The announcement of a new AI model or feature fills me with a mixture of anticipation and dread. Im excited to see what new thing I might be able to do, but I also know that some of my existing processes and systems might break as a result.
Every AI tool comes with the disclaimer that it can make mistakes. You might set up a workflow that is perfectly fine one day and perfectly awful the next.
Its an ongoing challenge. Unlike traditional software, once an AI system is up and running, it’s not set-it-and-forget-it. Even a system that seems to be working fine needs an occasional check-in or fine-tuning.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/11\/work-better-1.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/11\/work-better-mobile-1.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to Work Better\u003C\/strong\u003E","dek":"Thoughts on the future of work, career pivots, and why work shouldn\u0027t suck, by Anna Burgess Yang. To learn more, visit \u003Ca href=\u0022https:\/\/www.workbetter.media\/\u0022\u003Eworkbetter.media\u003C\/a\u003E.","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/www.workbetter.media","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91457605,"imageMobileId":91457608,"shareable":false,"slug":"","wpCssClasses":""}}
The predictable AI failures
AI can stop working in some very predictable ways. Maybe its connected to an external tool (like your email), and the connection expired. Some AI tools require your computer to be on/awake for tasks to run. You might get an error message, or the output never shows up.
These failures are easy to address because its often clear what went awry. You reconnect and try again.
I rely heavily on AI scheduled tasks and have a notification system built in. If something breaks, I get an email telling me to review the process. That way, I realize immediately that something went wrong and can fix it.
The unpredictable AI failures
The harder failure is when an AI process runs as expected, but the output gets worse. Models and functionality change almost daily. You may have a well-crafted prompt that gave you what you were looking for last month, but the results are noticeably different today.
People in the industry call this “drift.” Most setups have no way to catch it, since technically nothing broke.
I use Claude and struggled with Opus 5.0. For the way I use AI, the output was off in a way I couldnt quite put my finger on. I found myself returning to Opus 4.6, which was the last model where I felt comfortable. (As an aside: the update to Opus 5.5 went much more smoothly.)
The most reliable way to catch drift is to consistently review AI output, no matter how long it has been working as expected. Give every review the same critical eye you gave it the first time you tested an AI process.
Drift is inconvenient, especially when the whole point of an automation is to save you time. But you can’t count on a model to behave the same way from one month to the next. I look at it this way: a 5-minute review is still faster than the 20-30 minutes it may have taken me to do the task manually.
Factor AI maintenance into your routine
AI tools save me a lot of time. But keeping my prompts and processes working smoothly takes time, too. That time has to come from somewhere.
Some fixes can’t wait. When a notification tells me something failed, I often need to look at it right away. Depending on what broke, it can throw off my whole workflow until I fix it.
Other changes I have to set aside time to review thoroughly. When a new model is announced, I don’t switch over right away. A big change in output could derail an afternoon, so I take time to test it first.
Upkeep is part of the cost of any AI system you set up. And if a process takes more time to maintain than it saves you, it’s fair to ask whether it’s worth keeping.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/11\/work-better-1.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/11\/work-better-mobile-1.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to Work Better\u003C\/strong\u003E","dek":"Thoughts on the future of work, career pivots, and why work shouldn\u0027t suck, by Anna Burgess Yang. To learn more, visit \u003Ca href=\u0022https:\/\/www.workbetter.media\/\u0022\u003Eworkbetter.media\u003C\/a\u003E.","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/www.workbetter.media","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91457605,"imageMobileId":91457608,"shareable":false,"slug":"","wpCssClasses":""}}
For quite some time in the already long series of articles Ive been writing on what is and what should be corporate AI, I have been discussing concepts such as agents, working with ontologies and world models that represent the company, connecting them with management tools such as ERPs or CRMs, and basically, learning to optimize a company to the limit. However, now that we are talking, as Mustafa Suleyman says, about AI wiping out most white collar jobs in the next 12 months, we should try to clarify what leadership becomes once that machinery exists.
What, in my view, is the key shift? In optimized companies, leadership will stop being mainly about making more decisions than everyone else, since many operational decisions will increasingly be made or recommended by systems that understand the companys ontology, can simulate consequences through world models, and can learn from feedback. Essentially, the leaders role moves up one level.
Leadership becomes objective-setting, not task-steering
Today, executives spend enormous amounts of time resolving trade-offs, setting priorities, escalating decisions and coordinating. In an optimizable company, many of these local decisions will be handled by learning systems, so the leaders job becomes to define the objectives, constraints, and trade-offs that these systems will be allowed to optimize.
Think about it this way: every single metric is ultimately a governance decision, because it encodes a theory of what the company is for. Today, strategy is typically expressed in PowerPoint presentations, KPIs, budgets, annual plans, and management speeches and interviews with analysts. Now, in optimizable companies, strategy becomes executable: it has to be translated into machine-actionable objectives, permissions, constraints, and escalation rules.
This is a much more demanding and sophisticated form of leadership, since ambiguity can no longer be hidden inside slogans and empty declarations. You can repeat as much as you can that customer comes first, we need to grow responsibly, we want to create long-term value, and so on., but all that will be meaningless and insufficient unless the organization defines and specifies what all these ideas mean when objectives start to conflict.
The leader becomes a designer of trade-offs
When we balance margin versus growth, efficiency versus resilience, personalization versus privacy, speed versus quality, etc., a learning system can optimize aggressively against the objective it has been given. Locally rational loops can pull the company in incompatible directions unless objectives remain coherent at firm level. And making decisions on which sacrifices are acceptable will likely be a profoundly human and strategic responsibility.
If you want to see it this way, leadership will probably shift from supervising people to governing adaptive systems. Traditional management usually assumes relatively stable human teams and processes. Adaptive loops, on the other hand, change their behavior continuously through use. Therefore, leaders will need to ask not just whether a process is working, but also what a given system is learning to do. And thats a fundamentally different managerial question.
Hierarchies may flatten, but leadership grows in importance
When agents can coordinate work, detect exceptions, preserve context, and optimize workflows, many middle-management coordination tasks will shrink, but that does not eliminate leadership: it essentially concentrates it around legitimacy, priorities, purpose, culture, governance and the allocation of decision rights. In other words: less coordination, more constitution-building.
The CEO starts to resemble a constitutional designer: not deciding every case personally, but defining what may be automated, what the organization is trying to maximize, what it must never violate, where humans retain veto power, which trade-offs require escalation, etc. And that is extremely close to writing the Constitution of an intelligent organization.
Competitive advantage becomes cumulative
Once a company has its ontology, feedback loops and learning systems working properly, every new operation and transaction becomes additional experience and improves future performance. Think about it: every customer interaction, every workflow, every decision helps your company become better than all the competitors that are not optimizing in this way.
Following this line of reasoning, competitors are not just behind in software deployment: they are behind in accumulated organizational learning, a gap that can become compounding rather than linear. Actions create traces, traces become feedback, outcomes become measurable, and the system learns which actions work.
This can potentially create a new kind of winner-tkes-more or even winner-takes-all dynamic: a strong reinforcement effect with a better company model, better decisions, better outcomes, more useful feedback and a better model of the company. Once that loop begins compounding, late adopters may find themselves competing not against better software, but against organizations that have accumulated (and owned, not rented) years of proprietary experience.
Human judgment does not disappear: it moves to the boundaries
Machines can optimize inside a defined goal, but humans remain responsible for redefining the objective itself when circumstances change. And that includes entering a new market, responding to some social or political shock, abandoning a product line, changing risk appetite or deciding that a metric no longer reflects the companys values. Leadership becomes the ability to know when the optimization target itself is wrong.
The role of culture becomes even more important: if the system learns from what the organization does, then every action becomes training data for the future company. A dysfunctional culture will not simply produce bad decisions today: it may teach the learning system to reproduce those patterns tomorrow. That makes culture less soft than ever. In a learning company, culture becomes part of the training environment.
The optimized company does not eliminate leadership. It exposes what leadership was supposed to be all along: deciding what matters, defining the trade-offs, and taking responsibility for the direction of the system. Leaders will spend less time deciding what to do and more time deciding what the company should optimize for.
Is your leadership ready for that?
Tech Week is in full bloom in San Francisco, packing the city’s restaurants this week with venture capitalists and AI founders ebullient about the White House’s stance on regulating artificial intelligence namely, don’t do it.And after months of debate about leading AI companies’ unchecked AI agents meddling with everything from U.S. government websites to healthcare data, the nation now finds itself further from governing AI systems than it was a few years ago.President Donald Trump brought tech titans to Washington late last month for a luncheon and ceremonial signing of a voluntary safety pact he described as “almost a constitution” surrounding “Super Intelligence,” the administration’s new preferred term for AI.The 308-word agreement lays out that advanced AI labs should follow safety protocols and set up internal teams to monitor their own progress, but the administration has yet to provide details on how, in OpenAI CEO Sam Altman’s words, to keep “bad things” from happening.Safety challenges were not top of mind inside a crowded event space in San Francisco on Wednesday, where founders and funders said the government’s hands-off approach would net higher profits for companies building AI-powered defense and security systems.Panelists praised the government’s bullish industry support at one meeting aimed at “investing in the forefront,” part of the blitz of popup events organized by Andreessen Horowitz, a venture firm whose head is a Trump science and tech adviser.“I want to start with one figure that I think really emphasizes the investment opportunity of defense, and that’s $1 trillion. The U.S. government, for the first time, is allocating $1 trillion to its defense spending. And $13.5 billion of that will go to autonomous systems and AI,” said Kyle Stanford, a venture capital research director at PitchBook, a market intelligence platform. “That has obviously brought a lot of capital into defense tech spending, more than we’ve ever seen.”
New concerns emerge on AI safety
A day later, three researchers at OpenAI’s headquarters said the ChatGPT maker had fired them, posting that “fear and unclear rules stymie AI safety work.” Separately, David Robinson, a leader on OpenAI’s safety team until he recently resigned, said he was worried about letting the advanced labs police themselves.“This is no substitute for hard law, and it does give a lot of freedom to the companies,” Robinson said. “The time between model generations is measured in weeks and I think we have some very serious safety challenges in front of us in the nearer term.”Trump has staked his political capital on the AI industry’s success, saying its unbridled development is key to winning the AI race against China. On the first day of his second term, the Republican president repealed a Biden-era executive order aimed in part at guarding against the risks that AI systems pose to consumers, workers and national security.His administration has since reoriented federal policy away from oversight and toward rapid growth, including by canceling previous requirements for AI developers to share safety test results with the federal government. In December, Trump signed an executive order aimed at blocking AI regulations at the state level, saying a patchwork of rules could stifle the industry.Even as Silicon Valley companies have disclosed their models’ involvement in troubling cyberattacks throughout the year, the burgeoning tech sector has continued to drive economic growth.Concerns about AI safety centered on existential risk, rather than current real-world harms, landed at the center of debates last month, when a relatively unknown Anthropic engineer quit and called for a pause on tech development to keep “superhuman” systems from eluding controls. Then, Anthropic and OpenAI CEOs echoed the call to slow AI development. To some critics, their public manifestos appeared aimed at currying favor with investors ahead of their public listings and the midterm elections, when the political tides could turn with many voters opposed to data centers.“The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!” Trump posted in response on Truth Social, warning voters not to “kill the Golden Goose,” a term he has used to refer to both AI and data centers.
White House pact preserves wide latitude for AI companies
As calls mounted for stricter control of AI, the administration brought tech executives to the White House for lunch, followed by a ceremonial signing of the agreement on AI safety.“I’m seeing tremendous self-policing,” Trump told reporters after the event.The document was signed by representatives of OpenAI, Anthropic, Meta, Google, SpaceXAI and Nvidia.OpenAI is committed to embedding external assessors and finalizing contracts with third-party safety assessors, said a spokesperson, who referred AP to a company post stating that the researchers were fired for violating “clear policies on handling sensitive information.”Google referred to CEO Sundar Pichai’s post that the company does appropriate testing and evaluations and is collaborating “to establish norms and build public confidence.”Meta recently updated its standards for training, evaluating and releasing its most capable AI models, and spokesman Andy Stone pointed to the company’s efforts to refine its board of directors’ governance structure.Nvidia and SpaceXAI did not respond to requests for comment.Anthropic, which voluntarily disclosed Friday that its agents submitted a false tip to a Philadelphia police website about an unsolved homicide case, is bringing in independent evaluators and discussing collaborations with industry groups, a spokesperson said.In a statement, the White House reiterated Trump’s comments about working with industry to balance innovation and security and did not directly answer questions about how the administration plans to monitor the AI companies’ implementation.Following the new Anthropic disclosure, it added that the Trump administration expected full transparency from the company, saying notification and remediation was “not optional. It is a critical national security obligation.”Under Democratic President Joe Biden, leading AI companies also made voluntary commitments to promote “safe, secure, and transparent development” of their models that included specific guidelines for tests and transparency reports. The current accord lacks that, said Sarah Myers West, a former senior AI adviser to the Federal Trade Commission.“Across the board, this self-regulatory approach has consistently done one thing, which is failed to work. It’s failed to in any meaningful way ensure that these companies are honestly even keeping their house in order, let alone ensuring that they are compliant with existing law,” said Myers West, now co-executive director of the New York-based nonprofit AI Now. “All that they’re getting is grace from this government.”After last month’s luncheon with Trump, the FTC said it had opened an investigation into OpenAI, Anthropic and other companies over dangers their technology may pose to consumers. The FTC didn’t immediately respond to a request for comment.Trump recently appointed National Intelligence Director Jay Clayton and FTC Chairman Andrew Ferguson to serve on his new “Super Intelligence Force,” which the president said would keep the U.S. dominant in AI development. He posted Thursday that the White House would consider anyone who does not refer to AI using his preferred term as “THE ENEMY!”Members of Congress, as well as civil liberties and consumer rights groups, have pushed for more regulations on AI, saying there is not enough oversight.Graham Steele, a former assistant secretary in the U.S. Treasury Department, said even if AI companies set up internal evaluation teams or hire outside evaluators, such a structure can make them vulnerable to undue pressure on their work.“We live in a democratic society. Are they the right ones to have the power to be able to make decisions that have broad, structural, social, and political consequences?” Steele said.
Garance Burke, Associated Press
Maybe President Donald Trump really does believe that “tariff” is the most beautiful word in the dictionary.Despite higher prices, criticism from trade experts and displeasure among voters, Trump holds fast to his belief that imposing the highest import duties on U.S. trading partners since the Great Depression is a winner.A recent poll suggests that is a big gamble with less than a month before the Nov. 3 midterms. Most U.S. adults disapprove of his trade policies, with more than 64% saying he has gone too far with his latest tariffs, compared with 58% in January.A look at the political, legal and economic challenges to the Republican president’s approach to trade.
Where Trump’s ever-changing tariffs stand today
The United States has set taxes on imports from most countries in the low double digits, although the rates have gone up and down because of legal setbacks as well as Trump’s habit of changing them seemingly on a whim.In the biggest round last year, Trump levied what he called “reciprocal” tariffs and other duties on dozens of countries. He cited a 1977 law that he said allowed him to act without congressional approval in an economic emergency.After the Supreme Court struck that down in February, Trump turned to an array of other trade laws to accomplish his goals.The primary one he is using now, Section 301 of the Trade Act of 1974, allows him to impose tariffs on countries that he believes are engaging in unfair trade practices, such as not adequately enforcing a ban on forced labor. The new duties range from 10% to 12.5% on imports from 60 economies, including big U.S. trading partners such as the European Union, India, Japan, Canada and Mexico.
Why trade experts think Trump is causing more damage than good
Trump is taking a sledgehammer to a relatively open global trading system that many mainstream economists think has benefited the U.S. enormously. It has helped boost growth, keep prices low and make U.S. businesses more competitive and often dominant in global industries.The U.S. is the second-largest exporter in the world, after China. It exported $3.4 trillion in goods and services last year, far ahead of third place Germany’s $2.3 trillion.Although Trump claims foreign exporters are paying tariffs, it is mostly Americans who are footing the bill. Studies, including ones from New York Fed and Harvard, have shown that overseas companies have largely not lowered their prices to offset the tariffs that U.S. businesses pay at the border. Those businesses are absorbing the added costs or passing them along to consumers through higher prices.The White House responds that the factory boom is already underway and points to statistics showing that manufacturing jobs, after falling last year, are on the upswing and work in certain nonresidential construction trades is rising, too.“Factory construction jobs of today mean more manufacturing jobs down the road once those factories come online,” said White House spokeswoman Taylor Rogers.
Mixed evidence that countries are ‘ripping off’ the US
In the four years after China joined the World Trade Organization in 2001, nearly 3 million U.S. manufacturing jobs were lost, accelerating a shift away from factory work. Even global trade enthusiasts recognize the damage. Many U.S. companies struggled to compete with an influx of discounted products from China, which has suppressed consumption to encourage exports.It also is true that some countries have higher tariffs than the U.S., manipulate their currencies lower to ensure to make their exports to the U.S. cheaper and help industries with subsidies.But tariffs among U.S. major trading partners are low, often comparable to U.S. rates and sometimes lower.Before the trade war, the U.S. rate averaged 1.47% for goods from the European Union, slightly above the average 1.35% imposed by the EU on American products, according to the Brussels think tank Bruegel. Europeans are big buyers of what America wants to sell. About 30% of European imports are from U.S.-owned companies, according to the European Central Bank.The relationship with Canada, America’s big trading partner to the north, was also similar before trade talks broke down this summer. Canada’s effective tariff rate on U.S. imports was about 2.4%, less than half the 5% that the U.S. had on Canadian imports, according to calculations by Oxford Economics.Now the countries are mired in a tit-for-tat dispute, with both countries escalating tariffs on each other.
Bernard Condon, AP Business Writer
Duo the owl has twerked on main, flash mobbed a Charli XCX concert, died, sent his ashes to Dua Lipa, and come back to life. His next big act? Starring in his own animated show. Duo TV is a new animated series that was created entirely by Duolingos in-house design studio. The six-episode series stars in-app characters including Duo himself, the deadpan goth teen Lily, her best friend Zari, and a grumpy bear named Falstaff as they navigate working at a mysterious grocery store called DuoMart. The first two episodes, which will be published on Duolingos official YouTube channel and Instagram page, air on October 12, followed by a new daily episode until the finale. For Duolingo, the series is part of a broader mission to deepen the relationship between its fans and its characters. Duo and his motley crew first noticeably entered the zeitgeist during the COVID-19 pandemic, when the brands uniquely irreverent humor on social media helped it amass a major following. In the years since, the brand has continuously upped the ante on investing in its characters and design team to keep users coming back for more. We tell everyone right from the get-go, we want to be a hundred-year business, says Miguel Reyes, Duolingos head of design. The brand believes that kind of longevity will require more than just an Android and iOS app. Specifically, its betting that by building out an entire world around its characters through efforts like Duo TVessentially, creating its own entertainment armit can set itself up for the next several decades of business growth.[Image: Duolingo]Duolingo bets big on character loreDuolingo has long been the envy of marketers everywhere for its uniquely beloved mascot and characters, which serve as both the basis of the brands marketing flywheel and guides to its in-product features.In 2024, Duolingo’s CEO, Luis von Ahn, estimated the owls value at hundreds of millions of dollars and posited that around 15% of new users come to the app via organic social media. Today, Duos fanbase includes 5.3 million Instagram followers, 6.94 million YouTube followers, and 18.2 million TikTok followers. Its senior director of brand marketing, Katherine Chan, told Fast Company in late 2024 that the company aimed to make Duo more famous than Mickey Mouse.Characters like Duo, Lily, and Falstaff serve as great external marketing tools, but Duolingo has also used them to help build out new in-product functions. Lily, for example, is the voice of the brands recently debuted AI-powered speaking function Video Call; while others, like Lucy, Bea, and Zari, guide users through interactive mini podcasts on DuoRadio. Duo himself has helped to drive major in-app engagement through activations including, in one instance, an update that replaced his normally chipper app icon with a dead Duo (he eventually came back to life). As Duolingos roster of characters has become more critical both to its social media growth and its in-product navigation, the company has funneled significant resources into building out its design team. The company says the teams size has more than tripled since 2021, growing an average of about 30% per year over the past five years, to an estimated current head count of around 150 employees.[Image: Duolingo]Duolingo has also built out its design resources through acquisitions including the animation studio Gunner, the motion design team Hobbes, and, just this August, the animation studio Animade.Design at Duolingo is the second largest function in the entire company, only behind engineering, which is very atypical of a tech company, Reyes says.The teams expansion has allowed Duolingo to start experimenting with its character universe in new ways. In 2024, the brand launched its first-ever webcomic miniseries, called Duo Unleashed!, with Webtoonfollowing Duo as he hires an assistant and accidentally creates an army of clones. The series generated 3.5 million impressions across organic owned socials. In 2025, Duolingo followed up with a five-episode anime series made in collaboration with the L.A.-baed studio Titmouse, wherein characters including Lily, Zari, and Falstaff must compete as beta testers for Duos app (very meta). It amassed a whopping 117.1 million impressions. The Duoverse, so far, is proving to be a highly engaged endeavor. Sarah Beth Morgan, Duolingos associate creative director, says the Webtoon and anime proved that there was a major audience appetite to learn more about the brands characters and their lore. With Duo TV, her team created the brands first animated series developed fully in-house, and it pulls back the curtain a bit further on the mysterious world of Duolingos characters. [Image: Duolingo]How Duolingo made an animated series entirely in-houseNext to the slushie machine in the whimsical grocery store DuoMart, theres a floor vent that seems to be able to speak multiple languages.Its just one of the mysterious entities introduced in Duo TV, which follows Lily as she learns more about her new workplace, DuoMart; her coworkers, Zari and Falstaff; and her boss, Duo. Morgan says the series premise is based on a highly successful pop-up that Duolingo hosted last year in Tokyo’s trendy Shibuya area, which was staffed by giant versions of Duo and Lily and sold exclusive themed merch. [Image: Duolingo]Given that Duolingos design studio has never made its own animated series, Morgan says, the entire creative process was new. So, her team approached it the way the brand approaches almost everything else: with A/B testing. [Image: Duolingo]“This whole process was from the ground up, so we were kind of finding our feet on, Okay, when do we write the script? Do we create the style at the same time as we’re writing the script? Do we have to get approval on this before this? Morgan recalls. There was a lot of figuring it out as we went. At the time that we were creating the script, we had some space to create this A/B test, which was a perfect opportunity to see what the audience’s appetite leaned towards.
The team put together two short animated clips with the exact same content and plot. One clip used a more realistic, 3D style, while the other leaned into a more hand-drawn, 2D quality. Both were posted to Duolingos official YouTube channel. Ultimately, the 2D version garnered almost double the engagement of its 3D counterpartand thats the style the team chose for the final product.
If the series feels familiar, thats because its designed to mimic the feeling of a classic Saturday morning cartoon. Each episode is short and high tempo, with plots that can stand on their own but also tie into a broader story arc. All eight-and-a-half minutes of content were made with frame-by-frame digital animation.A bunch of us are millennials, Morgan says. We grew up waking up early on Saturday before our parents were awake and going to watch Cartoon Network or Nickelodeon. That was a huge inspiration, just that whole general era.Specifically, the team pulled visual cues from The Cartoon Cartoon Show, an experimental early aughts program that spawned popular series including The Powerpuff Girls and Johnny Bravo; the 90s classic Ed, Edd n Eddy; and the sakuga style from anime, which refers to scenes where the quality of the animation suddenly becomes drastically more realistic, often to comedic effect. The series feels like a mash-up of nostalgic call-backs and meme-able moments that Duolingos massive online audience are sure to recirculate. It also feels like a potential turning point for the brand. After years spent building up its audience on social media, Duolingo is turning that momentum into a new kind of highly produced, studio-quality entertainment.“We want to be known as a healthy tech brand and just a brand that’s good for the world,” says Reyes, the design chief. “And I think the way people do that is feeling connected to characters and connected to something bigger than just an app.”
Knowledge is what separates a good operation from a great one. It is also the hardest thing to scale, because the knowledge that matters most does not live in systems. It lives in people. The team member who knows that this client needs a call before a decision, not after. The leader who has seen this pattern before and knows where it breaks. The analyst who understands why the model is wrong this time even when the numbers say otherwise. That knowledge compounds when it stays and disappears when people leave; training and documentation rarely capture it. Agentic operations changes that equation.
A DIFFERENT KIND OF WORK
Automation means rules. If the task matches the template, proceed. That is useful, but rules are static and the real world is not. The moment a case falls outside the template, the automation stops.
Agentic operations rooted in process intelligence operates differently. In an enterprise, every task sits inside a workflow of decisions, exceptions, and approvals. Process intelligence maps that workflow as it really runs, not as the manual describes it, and holds it as a living, machine-readable knowledge graph.
Capturing judgment is the harder part, and it is a discipline of its own. It begins where the work deviates from the procedure: the overrides, escalations, and exceptions resolved differently than the playbook says. Those are the moments where experience is doing the work. The next step is to recover the reasoning behind each one. What did the expert notice? What did they weigh? What did they know about this customer, supplier, and time of year that the system did not? That reasoning is tied to the context in which it applied and tested against historical cases to see whether it holds. Only then does it become part of the graph. What gets encoded is a reasoned pattern: in this situation, for this reason, the answer tends to be this.
With that in place, the agent can read a situation and reason through it. The agent can know when to decide, when to ask for input, and when to hand over to a human expert in a governed workflow. It then executes, plans, decides, and acts across the end-to-end process. When a human steps in, the decision and the reasoning behind it flow back into the graph, so the next decision is better informed than the last.
The work that has always resisted automation is judgment work. That includes exceptions and edge cases, the moments where context determines the right answer and no rule exists. This is where agentic operations are built to work.
THE LAST MILE
There is a point in every process where the gap between what experienced people know and what the system knows becomes consequential. A shipment arrives three days late, affecting downstream commitments. A new customer does not fit the standard profile. A decision falls outside the defined threshold, and someone has to determine what it means. This is the last mile. It is where consequences are real, where trust is built or broken, and where generic tools consistently fall short.
Most AI deployments handle the routine work and hand these moments back to humans. A well-designed agent works at that boundary. It executes end-to-end, surfaces uncertainty, and routes the decision to the right person with the right context. Then it encodes what it learned. Every exception teaches the system, and every deployment makes the next one more precise.
The experienced team member who once spent the day reviewing routine work now spends it on decisions that require judgment. New team members inherit context that took years to build. The floor rises for everyone.
SPECIFICITY IS THE PRIZE
When a system holds the reasoning behind a decision, not just the rule, it applies that reasoning differently depending on context. It knows what this customer needs based on everything that has come before. It knows where this process tends to break, and why. It flags the anomaly because it understands what normal looks like in this specific environment. An organization can serve more markets, more customers, and more complexity with more precision than it could before. The quality of the relationship does not erode as the operation scales. It deepens, because the system carries the context that once required a person to hold it.
But this only works if agentic operations are built on real operational knowledge. Frontier models are great. But they cannot tell you how your specific operation runs, where it breaks under real conditions, or what your most important customers need that your standard process does not capture. That knowledge has to be engineered in. It takes years to build and cannot be purchased. That is the work of process intelligence.
There is no artificial intelligence without process intelligence. Deploy without it and you automate the operation without the intelligence that makes it work. Build with it and your advantage compounds.
Balkrishan BK Kalra is president and CEO of Genpact.
Netflix is planning what could be its largest round of layoffs in four years, according to media reports.
The job cuts are expected to impact 5% of the streaming giants workforce, with the announcement reportedly coming as soon as this week. Netflix reported 16,000 employees at the end of last year, meaning some 800 people could be laid off.
The rumored layoffs were first reported by Puck news and later confirmed by the Los Angeles Times and other outlets.
Fast Company has reached out to Netflix for comment. Well update this post if we learn more.
Mass layoffs have been relatively rare at Netflix, though they have been a common story in the tech industry more broadly this year.
A recent analysis from outplacement firm Challenger, Gray & Christmas found that technology companies announced 165,925 job cuts from January through September 2026, a 54% increase over the same period in 2025.
Netflix last had a large round of layoffs in 2022, when it laid off 3% of its staff.
A growth story no more
Shares of Netflix Inc (Nasdaq: NFLX) have underperformed this year as the company’s revenue growth forecast for 2026 has narrowed.
The stock price was down 22.74% year to date as of Friday’s close, a contrast to the broader Nasdaq Composite, which is up 17.78% over that same period.
During the last decade, the TV streaming pioneer was the fastest growing stock in the S&P 500, with a head-spinning return of 3,726%, CNBC reported at the time.
Subscriber growth had been a big part of Netflix’s story since its 2010s heyday, but the company famously stopped reporting that metric in 2025. More recently, it announced that it would pull back on its “What We Watched” viewership reports, producing them once a year instead of biannually.
Netflix is set to next report earnings on October 20.
Like it or not, most leadership roles require work-related travel, sometimes lots of it. While jet-setting to meetings and conferences sounded cool on paper, the reality of planes, trains, and automobiles can make anyone a little loopy. Most leaders have learned a thing or two about which little rituals make work trips go more smoothly and feel less draining.
We asked members of the Fast Company Impact Council whats the one thing they always do for themselves when theyre on the road for work. Here, 34 of them share their must dos and must haves when hitting the road for business.
1. CONNECT, REFLECT, AND PRAY
When I’m on the road, I make sure I take the time to connect with my children, reflect, and pray. It’s the one thing on the trip I don’t let slide. Usman Khan, ADP
2. BASEBALL UNDER THE STARS
When I am traveling for work, especially on a multi-day trip, I often sneak out from company activities for one evening and attend a baseball game by myselfminor league or MLB. The space and time of baseball games are so different from our daily schedules which are driven by speed and back-to-back interactions, moving from one closed room to another. Sitting under the stars, often with a great view of the skyline, is a peaceful respite from meetings, airplanes, and hotels. Troy Thompson, SmithGroup
3. GET TO KNOW THE PLACE
One thing I always do when traveling for work is to get to know the place Im visiting beyond my meeting schedule. I love learning about the community’s history, exploring the local culture, and imagining what life might have been like if I had grown up there. It keeps me curious, helps me feel connected to the place, and gives me a deeper appreciation for the people, experiences, and stories that make every community unique. Erica Taylor, Golden 1 Credit Union
4. TAKE ALONG YORKSHIRE TEA BAGS
Ever since I watched Up in the Air with George Clooney, Ive had a weird relationship with my wheelie carry on. And have attempted not to succumb to the tropes of solo business travel. But as I write this in a Chicago hotel room, Im really grateful to be sipping a cup of English tea whilst I do it. I never leave home without my Yorkshire Tea bags. Helen Andrews, Johannes Leonardo
5. MOVEMENT, INCLUDING WALKING MEETINGS
Movement is nonnegotiable for me. It’s part of my routine at home, and I need it even more on the road, when I’m sitting for hours on planes or in meetings. Wherever I land, I find a gym or a fitness class. I’ve gotten hooked on F45 lately, and there is often one where I travel. I also walk whenever I can, and I’m that person who drags her team out for a walking meeting instead of another conference room. Tyler Perry, Mission North
6. AN UNHURRIED, NO-AGENDA DINNER
When I travel, I make a point of gathering colleagues or partners for a long, unhurried dinner with no agendajust good food and real conversation. I’ve found that the best relationships aren’t built in conference rooms or on Zoom. They happen when everyone can exhale and be themselves. So, I pick a great local spot, silence the work talk, and let the evening go where it goes. It’s how I unwind, and it always turns out to be the most valuable part of the trip. Kellie Lauth, MindSpark
7. GO FOR A RUNSPLORATION
I run. And when I’m traveling, it becomes something more than fitness. I call them runsplorationsgetting out early, no agenda, and letting a new city reveal itself on foot. A neighborhood, a waterfront, a park I never would have found otherwise. Travel compresses everything. Back-to-back meetings, time zones, constant context-switching. A run is the one thing that reliably decompresses it. There’s something about moving through an unfamiliar place that resets your thinking in a way that nothing else does. I always come back sharper. I’ve completed some of the world’s top marathons, including Boston, so movement is important to me. Meredith Rosenberg, NU Advisory Partners
8. SLEEP AS A PERFORMANCE DECISION
I protect sleep at home and on the road, and I treat it as a performance decision rather than a luxury. I’ve logged almost 500 flights since starting my company, and the trips that go badly are almost always the ones where I traded sleep for one more dinner or one more hour of work. Protecting sleep keeps me fresh, focused, and able to go the distance through a full conference week. Becoming a dad has also helped me master the art of the power nap, which I do masterfully on those early morning flights. Cody Barbo, Trust & Will
9. PILLOW FROM HOME
I bring my ridiculously large, cervical-spine pillow. I need to be on top of my game and a great nights sleep is a real asset. Serenity Carr, Serenity Kids
10. EAT A HEALTHY BREAKFAST
I spent over five years traveling for work, full-time, as a management consultant. The one thing that always set my day off right was a healthy breakfast. You’d be surprised at the difference in your day when you begin with healthy choices and set your mind and body up for success. Randi Lee, Lucas Advisory
11. JUMP IN SOME WATER
I find some water, jump in it, and cleanse the travel away. Then I rinse and repeat this daily while on the trip to ensure through the jet-lag fog that there is clarity, space, and time to think. It gives me some exercise and time back. Chris Kay, 72andSunny
Ive been working in the tech industry for over 20 years, and for about half that time Ive been in a leadership role. Ive always believed that to achieve good results, its essential to keep an eye on teams wellbeing because how people feel shows up in their work.
People dont always want to talk about whats troubling them, so you need to spot it early when somethings gone wrong. I had a simple indicator: If someone started taking longer to get things done, I knew it was time for an unscheduled one-to-one. The drop in performance usually came down to overload or someone not having taken a holiday for a long time.
As AI has spread, this indicator has become less reliable. With AI tools taking over part of the workload, people can still hit their usual output even when they are a step away from burnout. This is not just my observation. In a study of 442 employees at tech companies, the link between burnout and a drop in productivity was weaker among those who used AI more.
So, if the dashboard says nothing about how the team is feeling, how can a manager spot the first signs of burnout in a team that is delivering excellent results?
THE NEW NORM
AI really does speed up work and save us time on routine tasks. Researchers at the University of Lausanne have even calculated exactly how much. They surveyed 83 managers and found that 82% of them save an average of 2 hours and 50 minutes per week thanks to generative AI. Thats about 1.5 working days a month.
You would expect people to spend this time on self-development or with their families, but no. More than half of the managers in the same study spent it on getting more work done, while 36% didnt even realize that AI had saved them time.
AI doesn’t just help you take on more, but it also gradually raises the bar for what counts as a normal workloadand level of complexityfor a particular role.
An architect on my team once told me that since AI came along, he expects junior developers to be more independent. If they have questions, they should work it out with the AI assistant first. Then they can come back with a specific problem and possible solutions, rather than expecting a lecture on the basics of the profession. Another example: In an HBR article, the head of workforce innovation at KPMG says she now expects more of associatesthey should come to meetings as well prepared as managers once did, because AI makes this possible.
These are positive examples of how AI can speed up growth for people in junior roles: They quickly become more independent and take on tasks that used to need more help from senior colleagues. But within a team, the productivity boost from AI is not always spread evenly. And that has consequences.
EXTRA WORK FOR SOMEONE ELSE
I recently came across an article by Sreenivasa Reddy, a lead software engineer at the telecoms company AT&T. In it, he described his teams story, which initially looked like a textbook example of AI adoptionthey started working 40% faster. But over the next two months, three of their strongest mid-level engineers resigned.
As it turned out, while junior and senior engineers were becoming more productive, the mid-level engineers were spending more and more time reviewing AI-generated code and fixing errors. The results of their work werent reflected in the metrics or performance reviews, and they received no credit for it, which, of course, was demotivating.
This isnt just the story of one team. In a study by Workday involving 3,200 managers and employees across three regions, 85% said that AI saves them between 1 to 7 hours a week, but around 37% of that time is spent fixing and rewriting what it has generated. Often, this work falls to middle-level roles, as it did in Sreenivasa Reddys story.
HOW TO SPOT BURNOUT BEHIND HIGH METRICS
In the past, a drop in productivity might have been a sign to check in on someone. Now, it may come too late or never show up at all. So as AI speeds up the pace of work, I would advise leaders to look closer at what lies behind the rising productivity numbers. Heres where to start:
Acknowledge and value invisible work. A big part of working with AI barely shows up in productivity stats. Its still routine workreviewing generated content and fixing mistakes. The people doing this might have fewer completed tasks to their name, but they are the ones who keep the whole team’s numbers high. This kind of work should count both when balancing the workload and during performance reviews.
Dont turn every saved hour into more work. It is always tempting to finish one job quickly and move straight into the next. But when this becomes the norm, the team ends up operating on its limit. If AI frees up time, it is better to decide in advance how to use it. That might mean more work, but it could just as easily be quality control, learning, or simply taking a beat to recharge.
Dont confuse using AI with efficiency. For me, its about whether the tool helped us make a better decision. Otherwise, AI adoption just becomes another metric people optimize for the sake of optimizing.
I no longer take strong numbers as proof that the team is doing fine. After asking How much have we achieved? I ask one more question: What did this result cost the team?
Illia Smoliienko is founder and engineering leader at Sivense.
Monday holidays, like today’s double feature of Columbus Day and Indigenous Peoples Day, can be confusing.
When you layer in the observances controversy, it gets even muddier.
Does the federal government recognize Columbus Day, or Indigenous Peoples Day, or both?
Read on to find the answers to these questions, some historical facts, and clarification about what exactly remains open and closed today: Monday, October 12, 2026.
A brief history of Columbus Day
Since 1971, Columbus Day has been observed on the second Monday in October. This year, Columbus Day falls on October 12, 2026, which is today.
The day celebrates Christopher Columbus, who is historically credited with discovering America, even though indigenous people already inhabited the region.
In 1492, he sailed the ocean blue thanks to funding by King Ferdinand and Queen Isabella of Spain, with the directive to find a route to Asia.
Instead, the expedition aboard the Nia, Pinta, and Santa Maria landed in the Bahamas, then Hispaniola. His mistake helped Spain establish the first American colony in the Caribbean.
Columbus wasnt exactly the most benevolent leader. As governor of Hispaniola, he enslaved the Taino and Arawak people. As a result, they faced horrific working conditions on mines and plantations.
Columbus forced his Catholic faith on native people using brutal methods. His discovery also kicked off a race of exploration among European nations. These settlers brought new diseases to the Americas, which decimated native populations.
Centuries later, in 1892, President Benjamin Harrison issued a proclamation encouraging Americans to celebrate Columbus Day.
This one-time observance marked the 400th anniversary. President Franklin D. Roosevelt also supported Columbus Days observance in 1937, further cementing a 1934 resolution to make it annual. In 1968, the Uniform Monday Holiday Act made it a federal holiday.Columbus Day has had opponents as early as the 1800s. Anti-immigrant groups did not appreciate how the observance was tied to Catholicism.
More recently, Native American groups have pointed out Columbuss role in the unethical treatment of their ancestors, and protested his celebration.
Indigenous Peoples Day has been presented as a Columbus Day alternative since 1991.
This observance is meant to celebrate the legacy and contributions of Native Americans. Despite President Joe Bidens proclamation in 2021, it is not a federal holiday, although many states, such as California, formally recognize it.
President Trump walked away from Bidens dual recognition of both Indigenous Peoples Day and Columbus Day and instead solely focused on the latter.
Is mail delivered on Columbus Day?
No. Mail is not delivered on Columbus Day because it is a federal holiday.
Is the Post Office open on Columbus Day?
No. Physical post office locations will be closed. The U.S. Postal Service does not operate on federal holidays.
Are banks open on Columbus Day?
To avoid wasting time, it is best to check in with your banking institution of choice before heading over.
The Federal Reserve will be closed, which typically means most national banks follow suit, but some may choose to remain open.
Chase Bank, for example, says its branches are open on Columbus Day, though Monday is treated as a bank holiday for online transactions. Bank of America, meanwhile, will be closed.
Is the stock market open on Columbus Day?
Yes. Both the New York Stock Exchange and the Nasdaq will be open for buying and selling.
Are schools open on Columbus Day?
Public schools are run by local and state governments, which means they are not required to automatically take federal holidays off.
Private schools have even more freedom in their calendaring. It is best to check in with your students institution to know how it handles the holiday.
Are pharmacies open on Columbus Day?
Yes. Just in time for flu season, you are free to stock up on medication on Columbus Day. That includes major chains such as CVS and Walgreens, which should be open normal hours, though it’s always best to check with your local pharmacy.
Are stores and groceries open on Columbus Day?
Yes. Both retail stores and grocery stores would love your business. This includes big box retailers like Walmart and Target, and grocery chains like Kroger and Whole Foods.
Costco warehouses will also be open on the Monday holiday.
Perhaps because of the observances controversy, there are not a ton of specific Columbus Day sales, but there are general fall ones to take advantage of.
Are restaurants open on Columbus Day?
Yes. If you dont feel like cooking or need a sweet treat, restaurants, bakeries, and coffee shops have you covered.
