The Gist
Preparation beats technology. Across four B2B companies on three continents, AI success depended on organizational readiness — clean processes, clear ownership and prepared people — not the sophistication of the AI itself. Field expertise is the real training data. POSCO’s AI-controlled blast furnace only worked once paired with veteran furnace operators’ decades of intuition, lifting daily molten iron output by 240 tons. Adoption is a multi-year culture program, not a rollout. Panasonic Connect’s time savings from its internal AI assistant more than doubled in year two as employee behavior matured, not because the technology changed.
I learned this lesson the expensive way, inside one of the largest companies on earth. During my years leading service and experience transformation at Samsung, I worked on AI projects across very different worlds (groups) — healthcare, mobile, telecommunications, high tech software enterprise — including conversation design programs that put intelligent systems directly in front of customers.
Some of those projects delivered. Others taught me more.
And the difference was almost never the technology. The projects that worked were the ones where the organization had done the unglamorous work first: clean processes, clear ownership, people who understood why the change was happening and wanted it to succeed. The projects that struggled had bought intelligence and skipped preparation.
Twenty years and many companies later, I watch organizations repeat the same mistake at scale. AI is not a miracle. It is an amplifier. Point it at a prepared organization, and it multiplies strength; point it at confusion and it multiplies that instead.
I wrote about the foundations of this in The Five Pillars of Successful AI and Customer Experience Transformation, and since then the evidence has only grown. So instead of theory, let me show you four B2B companies — on three continents, in four industries — that prepared before they deployed, and what each one can teach your leadership team.
1. Pair the Machine With Your Experts — Steel, South Korea
My favorite honesty in corporate communications comes from a steelmaker. POSCO admits openly that its first attempt to build an intelligent blast furnace, using a global IT company’s technology alone, did not deliver. The technology by itself wasn’t enough. The breakthrough came only when they combined it with something no vendor could sell them: the accumulated intuition of their own field experts, people who had read furnaces by eye for decades. That combination produced an AI-controlled furnace that lifted daily molten iron production by 240 tons, contributed to savings of 250 billion won across four years of smart projects and made the company the first in its country selected as a World Economic Forum Lighthouse Factory.
The preparation lesson: your veterans are not obstacles to AI. They are its training data. Companies that sideline their experts to “move fast” throw away the one asset competitors cannot copy.
How Did POSCO Increase Molten Iron Production With AI?
POSCO combined AI-controlled blast furnace technology with veteran field experts’ intuition, lifting daily molten iron output by 240 tons and contributing to 250 billion won in savings across four years of smart projects.
2. Prepare the Knowledge Before the Intelligence — Heavy Industry, Europe
At a German motion-technology manufacturer, generative AI now writes automation code for production machines from plain spoken language and walks maintenance crews through fixing errors step by step.
But look at what came before the magic: Schaeffler and its partner Siemens first connected the assistant to the complete engineering backbone — every relevant manual, guideline and piece of documentation — so the AI answers from the company’s actual knowledge, not from guesses.
The result addresses one of Europe’s hardest industrial problems, the shortage of skilled automation engineers, by letting less-experienced shop-floor employees grow into engineering roles with the AI as their guide.
The preparation lesson: an assistant without your knowledge is a stranger with confidence. Organize your documentation, your rules, your tribal know-how first — because that, not the model, is what your people and customers will actually experience.
Related Article: The Great AI Productivity Boom (Your Customers Will Never Feel)
Why Did Schaeffler Connect Its AI Copilot to Engineering Documentation First?
Schaeffler and Siemens linked their industrial copilot to the company’s complete engineering backbone — manuals, guidelines and documentation — before deployment, so the AI answers from actual company knowledge rather than guesses, helping less-experienced workers grow into automation roles.
3. Treat Adoption as Culture, Not as a Rollout — Electronics, Japan
In February 2023, before most boards had even discussed generative AI, a Japanese B2B electronics company gave an internal AI assistant to every single employee, from the CEO to the newest hire — deliberately built in a secured environment so nobody would need risky workarounds. Panasonic Connect reports that by 2024 the time saved through AI reached 448,000 hours a year, 2.4 times the previous year, as employees evolved from asking the AI questions to assigning it whole tasks. Read that progression again: the gains more than doubled in year two, not because the technology changed dramatically, but because the people did. The company treated AI skill as talent development, and the freed hours flow into more creative work.
The preparation lesson: adoption is not an IT rollout with a training video. It is a cultural program measured in years, and it compounds. Whoever starts building that culture now collects the compounding.
How Much Time Did Panasonic Connect Save With Its AI Assistant?
Panasonic Connect gave every employee access to a secured internal AI assistant starting in February 2023, and by 2024 the company reported 448,000 hours saved annually — 2.4 times the prior year — as employees shifted from asking questions to assigning full tasks.
4. Give People a Reason They Can Feel — Technology, North America
The fastest way to kill an AI program is to make it about the tool. A US communications company made it about time instead. Research for customer outreach used to cost each seller about four hours; with AI it takes fifteen minutes. Lumen Technologies puts the value of those four recovered hours per seller, per week, at $50 million a year — and the recovered time goes where it belongs: to customers. Sellers now take on complex work they used to avoid purely because it consumed too much time. Notice the customer experience logic hiding inside an efficiency number: the return was never “cheaper sellers.” It was more hours of human attention pointed at customers.
The preparation lesson: define, before deployment, exactly where the saved time will go. If the answer is “headcount reduction,” your people will resist and your customers will feel it. If the answer is “more time with customers,” everyone rows in the same direction.
How Much Value Did Lumen Technologies Gain From AI-Driven Time Savings?
Lumen Technologies cut seller research time from about four hours to fifteen minutes using AI, valuing the recovered time at $50 million a year and redirecting it toward customer-facing work.
The 5-Step Sequence for Preparing an Organization for AI
Four companies, four countries, four industries — one pattern. None of them started with the model. They started with the organization. If you are preparing your own company, this is the sequence I have seen work, at Samsung and everywhere since:
First, pick a real problem your customers actually feel, and write down how you will measure the change — before any vendor enters the room. Second, put your most experienced people inside the project, not in its audience; their knowledge is the raw material. Third, get your documentation and data into shape, because the AI will represent your company exactly as well as your knowledge allows. Fourth, run adoption as a multi-year cultural program with visible executive use, safe tools and skills treated as career development. Fifth, declare where the recovered time goes, and make sure a meaningful share of it lands on the customer.
None of this is glamorous. All of it decides the outcome. The steelmaker needed its furnace veterans. The manufacturer needed its manuals. The electronics company needed three years of patience. The telecom needed a promise about time. The AI came last in every story — and that is exactly why it worked.
Your organization is the product. AI just ships whatever you’ve built, faster.
Learn how you can join our contributor community.