- Future//Proof
- Posts
- 🤐 Nine in ten executives just admitted AI hasn't made them more productive
🤐 Nine in ten executives just admitted AI hasn't made them more productive
🎧 Grok Voice is closing 3,000 Starlink orders a week, Uber gets fined €825M for letting an algorithm fire people, and ChatGPT starts working straight from your inbox.
Avinash Mada
August 27, 2026
In partnership with
Welcome to the Future//Proof 🚀
👋 Hello , the AI Enthusiast.
In this week’s edition, we brought AI updates backed by high-quality research and data to give you deeper insights. You'll find the Top AI Breakthrough of the Week, a featured AI tool with a mini-tutorial, learning resources to help you master these tools, the top 3 AI news stories, and more.
Our goal is to help you improve your knowledge and stay ahead in the rapidly evolving AI landscape. You can submit your questions, queries, thoughts, opinions or anything regarding AI as a reply to this email and we'll feature and address them in our next newsletter.
🚀 Now Let’s dive in and explore the new AI Insights together!
⌛ Read Time: 6m:34s
Cut Lead Review From Hours To Minutes
Sign up for a free trial of Attio, the agentic CRM.
Ask Attio to build a daily workflow that surfaces the deals that need your attention today, like anything with a stage change, a recent reply, or a new signal in the last 24 hours.
Review your pipeline in Claude, synced live from Attio via MCP.
That's it.
An in-depth look at a major AI development, its industry impact, how it could affect your career, and a bold future prediction.
AI 90% of executives say AI hasn't boosted productivity. They're cutting jobs anyway.
Last week we covered hard data showing AI output exploding while working hours refused to fall, and this week we got the explanation, which turns out not to be a technology problem at all. Fortune published research on August 22 from University of Pittsburgh professor Mark Ma, backed by the Federal Reserve Bank of Atlanta, drawing on millions of Glassdoor employee reviews, roughly 10,000 corporate earnings-call transcripts, and hundreds of AI-related investment and layoff announcements from US public companies across five years. The headline finding is that about 90% of executives say AI has not yet delivered a productivity improvement at their own company, but three findings underneath it matter more. Management commentary about AI across those thousands of earnings calls was consistently upbeat, and that optimism had no significant relationship to actual productivity outcomes. Stock reactions to AI-cited job cuts averaged close to zero, and were negative or flat in more than half the cases studied, with the fintech platform Block a rare exception. Most importantly, employee sentiment toward AI turns out to be one of the strongest predictors of whether AI actually improves firm productivity, and Glassdoor comments mentioning AI were markedly more negative than reviews overall, driven by job insecurity, no training or upskilling, badly managed rollouts, and doubts the tools work at all. Ma's framing is a doom loop: companies cut staff to fund AI or to book the returns early, the cuts destroy the trust and job security that make people willing to use AI properly, the productivity gain never arrives, and so the company cuts again. He describes using AI to justify job cuts as a strategic miscalculation, and the researchers found some firms laid people off before investing, purely to free up the capital.
Potential Impact
Put this next to last week's Linear data and the picture completes itself: output is up, hours are not down, and the missing variable is not workflow design alone but whether your people believe AI is being done with them or to them. The efficiency story has stopped working as an investor story, because if the market prices AI-cited layoffs at roughly zero then the "we cut headcount because AI" press release has lost its magic, and what gets rewarded now is evidence of actual output or margin change. Adoption, meanwhile, is a trust problem wearing a technology costume: you can buy every licence in the category, but if your team quietly believes the tools are there to replace them, they will use them at the minimum viable level and the ROI will never show up in the numbers. That also makes training something other than a nice-to-have line item, since lack of upskilling was one of the named drivers of negative sentiment, and the companies getting returns are the ones that made their people more capable rather than fewer in number.
Implications for People/Careers
If you are an employee, this data is quietly good news and a clear instruction at the same time. The good news is that the "AI replaced them" narrative is running well ahead of the evidence, because nine in ten executives privately concede the gains have not materialised, and that gap between the story and the substance is your window. The instruction is that the safest position is being the person who makes AI work, not the person waiting to see if it does, since the sentiment research keeps pointing at the same divide between employees who received training and had a say in how AI was deployed and employees who found out through a layoff announcement. If your employer is not offering that training, source it yourself, because it is now a career-defence expense rather than a hobby. One more marker landed this week: AWS confirmed it will shut down Mechanical Turk on September 30, ending a 21-year run for the marketplace Jeff Bezos once called artificial artificial intelligence, after it stopped taking new customers on July 30 and was overtaken by AI-native data-labelling companies. Whole categories of task work are not being negotiated away, they are being switched off.
Our Future//Take
Our prediction is that within the next 12 months, the credible AI story in earnings calls stops being how many people you removed and becomes what your remaining people now produce. The companies that win the next phase will look almost boring from the outside, having trained their teams properly, rebuilt two or three core workflows end to end, kept the headcount they need, and quietly moved a real number like cost per ticket, cycle time or revenue per employee. The uncomfortable truth buried in this study is that AI rewards organisations that were already good at managing change and punishes ones that were not, which means the technology is not the variable and never was. So the question to carry into Monday is not where you can cut, it is what your team would build with AI if they trusted it was there to make their work better. Here’s your ₹25,000 AI Gift for FREE 🎁
Quick summaries of this week's top AI news, their relevance to your career, and our expert opinions.
SpaceXAI disclosed on Monday that Grok Voice now resolves more than 15,000 inbound support and sales calls per day for Starlink, diagnoses hardware faults, ships replacements, and fulfils over 3,000 orders a week across voice and chat. Musk confirmed it is deployed at scale for both support and sales. When xAI introduced the Think Fast voice model in April, it reported that Starlink's phone agent was converting 20% of sales inquiries and resolving 70% of support inquiries autonomously with no human in the loop, and that A/B tests against the old line showed a meaningful lift in both conversion and containment. Version 2.0 became the default on August 5. Starlink now has north of 13 million consumer subscribers.
Why It Matters to You
This is the first widely disclosed proof that a voice agent can carry a full-scale commercial call centre, not a pilot and not a deflection menu. Note what it is doing: it is selling. Three thousand completed orders a week is a revenue function, not a cost-saving one. If you run a business where a phone conversation stands between an interested person and a purchase, that gap just got a lot cheaper to close.
Our Take
The instinct is to read this as "call centre jobs are gone." The more useful read is that the economics of answering the phone at all have changed. Most small and mid-sized businesses in India lose more revenue to unanswered calls, after-hours inquiries and slow follow-up than they ever lose to bad agents. Start there, with the calls you are currently missing, not the staff you currently have. And note the containment numbers are not 100%. The design that works is an agent that handles volume and hands off cleanly, with the handoff built in from day one rather than bolted on after the first angry customer.
The Netherlands' Data Protection Authority hit Uber with an €825 million (about $966M) fine for suspending and deactivating driver accounts through automated systems without adequate human review, covering conduct from 2018 to 2022. It is the second-largest GDPR penalty ever issued, behind only Meta's 2023 fine. Deputy Chair Monique Verdier's summary was that a computer should not be making decisions of that consequence on its own. Uber called the penalty disproportionate, said it will appeal, and argued its current process already includes human review and an appeals route.
Why It Matters to You
Read the charge carefully, because it is not about AI capability. It is about automated decisions that materially affect a person, made without a human in the loop. That description covers a lot of ordinary business software: automated account suspensions, algorithmic performance scoring, AI-assisted CV screening, fraud flags that freeze payouts, automated deactivation of sellers or partners. If you use any of these, the exposure is not theoretical.
Our Take
Two practical moves. First, list every place in your business where software takes a consequential action against a person without anyone reviewing it, and put a named human in that loop with a documented appeals path. Second, keep the reasoning. Regulators increasingly want to see why a decision was made, and "the model decided" is not an answer. India's DPDP framework and the EU AI Act are both moving in the same direction, so build for the stricter standard once rather than retrofitting twice.
OpenAI's August 25 release notes expanded ChatGPT Work in three ways. Its browser can now operate on sign-in-gated websites for Plus and Pro users, with a confirmation step before consequential actions and secure credential handling through password-manager integration. Plus and Pro users can also trigger scheduled tasks from Gmail, Slack and GitHub webhooks, meaning an incoming email or message can start the work rather than a human remembering to. Free users get task sharing and up to three active scheduled tasks, though webhook triggers stay on paid tiers.
Why It Matters to You
This is the quiet shift from assistant to worker on standby. Until now, ChatGPT did something because you opened it and asked. A webhook trigger means it does something because an event happened, at 3am, whether or not you are at your desk. Combine that with logged-in browsing and a large slice of routine admin becomes delegable: pulling a report from a dashboard behind a login, drafting a response when a specific type of email lands, checking a portal every morning.
Our Take
Start with one trigger, not ten. Pick the single recurring event that reliably steals your attention, an inquiry email, a form submission, a daily report, and wire exactly that. Then be deliberate about the confirmation step on anything consequential. An agent that can log in and act is genuinely useful and genuinely capable of doing the wrong thing at scale, and this week's Uber fine is a decent reminder of which of those two outcomes is more expensive. AI Mastery for FREE (Sign up Now).
Discover a comprehensive guide to an AI tool, exploring its features, practical use cases, and learning resources to help you master it.
If this week's Big Picture is about AI failing where it is bolted onto vague processes, this week's tool is the opposite case: a tool that only works when you give it your actual material.
Gemini Notebook is the product formerly called NotebookLM, renamed on July 16, 2026 and folded into Google's Gemini brand. Google says it has passed 30 million users and 600,000 organisations. The distinguishing feature has not changed and is the whole point: it answers only from the sources you give it, with citations back to the exact passage. Upload your contracts, SOPs, research, call transcripts, board decks or a competitor's annual report, and you get a grounded assistant instead of a confident guesser.
The upgrade that arrived with the rename is a secure cloud computer inside every notebook, letting it write and run code against your sources for real data analysis rather than just summarising text.
⭐ Top Features
Grounded answers with citations. Every claim links back to the source passage, so you can verify in one click. For anything contractual, financial or regulatory, this is the difference between usable and unusable.
Wide source ingestion. Google Docs and Slides from Drive, PDFs, pasted text, websites and YouTube videos. Sync Drive content so the notebook stays current.
Secure cloud computer. Native code execution for calculations and data analysis grounded in your files, with no local setup.
Multiple output formats. Beyond text and Audio Overviews, it can produce charts, documents, spreadsheets, slides and structured data, so the analysis leaves the notebook in a form you can actually send.
Audio Overviews. Turns a stack of documents into a conversational audio walkthrough. Underrated for reviewing material during a commute.
Shared notebooks. Team members work from the same grounded source set, which quietly kills a whole category of "which version is right" arguments.
Resources for Learning
Official Help Centre: support.google.com for source limits, sharing controls and how the cloud computer works on your plan.
Feature Announcement: 9to5google.com for what the rename actually changed and what stayed the same.
A curated list of noteworthy AI tools and their key details to help you stay ahead in your field.
The August 25 update is the one to pay attention to. Scheduled tasks can now be triggered by Gmail, Slack and GitHub webhooks on Plus and Pro, and the built-in browser can operate on sites you are logged into, with a confirmation gate before consequential actions. Free users get task sharing and up to three active scheduled tasks. The differentiator is event-driven work: the shift from a tool you open to a process that runs when something happens. Start with a single trigger tied to your most repetitive inbound request.
The voice model running Starlink's support and sales line is available to businesses through xAI's API, and xAI has previously offered trial access. It is built for messy real-world audio, accents, background noise and interruptions, and for multi-step workflows where the agent has to look something up and act rather than just answer. The differentiator is proof at production volume, which is rare in voice AI. Worth evaluating against the incumbents if you are running inbound sales or support calls, and worth designing your human handoff before you design the script.
The workflow automation platform that connects your apps, your data and your AI models into one flow, with the option to self-host so sensitive data never leaves your infrastructure. It has become the default for teams that outgrew simple trigger-and-action tools but do not want to build from scratch. The differentiator is control: you can see and modify every step, run it on your own server, and mix AI nodes with ordinary logic. This is the tool that makes the "redesign the workflow" advice in this issue actually executable.
An AI notetaker that joins your calls, transcribes them, and produces summaries, action items and searchable meeting records that push into your CRM and project tools. Given that coordination overhead is the hidden tax in every AI productivity study, this is a rare tool that removes a step rather than adding one. The differentiator is searchable institutional memory: every commitment made on a call becomes findable months later. Agree recording consent norms with your team and clients before you roll it out.
A quick poll to help you recollect and engage with key points from the newsletter.
The Atlanta Fed-backed study found roughly 90% of executives say AI has not boosted productivity at their firms. According to the same research, what best explains the gap? |
Share your feedback on today's edition to help us improve and better meet your needs.
How was Today’s Edition? |
Share our Newsletter ⏩
Enjoying our insights on the latest AI breakthroughs? Don’t keep it to yourself! Share this newsletter with friends and colleagues who are passionate about technology and AI innovation.
If you haven’t subscribed yet, make sure to subscribe here to stay updated with cutting-edge AI news, tools, and tutorials delivered straight to your inbox!
Ask Us Anything AI ❓
Got questions? We've got answers!
Submit your questions, queries, thoughts, opinions or anything regarding AI and we'll feature and address them in our next newsletter. Your curiosity drives our content!
👇 Reply to this email with your questions, and we'll answer them in our next edition!👇