Five companies below rolled out AI systems that already show up in daily employee and customer workflows, not stuck in a strategy deck somewhere. What ties them together isn’t technology. It’s a decision made before anyone switched anything on: which repetitive task was actually worth handing to AI, and how anyone would know if it worked.
What Happens When AI Answers Employees’ IT Questions?
Image by rawpixel.com on Magnific
IBM built an internal chatbot called AskIT, now used by more than 280,000 employees in 40 languages. It runs on IBM’s own watsonx Assistant and was trained on the roughly 80% of IT requests that repeat constantly: password resets, access requests, the same handful of questions a helpdesk fields all day.
Within four months of launch, over 133,000 employees had used it. About 75% of tickets got resolved without ever reaching a human technician.
Does Giving Employees an AI Research Assistant Actually Save Time?
Photo by JUICE on Unsplash
McKinsey rolled out an internal AI tool called Lilli across the whole firm. Roughly 72% of employees use it, logging more than 500,000 prompts a month against the firm’s own research and knowledge base.
The reported effect: people are getting back around 30% of the time that used to go into manually digging through old project files.
That time doesn’t just disappear, either. It shifts toward the part of the job that’s still entirely human: figuring out what the actual problem is, stress-testing the recommendation, sitting across from a client who wants to know why they should believe it.
Can AI Actually Save Money Without Anyone Noticing It’s There?
Photo by Daniel Andraski on Pexels
McKinsey estimates that AI in supply chain operations can cut logistics costs by 5–20%, mostly through better demand forecasting and route or inventory optimization that catches inefficiencies a quarterly review would miss.
This is the least visible use case on the list. No chatbot, no new screen anyone opens. But it’s often the first one finance actually sees show up in the numbers, which might be exactly why it gets the least attention.
None of these deployments needed a data science team built from scratch. Someone just decided which repetitive decision was worth handing over, then checked whether it actually worked.
What Does “Faster Customer Support” Actually Look Like?
Image by magnific
Liberty London uses Zendesk’s AI to automatically read and route incoming support tickets. That single change cut first-reply time by 73% and lifted customer satisfaction by 9 points. Clean example, clearly AI-driven.
Dunlop Sports is a related but slightly different story. A full overhaul of its support setup, bringing voice, chat, and email into one place, cut answer times by 89% and holds a 95.3% satisfaction score. Worth being straight about this one: Dunlop’s numbers came from that broader overhaul, with AI-assisted support rolling out as the next phase, not the reason behind these particular results yet.
A fast wrong answer still frustrates someone. AI doesn’t replace judgment here. It just removes the wait before a real answer shows up.
How Much Faster Is AI Making Hiring?
Photo by MART PRODUCTION on Pexels
Paradox’s conversational hiring assistant, Olivia, is used by companies including FedEx and Unilever. It handles the first stretch of hiring: answering candidate questions, screening applicants, booking interviews, all over text. Across Paradox’s own published customer results, time-to-interview drops from an industry-typical 5–10 days to under 24 hours in many cases.
AI hiring tools were about to fall under the EU’s strictest AI category, requiring bias testing and human oversight, until the deadline got pushed from August 2026 to December 2027. The scrutiny’s still coming. Just later than most companies were bracing for.
What Do These Five Companies Have in Common?
Not budget size. Not access to better models than anyone else has. What repeats across all five is a decision made before any tool got switched on: which repetitive, high-volume task was genuinely worth handing to AI, and how success would get measured once it went live.
That decision, not the algorithms themselves, is what the Master in Artificial Intelligence is built to teach: where to point AI, and how to tell if it’s working. The program runs fully online, so it fits around the job you already have. We covered the career-facing side of this same shift in a recent look at what AI hiring actually rewards, if you want the applicant’s-side story to go with this one.
Sources
1. IBM (2026). “Using AI to deliver a digital-first employee experience” — AskIT case study, IBM CIO Organization.
2. McKinsey & Company internal AI tool “Lilli” adoption and time-savings figures — reported via DigitalDefynd, “8 Ways McKinsey Is Using AI,” June 2026.
3. McKinsey & Company estimate on AI and logistics cost reduction (5–20%) — cited via Capsule CRM, “Small business AI adoption statistics for 2026,” April 2026.
4. Zendesk (2026). Customer case studies: Dunlop Sports and Liberty London, zendesk.com/customer.
5. Paradox published customer results on time-to-interview reduction; client list confirmed via Paradox and third-party review sources, 2026.
6. European Union, Digital Omnibus on AI — provisional agreement postponing high-risk AI obligations under the EU AI Act from August 2026 to December 2027, reported via Pinsent Masons and Gibson Dunn, 2026.
