Build, buy or partner? How fashion companies should source AI capabilities

KEY TAKEAWAYS Summary by the editors

1. Most fashion companies should buy AI for common tasks, build only where their own data and processes create a real advantage, and partner when they need both scale and customisation.
2. MIT's NANDA initiative reported in 2025 that buying AI tools from specialised vendors and building partnerships succeeded about 67 per cent of the time, while internal builds succeeded only one third as often.
3. In McKinsey's 2026 State of AI survey, 32 per cent of respondents said their organisations had decided against buying at least one software product or feature because they could build it in-house with agentic coding tools.
4. Gartner estimates that only about 130 of the thousands of vendors claiming agentic AI offer genuine agentic capabilities, a practice it calls agent washing.
5. Contracts for AI services should settle data ownership, use of company data for model training, exit terms and cost at scale before a tool becomes embedded in daily work.

Fashion companies should usually buy AI for common tasks such as copywriting, translation, customer service and document processing, build only where their own data and processes create a genuine advantage, such as forecasting for their specific assortment, and partner with technology providers when they need both scale and deep customisation. Most end up with a mix, so the real decision is made use case by use case.

What are the options for sourcing AI?

What does the evidence say about building versus buying?

The clearest recent evidence comes from MIT's NANDA initiative. Its 2025 report The GenAI Divide, based on interviews, an employee survey and an analysis of public deployments, found that purchasing AI tools from specialised vendors and building partnerships succeeded about 67 per cent of the time, while internal builds succeeded only about one third as often, according to Fortune. The researchers linked failures mainly to tools that were not adapted to real workflows.

At the same time, building has become cheaper for some tasks. McKinsey's 2026 State of AI survey found that 32 per cent of respondents said their organisations had decided against purchasing at least one software product or feature because they could build the functionality in-house using agentic coding tools. That suggests the build option is becoming viable for smaller, well-defined applications, even if large custom AI systems remain risky.

Read also

Fashion technology trends 2026: what is real and what is hype

When does buying AI make sense for a fashion company?

Buying makes sense when the task is similar across companies and the vendor's scale gives it better data, models or maintenance than a single brand could achieve. In fashion this typically covers:

• Translation and first drafts of product descriptions.
• Customer service assistance and ticket routing.
• Document processing, such as reading invoices, delivery notes or emailed orders.
• Image editing, background removal and basic visual tagging.
• Productivity assistants for office work, meetings and email.

AI features are also increasingly built into software that fashion companies already use for planning, product information, e-commerce and ERP. Activating those features is often the fastest route, provided they can be tested against a baseline like any other investment.

When should a fashion company build its own AI?

Building makes sense when the use case depends on data or processes that are specific to the company and represent a competitive advantage: forecasting for a particular assortment and channel mix, size recommendations based on the brand's own fits and returns, or pricing logic tied to a distinctive commercial model. It also requires the skills to maintain the solution, because models degrade as products, customers and markets change.

Building rarely means training models from scratch. More often it means combining a foundation model or standard machine learning methods with the company's own data, rules and interfaces. The differentiation lies in the data and the workflow, not in the algorithm.

What does partnering look like in practice?

Large fashion companies increasingly co-develop with major technology providers. Zalando developed its shopping assistant with OpenAI and has scaled it to 25 markets in local languages, according to OpenAI. Levi Strauss & Co. announced in November 2025 a partnership with Microsoft to build an AI superagent in Microsoft Teams, an orchestrating assistant that answers employee questions by drawing on several behind-the-scenes agents, intended for corporate, retail and warehouse staff, according to Microsoft. Both cases combine a provider's platform with the company's own data and processes.

Partnering suits strategic use cases at scale, but it needs clarity on who owns what is built, who may reuse it and how responsibilities are split when something goes wrong.

How should fashion companies evaluate AI vendors?

Vendor claims need careful checking. Gartner estimates that only about 130 of the thousands of vendors claiming agentic AI offer genuine agentic capabilities; many rebrand existing assistants, robotic process automation or chatbots, which Gartner calls agent washing. Useful questions for any AI vendor include:

1. Can you show the tool working on our data, not just in a demonstration?
2. Which models do you use, and how do you handle model changes and errors?
3. Is our data used to train models for other customers, and can we opt out?
4. How does the tool integrate with our ERP, product information and e-commerce systems?
5. What does it cost at full scale, and how is usage priced?
6. How do we export our data and configurations if we leave?

Read also

How to run a software selection (RFP) without losing a season

How do you avoid lock-in and protect your data?

Lock-in risk grows as an AI tool becomes embedded in daily work. Contracts should settle who owns the input data, the outputs and any fine-tuned models, whether company data may be used for training, where data is processed, what happens to data at the end of the contract and how prices can change. Keeping product, inventory and customer data well structured in systems the company controls is the most effective protection, because it makes switching providers a manageable project rather than a restart.

The pragmatic answer to build, buy or partner is therefore a portfolio: buy for common tasks, build selectively where the company's data creates an edge, partner for large strategic programmes, and keep control of the data that all three depend on.

Frequently asked questions

Should a fashion company build or buy AI?

For most common tasks such as translation, customer service and document processing, buying is faster and more reliable. Building makes sense where proprietary data and processes create an advantage, for example in forecasting for a specific assortment. MIT's NANDA initiative found in 2025 that purchased tools and partnerships succeeded about 67 per cent of the time, while internal builds succeeded only one third as often.

What is agent washing?

Agent washing is Gartner's term for vendors rebranding existing products, such as assistants, robotic process automation or chatbots, as agentic AI without substantial agentic capabilities. Gartner estimates that only about 130 of thousands of vendors claiming agentic AI are genuine. Buyers should ask for demonstrations on their own data.

How do you choose an AI vendor?

Test the tool on your own data in a paid pilot with clear success criteria. Check how it integrates with existing systems, how your data is used and protected, what it costs at full scale and how you can export data if you leave. References and demonstrations are useful but not sufficient.

How can a fashion company avoid AI vendor lock-in?

Settle data ownership, training rights, data location, exit terms and pricing changes in the contract. Keep core product, inventory and customer data structured in systems the company controls. That makes switching providers a manageable project rather than starting again.

GuideThe complete guide to AI strategy for fashion companiesRead the complete guide