Cases
IKEA, Duolingo, Prosus, Toyota, Freeport-McMoRan and DBS. In all six, it was not only the process that changed, but also the work of the people in it.
AI assistant Billie handled 47% of the queries to the call centre. Ingka, IKEA's largest franchisee, retrained 8,500 call centre employees as interior design advisers. In financial year 2022, the remote interior design channel accounted for 1.3 billion euros in revenue.
The gain lies in what people go on to do once the routine work is gone.
With the same number of people, Duolingo now produces four to five times as much course content in the same amount of time. Since its AI-first announcement in April 2025, no full-time employee has been laid off and the company is still hiring. It is, however, phasing out external contractors for work AI can do.
AI as added capacity, more output from the same team.
40,000 employees together run more than 60,000 AI agents. Without any mandate from headquarters, companies in different countries arrived at the same twenty use cases. Around 2% of the agents deliver most of the impact.
The value sits in a few well-chosen processes, not in as many use cases as possible.
Global capacity planning ran on dozens of spreadsheets, large planning teams, weeks of work and several layers of approval. Toyota replaced this with a process in which AI brings demand and supply together and takes planners through scenarios in minutes. The planning team shrank by more than 80%, not through layoffs, but by moving people to work that calls for more judgement.
Take out the hand-offs and the data shuffling, and people are free for the work they are really needed for.
Instead of building a new processing plant at 1.5 to 2 billion dollars, the mining company got more out of an existing one. Cross-functional teams began adjusting process settings every hour instead of once per shift. Within months, copper production at the Bagdad mine rose by 10%. Around 60% of the code developed there could be reused at seven other sites.
The result came from changing how the work was done, not from adding a model.
The bank organised its work around customer journeys, such as opening an account or buying a home, with a cross-functional team for each journey. Building and deploying an AI model took 15 to 18 months in 2018 and two to three months in 2023. AI generated an estimated value of more than 1 billion Singapore dollars.
Speed comes from how the organisation is set up.
Patterns
Per process
Redesign does not have to change the whole organisation at once. These examples cover one process each, with a result that could be measured straight away.
A voice assistant handles first customer contact and simple requests. Within five months, the system understood about 75% of customer needs.
Result: a 6% increase in customer satisfaction and more than 10 million dollars in annual impact.
Source: McKinsey, Rewiring customer experience for the agentic era, July 2026An AI assistant supports service agents on complex customer cases with the right information at the right moment.
Result: 24% more cases resolved at first contact and 30% higher agent productivity.
Source: McKinsey, Rewiring customer experience for the agentic era, July 2026AI decides which offer a customer sees, when and through which channel, based on signals from all channels combined.
Result: 40 million euros in margin improvement.
Source: McKinsey, Rewiring customer experience for the agentic era, July 2026Marketing processes were redesigned end to end, and more than 2,000 people in marketing and R&D were trained.
Result: creative assets adapted 30% faster and 20 to 40% time saved on brand performance reporting.
Source: BCG, Driving sustained structural cost advantage with applied AI, July 2026Every back-office process was reviewed function by function: eliminate, simplify, then automate. More than a million activities were automated.
Result: around 4.5 billion dollars a year in lower operating costs, 30% of the cost base. About half of that came from AI.
Source: BCG, Driving sustained structural cost advantage with applied AI, July 2026AI mapped procurement spend and drafted negotiation letters for the long tail of suppliers. The company completed more than 3,500 negotiations in under 100 days.
Result: 900 million dollars in cost savings.
Source: BCG, Driving sustained structural cost advantage with applied AI, July 2026Drug discovery was redesigned with AI at every stage.
Result: 40 to 50% shorter cycle times and 20% lower molecular synthesis costs.
Source: BCG, Driving sustained structural cost advantage with applied AI, July 2026Thousands of employees were trained first through a dedicated digital academy, before AI was deployed in smelting operations.
Result: more than 120 million dollars in impact and a 170% ROI.
Source: McKinsey, The operating model advantage, July 2026Decisions on pricing, promotions and portfolio used to run through spreadsheets and separate meetings. They are now based on AI scenarios, and employees learn to work with the system while making real decisions.
Source: McKinsey, The decision dividend, August 2026AI agents handle customer queries by chat and phone, more than a million conversations a month.
Result: 70% of queries resolved without an agent.
Source: DecagonAI agents answer questions from readers and writers.
Result: more than 90% of questions resolved without human involvement.
Source: DecagonAI agents hold sales conversations with customers.
Result: 1 million dollars in revenue from conversations handled entirely by AI.
Source: DecagonQuotes for sabotage and terrorism cover in the London Market were prepared by hand from brokers' email submissions, which took up to three days. AI now extracts the data from those submissions, while the underwriter stays in charge of the assessment.
Result: in the proof of concept, a quote in three minutes instead of up to three days.
Source: HiscoxA dispute required review of more than 600,000 documents on a limited budget. AI filtered out more than 90% of the documents.
Result: the document review was completed 50% under budget.
Source: CasepointKnowledge base
The examples are there, and the technology is already working at these organisations. We have brought together publications from consultancies such as McKinsey, BCG, PwC and Bain, technology companies such as OpenAI, Microsoft, Google, Salesforce and IBM, and institutions such as MIT, Carnegie Mellon and the World Economic Forum, more than 2,000 use cases and a process library of over 500 processes, before and after redesign. What remains is the question of how it will work inside a specific organisation.
Next step
The Scan maps this out. Introductory price: 2,500 euros excluding VAT.