> ## Content Index > Fetch the complete content index at: https://www.ceciliaweckstrom.com/llms.txt > Use this file to discover other available public pages before exploring further. # Loved beyond reason - URL: https://www.ceciliaweckstrom.com/the-tenon/loved-beyond-reason/ - Published: 2026-09-27T06:30:51.000Z - Updated: 2026-09-27T13:13:00.000Z - Description: AI will not create the gaps in your business. It will show them to your customers first. - Author: Cecilia Weckstrom - Tags: Brand Building, AI capability, Strategy Execution, The Departure Test **The short version*. An AI agent has no habit, so it removes the inertia most brands book as loyalty and leaves only price and the customer who wants you anyway. That customer is won or lost wherever the business disagrees with itself, and automation at the surface takes every such disagreement public within seconds. Love is still made in marketing. It is now tested everywhere else, and the marketing leader, who makes the promise and hears the customer first, is the one person placed to see both ends. This piece sets out what it takes to close the gaps: a goal every function shares, principles a person and a machine can both decide by, what customers say reaching the people who set price, policy and product with the power to change them, and one owner for the places where departments meet.* A man whose grandmother had just died asked an airline's chatbot about bereavement fares. The bot told him to book now and claim the discount within ninety days. He did. The airline refused, because that had never been the policy. When the case reached a tribunal, the airline argued that its chatbot was a separate legal entity, responsible for its own words. \[E1\] It lost, and paid him a few hundred dollars. The ruling now appears in every deck about AI risk, read as a story about a machine making things up. I read it that way twice before I saw the other story. Read the defence instead. A company stood in front of a judge and said that its own front door was not part of the company. Nobody had decided who owned the join between what the surface promised and what the policy allowed, and the airline's lawyers said so out loud. The money was nothing. The cost was that every customer, and every court, now had it in writing. The gap was there long before the bot arrived. The bot found it in seconds, and took it public. > AI will not create the gaps in your business. Instead, it will find them faster than your customers ever could, and it will show them. ## Reason has been automated Most of what a brand calls loyalty is habit. The last thing bought or the pack at eye height. The name that comes to mind first because it always has. That share has been safe for a century, because people do not re-run the comparison every time they buy. An agent has no habit. It runs the comparison from scratch, on the facts, every time. So AI does not threaten your brand. It strips out the inertia and leaves two things standing: price, and the person who says no, I want that one. In 2004 Kevin Roberts published Lovemarks and gave us "loyalty beyond reason": the brands people choose past the point where the arithmetic would let them. He was right that love beats reason. I have spent twenty-five years inside a brand that proves it every day, so I say what follows with respect. \[L3\] He wrote for a world that has gone. Two things have moved since. *People no longer receive the brand; they make it with you, and the machines read what they make at a scale and speed no individual could before.* A model asked what a brand is does not consult the brand book. It consults the reviews and the forums, the things people posted for nothing. And reason itself has been automated. In 2004 "beyond reason" meant a customer's loyalty outrunning the customer's own sums. In 2026 the sums are done by software, and software does not tire, forget or forgive. **Beyond reason is no longer a nice place for a brand to sit. It is the only margin the machine cannot arbitrage.** Two machines are arriving at once, and it pays to keep them apart. *One answers on the company's behalf:* the service bot, the personalised page, the content pipeline. *The other buys on the customer's behalf,* and in most categories it is not doing that yet, but substituting for search. The first is already at work in your business, and every case in this piece is about it. *The answering machine exposes the seam. The buying machine will price it.* If your category is years away from agents, you are not off the hook. You are early. \[L4\] The current conversation about brand love in the age of AI says: keep the brand lovable, write down what it stands for so the machines can read it, win the shortlist. All sensible, and all of it work the brand team can do on its own. That is the mistake. > Love is still made in marketing. It is now tested everywhere else, at machine speed, and the marketing leader is the only person in the building who sees both ends. ## The trap Automation at the surface does not settle a disagreement underneath it. It multiplies it, and takes it public. The cases that make the news are service cases, because a chat window is where a seam is easiest to photograph. The seams themselves are everywhere: the campaign promise against the shelf price in a market that never signed up to it, the claim on the pack against what the supply chain can prove, the launch date against what a market can stock, the loyalty scheme against the returns desk. Each is a place where two functions were both right and nobody owned the join, and each is about to be answered for by software that speaks in the brand's voice and knows one department's rules. **Klarna** told its own story. It announced that an AI assistant was doing the work of seven hundred agents and celebrated the savings. A year later its chief executive said that cost had been too dominant a factor in the design, that quality had fallen, and that a human would always be available from now on. From a brand perspective, he said, and he is right. His remedy is the wrong lesson. Rehiring people treats the symptom; the seam was that cost was the only number the service had been designed against, and no human on a phone line closes that. Donella Meadows, whose 1999 essay on leverage points is still the most cited work in the systems change field, put parameters and numbers at the bottom of her list, the cheapest lever and the one that changes least. Klarna pulled it hard in both directions and got what a bottom lever buys. \[E2\] **Duolingo** should worry anyone who runs a loved brand, because Duolingo is one: an owl with a personality, a real community, a surface that says we are the playful humans in a serious category. In April 2025 the chief executive told the company it would be "AI-first", phasing out contractors where AI could do the work; it had already shipped a hundred and forty-eight courses in a year that way. Within days people were filming themselves deleting the app, ending streaks of a thousand days on camera, and the brand wiped its own social feeds and went quiet. It is tempting to file this as a memo that went out badly. *Look at what disagreed with what. The surface promised playful humans. The operating decision was content by machine.* Learners in the smaller courses had already noticed errors a human would have caught, and the memo told them why. The seam was real, the fans were standing on it, and the speed of machines lit up everything the company had not yet reconciled. \[E3\] ## Love, from the other side of the counter Customers never see the org chart. They do not know that the returns policy belongs to finance, the chatbot to service, the promise to marketing and the product to somebody else again. They experience one thing, and judge whether it behaves like one thing, with one mind and decent judgement. That is what people feel as care. Not warmth or tone of voice, not the owl. **Coherence.** \[E4\] > It is also what the machines will measure, whether or not anyone intends it. An agent comparing you to a rival is not reading your values statement. It is reading what happened to ten thousand people at your seams, in their own words, on the forums the model was trained on. ## A scene from inside I want to make "systems design" concrete, because as a phrase it is worth nothing, and every consultancy in the world now says it. Years ago I was handed LEGO.com. It was the victim of three prioritisation systems that could not talk to each other. Enterprise IT ran a waterfall process on corporate priorities. Our internal agency ran on the marketing calendar and the product launches. The markets ran on their own launch dates and targets, tied to the global campaigns but never the same as them. Each was reasonable. Together they guaranteed that whatever the customer saw arrived late, or generic, or both. Nobody was wrong, and nobody owned the join. I lost the first year trying to reconcile them. I clarified who decided what, redrew the processes, reshaped my own team. Each value stream made perfect sense from inside the part of the company that ran it, and none was going to bend for a team that happened to sit where all three met. What worked was to stop reconciling and move the decision up a level, to where all three could find common ground: one shared goal for the site, and decisions made on principles rather than the minutiae each stream was fighting over. The object came first. A campaign builds things to be thrown away: a page or a game, made for one launch and written off when it ends. We separated the functionality from the skin, so the same thing could be reused and reskinned for the next campaign at a fraction of the cost. An asset built once and discarded is an expense. A component that three markets reuse and next year's campaign reskins is an investment, and every campaign after it got cheaper and faster because of the ones before. Then the rule, which is the part worth copying. Nothing got built unless at least two markets wanted the functionality and it could be reused in a future campaign. A person could apply that rule in a meeting and a machine could apply it in a queue, and it gave the same answer either way. Then the yes. I decided what got built next, ranking the candidates on the value they would compound over time, their fit with the global marketing calendar and how far they carried the strategic objectives. Nobody lost a veto. IT kept its process, the agency its calendar, the markets their targets, and money could now reach the site from both directions, local and global, to fund the same components. The three value streams went on existing. They just stopped colliding at the customer, because a decision about a hundred small things had become a decision about a few principles, and all three could sign up to those. In Meadows' terms nothing at the bottom of her list moved. A shared goal, a rule, a flow of money, an owner at the join: the top of it, reached by design rather than restructure. \[L1\] That site still carries over forty million people a month, and it has not needed me for years. ## What it takes So here is the mechanism, for the leader who takes this upstairs and finds that nobody in the room thinks it is their budget. The chief executive hears a service problem and the finance director hears a cost line. The argument is about wiring, and it belongs to all of them. - **Start with the number.** If the only figure every function shares is revenue, the seams are where everyone hides, because a cost pushed over the join to the next team still counts as a win. *One measured goal* that crosses the hands changes that arithmetic before anything else does. To finance, this is not a new cost. It is the end of paying twice for the same customer, once to promise and once to apologise. - **Then the principles.** *A person at the counter and a model at the counter need the same finite set to decide by*, because a script breaks the moment the case is not on the list, and the machine will find that case within the hour. Two markets or it does not get built: short enough to remember, hard enough to lose an argument to. Finite rules, infinite expression. - **Then the flow back up.** *What a customer says at the surface has to reach product, pricing and policy as something that can change a decision.* In most companies it arrives as a ticket, and a ticket gets closed. And it is not only complaints. What your fans make and say is the corpus the buying machine reads, so the people who love you are already writing your brand's entry in the comparison. The question is whether anyone inside reads it with the power to act. - **And somebody has to own the seam itself**. Brand owns the campaign, service owns the bot, commerce owns the shelf, and the join between them has no name on it. The Air Canada defence is what nobody-owns-the-join sounds like when a lawyer has to say it. *The owner should be the marketing leader, because marketing makes the promise and hears the customer first,* which puts it at the join whether it wants to be there or not.That is an operating job, not a communications one, and it comes with a budget line marketing has never carried: the cost of every promise the business makes and then fails to keep. - **Design is one part of this. The rest is culture**: how people are rewarded, whether anyone is allowed to say that the promise and the policy disagree, and what happens to them when they do. I did not learn that in marketing. I learned it across seven roles and six functions, watching value pass through many hands and asking why it arrived intact in some places and shredded in others. ## What comes after the diagram There is a temptation, once you accept all this, to draw the system. A study published this month reviewed forty-six guides to changing systems, from 2014 to 2025, across philanthropy, government and development. Few cited each other. *A third said nothing at all about how you would know whether the change had held*. Even the discipline built to change systems has spent a decade drawing them and a fraction of that time checking whether the drawing worked. \[E5\] The map is the beginning. The seams get closed on a Tuesday, in a budget meeting, by someone with the standing to say that the promise on the website and the policy in the back office will agree from now on, and here is who decides when they do not. If you want to know whether you have such a person, ask one question first thing. *When did something a customer said at the surface last change a policy underneath, and who made that happen?* If nobody can name the occasion, you have found the seam, and found that nobody owns it. That is the work. It does not photograph, and it is the only thing the machines cannot price. ## The searchlight is already on I am leaving the LEGO Group this week, after twenty-five years inside a brand that other people, not I, have measured as the most loved and reputable in the world, and that is loved beyond any reason I could ever have put in a spreadsheet. I have watched what that love survives and what it does not, and I have never been more certain of the answer than I am now, with the buyer turning into a machine.\[L2\] The seams will show whether you fix them or not. AI only decides how soon. --- ### Sources **Evidence** - *Moffatt v. Air Canada*, 2024 BCCRT 149, British Columbia Civil Resolution Tribunal, 14 February 2024\. Negligent misrepresentation; the "separate legal entity" argument rejected. [\[Link\]](https://decisions.civilresolutionbc.ca/crt/crtd/en/item/525448/index.do?ref=ceciliaweckstrom.com) - On 28 April 2025, Duolingo CEO Luis von Ahn announced an “AI-first” strategy in a memo published on [LinkedIn](https://www.linkedin.com/news/story/duolingos-ai-growth-sees-backlash-7282466/?ref=ceciliaweckstrom.com), including a plan to phase out contractor work that AI could perform and to make AI capability part of hiring and performance assessment. Days later, Duolingo announced 148 new courses, saying that generative AI had enabled it to produce in about a year what had previously taken far longer. The policy triggered substantial online backlash; [Fast Company](https://www.fastcompany.com/91338068/duolingo-deletes-tiktok-ai-backlash-returns-with-strange-message?ref=ceciliaweckstrom.com) reported that the company temporarily wiped its TikTok and Instagram posts, while Ad Age was cited in contemporaneous coverage as reporting losses of more than 400,000 TikTok followers. In an August 2025 [New York Times ](https://www.nytimes.com/2025/08/17/business/duolingo-luis-von-ahn.html?ref=ceciliaweckstrom.com)interview, von Ahn said Duolingo was still hiring at its previous pace. [\[Link\]](https://theoutpost.ai/news-story/duolingo-embraces-ai-first-strategy-plans-to-replace-contractors-with-artificial-intelligence-14789?ref=ceciliaweckstrom.com) - Klarna: February 2024 announcement that its AI assistant handled the work of 700 agents; Sebastian Siemiatkowski to Bloomberg, May 2025, on cost as "a too predominant evaluation factor" and the return of human support. [\[Link\]](https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/?ref=ceciliaweckstrom.com) - Sinch, 2026: 74 percent of companies that deployed a live AI customer agent have rolled one back; 81 percent among those with mature guardrails. The number behind "more governance is not the answer". [\[Link\]](https://sinch.com/blog/ai-chatbot-failures/?ref=ceciliaweckstrom.com) - Emily F. Gates, Clara Shim, Nishat Akhi and Stephen Gyan, "Decoding a Decade of Systems Change Guidance, 2014–2025", *Systems Research and Behavioral Science*, 17 September 2026\. Forty-six guides; only 18 cite another guide; 16 of 46 do not discuss evaluation at all. [\[Link\]](https://doi.org/10.1002/sres.70121?ref=ceciliaweckstrom.com) **Building on** - Donella Meadows, *Leverage Points: Places to Intervene in a System* (1999). Parameters at the bottom of the list; rules, information flows and goals near the top. The scale the Klarna passages and LEGO.com passages are read against. Gates et al. (above) find it the most-cited work across the whole decade of systems change guidance, which corroborates leaning on it here. [\[Link\]](https://donellameadows.org/archives/leverage-points-places-to-intervene-in-a-system/?ref=ceciliaweckstrom.com) - Talkwalker, *Love Brands* report (2020): the LEGO Group ranked the world's most loved brand. RepTrak, *Global RepTrak 100*: the LEGO Group ranked first for reputation in 2024, 2025 and 2026\. [\[Link\]](https://www.reptrak.com/globalreptrak/?ref=ceciliaweckstrom.com) - Kevin Roberts, *Lovemarks: The Future Beyond Brands* (powerHouse Books, 2004). "Loyalty beyond reason", the idea this piece dates and extends. - PHD × WARC, *From Abundance to Agents* (July 2026). The delegation thesis with the industry's own numbers on it; the report stops where this piece starts. [\[Link\]](https://www.phdmedia.com/new-phd-x-warc-research-from-abundance-to-agents-how-the-delegation-of-choice-is-transforming-marketing/?ref=ceciliaweckstrom.com)