How do you get employees to actually use AI? An adoption plan for fashion teams
Licences alone do not change how people work. A practical adoption plan for fashion companies: use cases by team, training, guardrails, champions and measurement.
KEY TAKEAWAYS Summary by the editors
- Buying AI licences does not create adoption; usage grows when tools are tied to specific tasks in each team's weekly routine.
- McKinsey's 2025 Superagency survey found employees were three times more likely than leaders estimated to be using generative AI for a meaningful share of their work.
- In the same survey, 48% of US employees said formal training would most increase their daily use of generative AI.
- McKinsey's State of AI 2026 found that nearly three-quarters of AI high performers had fundamentally redesigned workflows, compared with about a quarter of other respondents.
- An adoption plan should define use cases per function, provide role-specific training, publish clear guardrails, appoint champions and track usage alongside business outcomes.
Employees use AI when it saves them time on tasks they already do, when they know what is allowed, and when someone shows them how. A workable adoption plan for a fashion company therefore starts with specific use cases per team, adds role-based training and clear guardrails, uses internal champions, and measures both usage and business results. Rolling out licences without these steps usually leaves most seats idle.
Why do AI licences often go unused?
The gap is rarely the technology. McKinsey's Superagency in the workplace report, published in January 2025 and based on a survey of 3,613 employees and 238 C-level executives, found that leaders estimated only 4 percent of employees used generative AI for at least 30 percent of their daily work, while employees self-reported 13 percent. Only 1 percent of executives described their rollout as mature. In other words, informal use is often ahead of the official programme, but it is scattered and unmanaged.
The same survey shows what employees want: 48 percent of US respondents named formal training as the initiative most likely to increase their daily use, followed by seamless integration into workflows (45 percent) and access to tools (41 percent). More than a fifth reported receiving minimal or no support.
Which AI use cases should each fashion team start with?
Generic encouragement to 'try AI' produces little. Better to give each team two or three concrete starting points tied to real deliverables.
| Team | Starter use cases | What must be checked by a person |
|---|---|---|
| Design | Mood board variations, trend research summaries, colourway exploration | Originality, IP risk, feasibility |
| Merchandising and planning | Summarising sell-through reports, drafting range review notes, scenario questions on spreadsheets | Numbers against source systems |
| E-commerce and content | Product description drafts, translations, SEO metadata | Accuracy of materials, care and fit claims |
| Sales and wholesale | Account briefings before appointments, follow-up emails, order summaries | Prices, terms and delivery dates |
| Customer service | Reply drafts, policy look-ups, case summaries | Refund decisions, tone with upset customers |
| Supply chain and sourcing | Supplier email drafting, document comparison, shipment exception summaries | Commitments to suppliers, compliance data |
Sequence matters as well. Launch with the teams where a starter use case is closest to an existing deliverable, such as product content or sales follow-ups, so that early results are visible quickly. Teams whose work depends on data that is not yet reliable, for example planning with incomplete sell-through history, can follow once the inputs are fixed. Publishing a simple rollout calendar tells everyone when their turn comes and reduces the pressure to experiment with unapproved tools in the meantime.
How should AI training be structured?
Training works best when it is short, role-specific and repeated. A one-off webinar on 'what is generative AI' is not enough. A practical structure has three layers:
- Foundations for everyone: what the approved tools can and cannot do, how to check outputs, what data may never be pasted in, and how errors and hallucinations look in practice.
- Role clinics: 60 to 90 minute sessions per team using that team's real documents and tasks from the table above.
- Office hours and refreshers: a regular slot where champions or the AI team help with specific problems, and short updates when tools change.
There is also a compliance dimension. Since 2 February 2025, Article 4 of the EU AI Act has required deployers of AI systems to take measures to support AI literacy among their staff. The European Commission says no certificate is needed, but organisations should be able to show what they did, so keep simple records of who attended which session.
Some retailers have formalised this. Debenhams Group launched an AI Skills Academy in August 2025 with Multiverse, open to all employees and funded through its Apprenticeship Levy, according to FashionUnited.
What guardrails do employees need before they use AI?
Uncertainty about rules slows adoption as much as lack of skill. Publish a one-page policy that answers the questions people actually ask: which tools are approved, which data classes (customer data, unreleased designs, supplier prices, contract terms) are allowed in which tool, who is accountable for an AI-assisted output, and when AI use must be disclosed. Pair it with a simple route for requesting new tools so that people do not turn to unapproved consumer apps.
How do champions and managers drive adoption?
Peer examples persuade more than central announcements. Appoint one champion per team, give them a few hours a month, and ask them to collect and share working prompts and short before-and-after examples. Managers matter too: when a range review or sales meeting template explicitly includes an AI-prepared summary, usage becomes part of the routine rather than an optional extra.
The deeper lever is process redesign. McKinsey's State of AI 2026, based on 1,719 respondents surveyed in May and June 2026, found that nearly three-quarters of AI high performers had fundamentally redesigned workflows, compared with about one-quarter of other respondents. Adding a chatbot to an unchanged process rarely shifts results.
How do you measure whether AI adoption is working?
Track two kinds of measures together. Usage metrics (active users per week, use by team, repeat use) show whether the tools have entered daily work. Outcome metrics show whether that matters: time to publish a product page, hours spent on range review preparation, first response time in customer service. Review both monthly for the first six months, retire use cases that do not stick, and expand the ones that clearly save time or improve quality.
Expect uneven progress. Some teams will adopt quickly, others will need more support or better data before AI is useful. A plan that is honest about this, and that adjusts based on evidence rather than enthusiasm, will get further than a big-bang launch.
Frequently asked questions
How long does it take for employees to adopt AI tools?
There is no fixed timeline, but most companies should plan for several months of training, champion support and process changes before usage stabilises. Reviewing usage and outcome data monthly during the first six months helps identify which teams need more help.
What is the best way to train fashion employees on AI?
Short, role-specific sessions using each team's real tasks work better than generic courses. Combine foundations for everyone with team clinics and regular office hours, and keep records to support EU AI Act literacy obligations.
Should companies ban employees from using public AI tools?
Blanket bans tend to push use underground. A clearer approach is to provide an approved tool with business data protections, define which data may be used where, and offer a quick route to request other tools.
How do you measure AI adoption?
Combine usage metrics, such as weekly active users and repeat use by team, with outcome metrics tied to the task, such as time to publish product content or customer service response times. Usage without outcome improvement suggests the use case or process needs rethinking.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
