Parag Parekh is chief digital officer of Ingka Group, the largest IKEA franchisee, whose organisation he described as 160,000 people strong. He spoke to McKinsey’s Hai-Ly Nguyen and Holger Harreis for an interview published on 11 June 2026. The conversation covered how Ingka decides which AI initiatives to fund, where the returns are coming from, how the customer experience is changing and what agentic commerce could mean for a retail brand.
Three areas, one ranked list
Parekh described AI’s impact at Ingka in three areas. In the customer experience, generative AI and agentic commerce are moving relationships away from single transactions and towards solving customers’ problems. In supply chain and logistics, AI is used to take out costs. In the back office, the aim is to free coworkers from repetitive tasks so they can spend more time with customers.
To decide what to fund, the company maps initiatives on two axes, customer-focused against coworker-focused and growth against cost, then ranks the whole portfolio by its business and customer impact. Recommendation engines, pricing and agentic commerce sit in the customer-growth quadrant, while supply chain and last-mile optimisation sit in the cost quadrant. Parekh warned of "the risk of doing everything but maybe nothing" if AI initiatives are allowed to multiply.
Where the returns are
The strongest returns, he said, still come from traditional AI that has had several years to mature. The clearest example is goal-based order allocation, which uses weighted AI models to choose the best fulfilment route for each order across more than 400 fulfilment and shipping points, and which has improved Ingka’s cost ratio in shipping and logistics. Personalised cross-selling and upselling is another source of compounding returns. Generative AI in customer service, particularly for multilingual support and routine interactions, is promising but has not yet reached scale.
From buying a sofa to designing a room
Parekh described a change in what customers come to IKEA for. More of them now ask for help designing a room or shaping their home, and fewer arrive knowing exactly which product they want. Using AI with LiDAR room scanning, the company asks a few questions about style, colour and budget and generates four or five options from a 3D scan of the room, which the customer then refines. He expects the same experience to move into voice and chat-based agentic formats.
The design service shows what the economics make possible. Helping a customer design a space used to take an average of six hours of a coworker’s time and cost the customer around €70 a session. Automation has cut that to about 30 minutes per room per coworker, which puts the service within reach at IKEA prices.
Adoption is the measure
For Parekh, the hardest part of scaling AI is managing the change. Ingka judges each digital product on its deployment and adoption, measured end to end. It has run onboarding workshops for senior leaders, completed AI literacy programs for 40,000 coworkers and launched pilots in three countries with store-level ambassadors who find and showcase high-impact use cases.
Inside the digital organisation, AI is compressing the software development cycle into rapid prototyping and experiments, which is changing the work of product managers, designers and engineers. A program called Digital Next is working out what group digital should look like in 2028, and Parekh said every function, from HR to finance, should be asking which of its processes AI will reset.
The agentic commerce question
Parekh sees agentic commerce as the latest stage of a shift that has run for 15 to 20 years, in which customers increasingly decide where they meet brands. Over the past two to three years the share of search has moved sharply towards agentic search, and he expects that brands will need to appear in those experiences. The open question is whether purchases end up on the agentic platforms or are sent back to the brand. He expects that to play out over the next two to five years, with both sides inclined to collaborate.
What it means for leaders
- A short, ranked portfolio is easier to fund and to stop than a long list of pilots. Ingka’s two axes give a board a simple way to choose.
- Traditional AI in fulfilment and personalisation is still where Ingka sees its clearest returns, and generative AI budgets are worth judging against those returns.
- Measuring adoption end to end, and training at scale, is how Ingka turns tools into results.
- Retailers need to decide how they will appear in agentic search and shopping assistants while the commercial terms are still unsettled.
