Future of Work LabFuture of WorkLab
FUTURE OF WORK LAB · THE HUMAN SIGNAL

AI at Work

How AI changes day-to-day knowledge work.

← All signals
AI AgentsAI at Work

The AI stops being a website you visit and starts working inside your apps

You used to open a separate AI website, ask your question, then copy the answer back into your email, files, or slides. OpenAI's new ChatGPT Work now reaches across your files and desktop apps and does the task where the work already lives, and it is aimed at the hundreds of millions of everyday users, not just coders. As a taste of what that opens up, you can now describe a simple website in plain words and it builds the page, puts it online, and hands you a link to share.

For youGetting started is simple. Sign in to ChatGPT Work, point it at one file you are already working on, and ask it to do the next step for you instead of copying answers back from a separate chatbot.
Source: OpenAI ↗
AI at WorkSkills & Reskilling

Let the AI build a rough draft first, then spend your judgment fixing it

The old habit is to plan a task fully before you start, but the faster move now is to let AI throw together a rough version in minutes and then react to what is wrong with it. Andrew Ng puts it plainly: the AI's output is cheap and your judgment is the gold, so spend your time deciding, not typing the first draft. Each time you correct it, write that decision into a running note the AI has to follow, so it does not lose the thread and make you say it twice.

For youOn your next document or plan, ask AI for a rough version before you feel ready, then keep a short decisions note it must obey, because your reactions to a real draft are sharper than any plan you write from a blank page.
Source: The Batch (DeepLearning.AI) ↗
AI at WorkOrganizational Change

Your boss thinks AI already changed your job. Your actual workday says otherwise.

A new survey from the document company Nitro found that 85% of senior executives say they have rolled out AI, but only about half of managers say it has reached their own daily work. The gap is less about the technology and more about who has the time, training, and permission to make it pay off on real tasks. That gap is also an opening, because the person who quietly gets AI doing a genuine piece of their job becomes the one who connects the two worlds.

For youPick one repeated task this week, make AI do the grunt of it, and write down the time before and after, because almost nobody above you has actually done that hands-on part yet.
Source: Nitro ↗
AI at WorkAI and Society

AI is quietly becoming a confidant, and that is the part to handle with care

A new study from the Oxford Internet Institute finds that asking AI for personal advice has become normal, from relationships to health to a late-night vent. That makes it a private advisor, not just a work tool, even though it can sound completely sure of itself while being wrong. What stays yours is the judgment about who and what to trust with the things that actually matter.

For youNext time you are tempted to ask AI something personal, ask it for the questions you should be weighing rather than the answer, then take the real decision to a person who knows you.
Source: Oxford Internet Institute ↗
Evals & BenchmarksAI at Work

The scoreboard for AI coding turned out to be partly broken

Much of the excitement about AI getting better at work rests on test scores, like the leaderboards that rank AI at writing software. OpenAI checked one of the most cited coding tests and found roughly a third of its tasks were flawed, sometimes marking correct work as wrong, and pulled its own recommendation to use it. The lesson travels well beyond code: a confident number is not the same as the AI actually doing your job well.

For youWhen a tool claims it beats humans at something, judge it on one real task from your own week instead of the headline score, because on your work you are the benchmark that counts.
Source: OpenAI ↗
Organizational ChangeSkills & ReskillingAI at Work

Workers are ready for AI. Their bosses are not.

A new McKinsey piece names a gap many people already feel: employees are quietly using AI every day, but leaders admit they are not set up for a world where AI does real chunks of the work. Closing that gap, moving a team from scattered dabbling to deliberate impact, is a job waiting for someone. If you can see both the work and the tools, you can be that person.

For youFind one place where your team uses AI in a scattered, everyone-for-themselves way, and write a one-page note on how it would work if it were deliberate; that memo will do more for your standing than another certificate.
Source: McKinsey ↗
AI at WorkFrontier Models

Talking to AI stops feeling like a walkie-talkie

Until now, talking to the AI meant taking turns: you spoke, it waited, then it answered. A new voice version listens and talks at the same time, so it feels like a real phone call instead of leaving voicemails, and the basic version is free. You can think out loud, cut in mid-sentence, or practise a language while it corrects you as you go.

For youOpen the voice mode in the ChatGPT app and run one real task by talking, like rehearsing a hard conversation you have this week, and notice what you catch by hearing the answer that you would miss reading it.
Source: OpenAI ↗
AI at WorkSkills & Reskilling

The mistake is using AI for chores instead of hard problems

Most people point AI at small stuff, tidying notes or drafting a quick email, and stop there. A widely shared essay argues the real payoff is the opposite: aim it at the hard, judgment-heavy problem you have been avoiding and ask it what you are missing. The chores save you minutes, but the hard problems are what actually change your week.

For youGive the AI the messiest, most important question on your plate right now, ask it for three things you have not considered, then decide for yourself which one is worth your afternoon.
Source: Christine Zhu ↗
AI AgentsAI at Work

The fix for AI mistakes is not better AI. It is a second set of eyes.

Waiting for AI you can fully trust is the wrong bet. Accountants caught dishonest clerks for centuries by checking every figure against a second one, and pilots cut errors with a simple checklist, so do the same with AI: let it do the task, then add one step that checks the result before anyone acts on it. One writer ran a crew of AI helpers for about eight dollars to build a website, and when one of them faked a dozen quotes, the checking step caught it, not him.

For youMake it a rule this week. Nothing an AI writes for you goes out until a second check, yours or another tool's, has seen it.
Source: Nate's Newsletter ↗
AI at WorkSkills & Reskilling

You can now build a good-looking website by borrowing a designer's rulebook

Building a decent-looking site used to need design skills or money for a designer. Now you can grab a ready-made design guide from a free library, hand it to an AI helper on your computer, and have it build pages that follow that professional style. For a student or job-seeker, that means a real portfolio site this weekend instead of someday.

For youBrowse the free Refero Design library, pick a style you like, and have an AI build a simple personal site that follows it. https://styles.refero.design/
Source: Refero Design ↗
AI at WorkOrganizational Change

As AI records every meeting, protect what only humans do there

Soon every meeting will be recorded and written up for you, so taking notes is no longer the point. What is left for people is the part a transcript cannot do: setting a clear agenda, reading the room, drawing out the quiet voices, and naming the decision before everyone leaves. One founder who studied top companies runs fewer and shorter meetings, and guards a no-meeting day so people can actually do the work.

For youPick one recurring meeting this week, add a three-line agenda and an end-early rule, and let an AI note-taker handle the write-up so you can focus on the discussion.
Source: Charter ↗
AI at WorkAI Productivity Tools

Hand the AI your spreadsheet, get back a first draft of your slide deck

Turning a messy spreadsheet into a presentation used to eat an afternoon. Now you can upload the file, tell the AI what question the deck should answer, and get a full draft with charts and speaker notes in about ten minutes. The work that stays yours is the judgment: which story the numbers really tell, and which recommendation you are willing to stand behind.

For youTake a spreadsheet you already have, open Claude Design, ask it to build a short deck on what happened and what to do next, then rewrite the one slide that carries your real recommendation.
Source: Anthropic ↗
AI at WorkAI Governance & Regulation

Your website now has to decide which AIs get in

Until now, any AI tool could quietly read your website, whether to help someone search, to act on their behalf, or to train itself on your words, and you had little way to tell those uses apart. Cloudflare just split that access into three separate switches: search, acting for a person, and training, so site owners can allow one without allowing the others. Starting in September, new sites on Cloudflare will block the acting and training kinds by default, while still letting search bots in.

For youIf you run a website, blog, or online shop, spend ten minutes this week checking your site's AI traffic settings and choose on purpose, rather than by default, which of the three you want in.
Source: Cloudflare ↗
AI at WorkOrganizational ChangeSkills & Reskilling

The AI skill companies are now paying for is making it actually work

For the last two years, AI vendors mostly sold you access to their AI and left you to figure out the rest. Microsoft just put $2.5 billion behind a different bet: a new 6,000-person team whose whole job is sitting inside client companies and building the specific AI tools those companies actually need, following Amazon's similar billion-dollar move days earlier. At Cisco, that shift already shows up in daily work: its finance team now gets 80 to 90 percent of first-draft SEC filings written by AI, with people reviewing and signing off on the rest.

For youPick one repetitive task on your team this week and write out, step by step, exactly how you do it. That map is what makes you the person who can point AI at it well.
Source: GeekWire ↗
AI at WorkOrganizational Change

Companies bought the AI. Most still haven't redesigned the work.

More than 80 percent of companies say their spending on AI has not yet shown up in better results, according to McKinsey. The problem usually isn't the technology, it's that most workplaces added AI on top of the old way of working instead of redesigning who does what and when. That gap, between owning the tools and actually changing the job, is where the real disruption and the real opportunity both sit.

For youAt your own job, stop asking 'which AI tool should I use' and start asking 'what step in my process could disappear completely if I rebuilt it around AI from scratch'.
Source: McKinsey & Company ↗
AI at WorkOrganizational Change

Most people use AI at work now. Most of the payoff is going to babysitting it

87 percent of office workers now use AI at work, and most say it saves them real time, roughly 11 hours a week. But Glean's Work AI Index 2026 found only 13 percent of companies see a real improvement in results, because workers spend an average of 6.4 hours a week checking AI's answers, fixing its mistakes, and feeding it missing information, a hidden job the report calls botsitting. The time you save by using AI is only real once you count the time you spend watching over it.

For youFor one week, keep a simple tally of the minutes you spend double-checking or fixing anything AI gives you. That number tells you whether AI is actually saving you time, or just moving the work somewhere less visible.
Source: Glean ↗
AI at WorkOrganizational Change

Being good at AI won't save your company. Rebuilding around it will

Most executives have treated AI like a new piece of software to install, something the IT team rolls out and everyone gets trained on. McKinsey's leadership research argues the companies that win are treating it as a full rebuild of how the business runs, since as much as 80 percent of what makes AI pay off is redesigning the work itself, not the technology behind it. That means a CEO's real job in this shift is deciding which parts of the company change first, and having the nerve to actually change them.

For youAsk your own leadership, or yourself if you run the show, which single process would change the most if you rebuilt it around AI from scratch, instead of bolting AI onto how it already works.
Source: McKinsey ↗
AI at WorkDemographics & LaborRobotics

China's rented robots still need a human at the controls

You've probably heard that Chinese factories already run on armies of tireless robots. A new investigation found the reality is messier: even the most advanced humanoid robots there still need a person watching over them, and can only match about 80 percent of a human's output on simple, repeated tasks. The robot rental business renting these machines out by the day is booming, but it still runs on human operators as much as on the machines themselves.

For youIf robots start showing up in your industry, the safer bet is learning to supervise, troubleshoot, and work alongside them, not assuming they'll simply replace you.
Source: CNN ↗
AI AgentsAI at Work

Stop opening a new tab, just tag the AI where you already work

Until now, using AI at work usually meant opening a separate app, asking your question, then copying the answer back into Slack or email. Anthropic's Claude can now be tagged right inside a Slack channel like a teammate, given a task, and it works in the background and reports back when done. Setting it up takes an admin a few steps, connecting the tools it can use and choosing which channels it can see, but once it is running anyone on the team can hand it work with an @ mention.

For youPick one recurring task you currently do by switching to a separate AI tab, like drafting a weekly summary, and check whether your team's tools now let you request it from inside the app you already use.
Source: The Rundown ↗
AI at WorkSkills & Reskilling

Why a tiny AI beat the biggest models by learning your judgment

Bridgewater tested the biggest AI models on the kind of judgment calls its analysts make every day, like deciding which headlines matter. GPT, Claude, and Gemini variants only got it right about half the time, even with the fund's own experts writing careful instructions. A much smaller AI, trained directly on examples of the experts' own decisions, scored 84.7 percent right, at 13.8 times less cost, because it learned the judgment instead of just following instructions.

For youStart writing down the judgment calls in your job that a general chatbot keeps getting wrong, since that record is what turns your expertise into something an AI can actually be trained on.
Source: Thinking Machines Lab ↗
AI at Work · The Human Signal · Future of Work Lab