacity to supply it can produce useful information without fulfilling its entire promise. Conversely, a logistics facility can be profitable for operational reasons that predate any AI model. Economic discipline means asking what additional decision each investment enables and what alternative use the capital could have had.

## The customer sees an image, not a promise of fit

**Company statement.** Zara Try-on uses AI to generate a synthetic avatar from customer photographs and shows it wearing real products. Inditex dates its launch to mid-December and described it in its FY2025 results. The source does not establish that it predicts the physical fit of a size or provide a comparison of conversion or returns. [Inditex, FY2025 results, “Customer experience”](https://www.inditex.com/itxcomweb/se/en/press/news-detail/b870d5ec-6b7e-491d-b38e-340cd69036df/fy2025-results).

**Our interpretation.** The relevant change is from viewing a catalogue image to seeing a personal representation wearing the garment. The decision remains with the customer. The tool may make exploration easier or increase visual confidence, but a convincing likeness does not guarantee that a seam, fabric or size will behave the same way on the body. The distinction matters commercially: reducing uncertainty about appearance and reducing fit-related returns are different outcomes, with different metrics and errors.

**Our interpretation.** Evaluating the product would require comparing similar customers and purchases, measuring adoption, conversion and returns by reason, and checking whether any effect persists after costs and changes in product mix. Even an increase in conversion could shift purchases that would otherwise have happened in another channel. Without looking across the whole business, redistribution can be mistaken for incremental growth. We also do not know whether people who choose to use the avatar were already buying differently before they used it.

## Photography is production, too

**Statement reported by Reuters.** In December 2025, Inditex confirmed the use of AI to create new clothing variations in images of real models, with their agreement and compensation. Reuters reported industry concerns about future commissions, although it did not measure net job losses. [Reuters, Helen Reid, 18 December 2025, opening paragraphs and industry response](https://www.investing.com/news/stock-market-news/zara-turns-ai-generate-fashion-imagery-using-reallife-models-4414956).

**Our interpretation.** The potential change goes beyond producing digital variations from a photograph. A shoot, a licence, creative direction and approval can be reorganised at different stages: producing variations after an image is captured could make some assets cheaper and shift work towards selection, editing and rights management. It could also increase the volume of assets without proportionately reducing shoots. Neither trajectory follows from a statement that the technology is complementary. Assessing them would require observing commissions, remuneration, licence duration and access for new professionals over several cycles.

**Another company's statement.** H&M presented a campaign using digital twins and an identified creative team in July 2025. It is a useful reference for the promotional-imagery process, but one campaign does not make it possible to compare the companies' scale, costs or employment effects. [H&M, “H&M continues its exploration of creativity with AI”, 2 July 2025](https://hmgroup.com/news/hm-continues-its-exploration-of-creativity-with-ai/).

**Our interpretation.** If generation tools become widespread, access to them will be less defensible than control over a combination that is hard to copy: commercial identity, access to product, image rights, editorial judgment and distribution. The division of value, however, will depend on contract terms. A model could be paid for a specific use while losing future opportunities if a licence permits many variations. The outcome depends on terms and volumes that are not public here.

## Data for coordination: declared capability and proven decisions

**Company statement.** The website for Inditex's data and technology teams describes updates to supply-chain and store data, analytics and machine-learning models. It mentions predictive models and applications such as optimisation, recommendations and personalisation. The passages reviewed do not identify a specific forecasting, purchasing, allocation or replenishment model with measured results. [Inditex, Data & Tech Teams, “How we work” and “Our technologies”](https://xtend.inditex.com/tech-teams).

**Our interpretation.** That disclosure is enough to take an ambition beyond the digital shopfront seriously, but not to describe what part of replenishment an algorithm decides. Before a recommendation, someone defines what is measured and which errors matter. Afterwards, a person or system must accept, modify or reject it and deal with the consequences. A model can suggest quantities for a store without having authority to buy them. It may not be present in that process at all. The public information reviewed does not resolve which case applies to Inditex.

**Our interpretation.** It helps to separate three layers: stock data, an estimate of what might sell, and the decision to move or produce merchandise. Tags and inventory visibility help with the first, but they do not by themselves demonstrate the other two or establish a causal effect on margin. Strong operational evidence would describe the previous process, the intervention, the locations and periods affected, the rate at which recommendations were accepted, the errors, and a comparison with decisions made without the system.

**Company statement.** In 2021 Inditex described an automatic task-rotation system in distribution centres based on Orquest and AI, intended to support ergonomics, health and time optimisation. Its 2024 sustainability report mentions SmartPlanner in the context of workforce planning, without explicitly identifying it as AI in that passage. The documents do not establish that these are the same system or that the first remained in use in 2026. [Inditex, 2021 annual report, printed p. 202](https://static.inditex.com/annual_report_2021/en/documents/annual_report_2021.pdf) · [Inditex, 2024 non-financial information, printed p. 236](https://static.pullandbear.net/2/static2/itxwebstandard/legal/pdf/Sustainabilityreport2024_en.pdf).

**Our interpretation.** Rotation illustrates a decision different from sales forecasting: allocating tasks directly affects people's working days. The stated ergonomic objective may be valuable, but assessing it would require knowing the variables, coverage, outcomes and whether an allocation can actually be corrected or challenged. The absence of those details in the passages consulted does not prove that controls are missing, although it prevents us from judging how they work. Nor do we know what employment changes the system caused.

{{ include_html("snippets/analisis-de-empresas/inditex/evidencia.html", anim_shell="on") }}

## Building, buying and learning beyond the core business

**Company statement.** At the July 2026 shareholders' meeting, the chief executive said AI was integrated across the organisation and supported productivity, operational efficiency, technological development and the customer experience. This is a qualitative statement of scope, not a public inventory that verifies universal adoption or autonomy in every process. [CNMV, Inditex shareholders' meeting communication, 7 July 2026, PDF p. 2](https://www.cnmv.es/webservices/verdocumento/ver?t=%7B9fbdd78c-1616-4873-9c74-74f4deec7c5d%7D).

**Company statement.** The technical website distinguishes between the data platform, data science and the integration of models into production, and names Azure and Google Cloud as cloud providers with which it works. That does not identify the provider behind Try-on. [Inditex, Data & Tech Teams, “How we work” and “Cloud Excellence”](https://xtend.inditex.com/tech-teams).

**Our interpretation.** There are reasons to expect a combination of internal capabilities and external technology: the company describes its own data and modelling teams, while Zara's Irish policy mentions technology and analytics providers supporting Try On. This does not reveal how intellectual property is divided, the model's provider or a complete cloud architecture. [Inditex, Data & Tech Teams](https://xtend.inditex.com/tech-teams) · [Zara Ireland, privacy policy, §5, p. 11](https://static.zara.net/static/pdfs/IE/privacy-policy/privacy-policy-en_IE-20251202.pdf).

**Our interpretation.** The relevant economic decision is more than whether to build or buy. It includes who defines the problem, retains access to data, evaluates errors, can switch providers and is accountable when an output harms a customer or worker. An external solution can provide speed and specialist knowledge, and it can also create dependence on costs, licence terms and services that are not visible in the initial price. An internal team may integrate a tool more closely with the business, but needs talent, maintenance and the ability to challenge results. Without contractual and technical information, we cannot assign Inditex a precise degree of independence.

**Documented fact.** The 2024 sustainability report records an investment related to Epoch Biodesign and AI-enabled enzyme development in materials research. It is an R&D option, distinct from an already deployed commercial capability or a proven return. [Inditex, 2024 non-financial information, printed pp. 98 and 196](https://static.pullandbear.net/2/static2/itxwebstandard/legal/pdf/Sustainabilityreport2024_en.pdf).

**Our interpretation.** This bet reflects a different logic for allocating resources: funding learning with uncertain results and a potentially long horizon. Its value should not be confused with immediate savings from an internal application. Judging it would require technical milestones, the cost of scaling and verifiable materials outcomes, as well as the share of value Inditex could capture. A stake or research investment buys an opportunity. It does not guarantee industrial viability.

## Governance: what is written and what remains to be tested

**Company statement.** In 2024 the security team said it was working with other areas on an initiative to develop a framework for responsible AI use. In 2025 Inditex identified technological transformation and AI among emerging risks, alongside possible failures, attacks and data exposure. The 2025 sustainability report adds that the cybersecurity advisory committee examined AI's impact and that training covered AI- and deepfake-related threats. This evolution in disclosure does not demonstrate that the framework had been completed or abandoned. [Inditex, 2024 non-financial information, “Cybersecurity”, printed p. 320](https://static.pullandbear.net/2/static2/itxwebstandard/legal/pdf/Sustainabilityreport2024_en.pdf) · [Inditex, 2025 sustainability report, printed pp. 197–198](https://www.inditex.com/itxcomweb/api/media/f8d28973-3e94-45a5-9b0f-9e4d4b1c14b3/SustainabilityReport.pdf?t=1773404719972) · [Inditex, 2025 corporate governance report, “Emerging risks”, p. 83](https://www.inditex.com/itxcomweb/api/media/bc056846-807b-4924-b11c-c85797226a7f/IAGC2025_ENG.pdf?t=1773251265953).

**Company statement.** Zara Ireland's policy says photos are deleted after the avatar is created. The avatar is retained for up to a year for account holders, extendable with consent, or 24 hours without an account. This is a statement for Ireland, not an audit of implementation or a universal rule. [Zara Ireland, privacy policy, §§2, 4 and 5, pp. 6 and 10–11](https://static.zara.net/static/pdfs/IE/privacy-policy/privacy-policy-en_IE-20251202.pdf).

{{ include_html("snippets/analisis-de-empresas/inditex/privacidad.html", anim_shell="on") }}

**Our interpretation.** Deleting the photograph and retaining the avatar are separate decisions, as are consent, security and rights of access. Convincing governance would show how these commitments are checked, what happens after an error or misuse, and who can stop an application. For systems that affect workers, explanations and practical review of decisions matter too. We have not identified enough public evidence to assess those mechanisms in each case. That gap is neither an incident nor proof of non-compliance.

{{ include_html("snippets/analisis-de-empresas/inditex/cronologia.html", anim_shell="on") }}

## Two explanations and three observable scenarios

**Our interpretation.** The coordination and distribution hypothesis has a resource-allocation logic: commercial and logistics infrastructure, the declared technical teams and visual applications offer concrete places where an improvement could be amplified. But an alternative deserves equal scrutiny. Inditex may be spreading AI tools internally to protect efficiency and service quality without there yet being a unified coordination strategy that can be demonstrated from outside. The lack of detail could reflect either broad undisclosed adoption or limited experiments. Neither reading permits a ranking of competitors by “AI leadership”.

**Our scenario, next 12–24 months: demonstrated integration.** If the company publishes an operational case with the previous process, scope, human intervention and a consistent comparison of availability, errors and cost, the coordination thesis would gain support. This assumes that the system affects real decisions at sufficient scale rather than being an isolated test. The value-creation case would weaken if an improvement disappeared against a comparable group or its full costs exceeded its benefits. Publication of one successful case would not prove that every process works the same way.

**Our scenario, next 12–24 months: value concentrated in visual experience.** If verifiable adoption and outcome data emerge for Try-on or content production while attributable operational impact remains undocumented, the more prudent reading would give visual distribution greater weight than coordination. This assumes that the measurements distinguish users, purchases and incremental costs. The scenario would be weakened if robust evidence of other internal applications were subsequently disclosed. Lack of publication alone does not prove that they do not exist.

**Our scenario, next 12–24 months: limited or contested benefit.** Use could grow without improving net returns, or create privacy, quality or working-condition frictions that make it more expensive to operate. The assumption is that costs and side effects are large enough to offset the value created. Comparable, auditable results showing net benefits, together with effective controls, would refute that conclusion for the uses measured. We assign no probabilities: there is not enough data to justify them.

**Our monitoring criteria.** To test this analysis, we would look for:

- A breakdown of AI investment and costs.
- Customer trials comparing conversion and reasons for returns.
- Operational cases documenting autonomy and measured outcomes.
- Contracts and changes in creative commissions.
- Evidence on retention, oversight and recourse in the affected markets and systems.

For Epoch, the relevant milestones are technical and industrial, not a generic mention of AI. Each signal needs a date, scope and denominator before it can support a conclusion.

**Conclusion · our interpretation.** Inditex's position makes it plausible that a small improvement in a well-integrated decision could have considerable value. It also makes it easy to attribute to AI outcomes that depend on decades of commercial organisation, infrastructure and human work. The most visible fact today is a new way of presenting product. The stated internal ambition is broader. The space between the two is where the next evidence needs to go.

## Main sources

- [2025 sustainability report](https://www.inditex.com/itxcomweb/api/media/f8d28973-3e94-45a5-9b0f-9e4d4b1c14b3/SustainabilityReport.pdf?t=1773404719972). Inditex, sustainability 2025, printed pp. 197–198 (PDF pp. 110–111). Cybersecurity advisory committee and training.
- [Shareholders’ meeting communication filed with CNMV](https://www.cnmv.es/webservices/verdocumento/ver?t=%7B9fbdd78c-1616-4873-9c74-74f4deec7c5d%7D). Inditex meeting communication filed with CNMV, 7 July 2026, PDF p. 2. Chief executive statement.
- [FY2025 results](https://www.inditex.com/itxcomweb/se/en/press/news-detail/b870d5ec-6b7e-491d-b38e-340cd69036df/fy2025-results). Inditex, 11 March 2026. Sales, strategy, customer experience and outlook. The page may be updated.
- [2026 half-year financial report](https://www.cnmv.es/webservices/verdocumento/ver?e=x2nr8h8c2xhhP723fHLUSFzWKkWgDybUdOsLCgb8yTD1xVx9TYwe6nqlQTLZolJy). CNMV, half-year report signed 8 September 2026, pp. 3 and 11.
- [H1 results](https://www.inditex.com/itxcomweb/km/en/press/news-detail/a2c9b82b-1e27-4582-a7c9-746bc88b744e/interim-half-year-2026-results). Inditex, 9 September 2026. Cash table and outlook.
- [2024 non-financial and sustainability information](https://static.pullandbear.net/2/static2/itxwebstandard/legal/pdf/Sustainabilityreport2024_en.pdf). Inditex, 2024 non-financial information, printed pp. 98, 196, 236 and 320.
- [2021 annual report](https://static.inditex.com/annual_report_2021/en/documents/annual_report_2021.pdf). Inditex, FY2021, printed p. 202.
- [2025 corporate governance report](https://www.inditex.com/itxcomweb/api/media/bc056846-807b-4924-b11c-c85797226a7f/IAGC2025_ENG.pdf?t=1773251265953). Inditex, corporate governance 2025, p. 83. Emerging risks.
- [Data & Tech Teams](https://xtend.inditex.com/tech-teams). Inditex, no visible date. Capabilities declared by its technical teams.
- [Privacy policy](https://static.zara.net/static/pdfs/IE/privacy-policy/privacy-policy-en_IE-20251202.pdf). Zara Ireland, §§2, 4 and 5, pp. 6 and 10–11. The path contains 20251202, but the original publication date is unverified.
- [Coverage of Zara imagery](https://www.investing.com/news/stock-market-news/zara-turns-ai-generate-fashion-imagery-using-reallife-models-4414956). Reuters, Helen Reid, 18 December 2025. Open syndication on Investing.com.
- [Digital-twin campaign](https://hmgroup.com/news/hm-continues-its-exploration-of-creativity-with-ai/). H&M, 2 July 2025. Company statement about a digital-twin campaign.

All sources were accessed on 6 November 2026. The precise publication date of the cited annual reports has not been verified.