AI in Knowledge Work Productivity

Gartner just surveyed 350 large enterprises deploying AI. 80% cut jobs. Some by as much as 20%. The result? The companies that cut the most showed nearly identical financial returns to the ones that cut the least. In several cases, the ones that cut less performed better. No correlation between AI-driven layoffs and improved ROI. None. Gartner's Helen Poitevin was direct: "Workforce reductions may create budget room, but they do not create return." Cutting people frees up cash. It does not generate value. Most leadership teams are conflating the two. So what actually works? Upskilling staff to work alongside AI. Redesigning roles around what humans do well vs. what AI does well. Building operating models where people guide autonomous systems instead of getting replaced by them. There's a real difference between using AI to do the same work with fewer people and using AI to unlock work that was previously impossible. The first saves money on paper. The second compounds over time. We've already seen the pattern. Klarna cut 700 CS roles, watched quality decline, and started rehiring. IBM automated HR functions and reversed course. The Commonwealth Bank of Australia reversed 45 AI-driven layoffs after realizing those roles were never redundant. Gartner predicts half of companies that attributed headcount cuts to AI will rehire under new titles by 2027. If someone in your org is building an AI business case around headcount reduction, share this data. The assumption that fewer people equals better margins equals better returns is not supported by the evidence. AI is not leading to a jobs apocalypse. It's changing the shape of what people do. The companies that understand that difference will be the ones worth working for, and buying from, three years from now. Read the full piece on State of Brand here: https://lnkd.in/ggH-NXyM

The Voice Stack is improving rapidly. Systems that interact with users via speaking and listening will drive many new applications. Over the past year, I’ve been working closely with DeepLearning.AI, AI Fund, and several collaborators on voice-based applications, and I will share best practices I’ve learned in this and future posts. Foundation models that are trained to directly input, and often also directly generate, audio have contributed to this growth, but they are only part of the story. OpenAI’s RealTime API makes it easy for developers to write prompts to develop systems that deliver voice-in, voice-out experiences. This is great for building quick-and-dirty prototypes, and it also works well for low-stakes conversations where making an occasional mistake is okay. I encourage you to try it! However, compared to text-based generation, it is still hard to control the output of voice-in voice-out models. In contrast to directly generating audio, when we use an LLM to generate text, we have many tools for building guardrails, and we can double-check the output before showing it to users. We can also use sophisticated agentic reasoning workflows to compute high-quality outputs. Before a customer-service agent shows a user the message, “Sure, I’m happy to issue a refund,” we can make sure that (i) issuing the refund is consistent with our business policy and (ii) we will call the API to issue the refund (and not just promise a refund without issuing it). In contrast, the tools to prevent a voice-in, voice-out model from making such mistakes are much less mature. In my experience, the reasoning capability of voice models also seems inferior to text-based models, and they give less sophisticated answers. (Perhaps this is because voice responses have to be more brief, leaving less room for chain-of-thought reasoning to get to a more thoughtful answer.) When building applications where I need a more control over the output, I use agentic workflows to reason at length about the user’s input. In voice applications, this means I end up using a pipeline that includes speech-to-text (STT) to transcribe the user’s words, then processes the text using one or more LLM calls, and finally returns an audio response to the user via TTS (text-to-speech). This, where the reasoning is done in text, allows for more accurate responses. However, this process introduces latency, and users of voice applications are very sensitive to latency. When DeepLearning.AI worked with RealAvatar (an AI Fund portfolio company led by Jeff Daniel) to build an avatar of me, we found that getting TTS to generate a voice that sounded like me was not very hard, but getting it to respond to questions using words similar to those I would choose was. Even after much tuning, it remains a work in progress. You can play with it at https://lnkd.in/gcZ66yGM [At length limit. Full text, including latency reduction technique: https://lnkd.in/gjzjiVwx ]

As AI drives massive productivity gains, businesses may consider cutting back on their human workforce to boost efficiency—but I don’t believe that’s the right choice. The real promise of AI isn’t in replacing humans—it’s in amplifying their potential. For centuries, humanity’s greatest leaps forward were not achieved by replacing humans completely with new tools we’ve built but instead by using the tools to accelerate human agency and potential. The automobile amplified human movement. The computer amplified human creation.  And today, AI is amplifying human intelligence. There are certain jobs—especially those involving repetitive, robot-like tasks—that AI will transform. After all, robots will always be better robots than humans. But humans thrive when they’re empowered to be better humans. With AI, people will unlock new skills and deepen natural talents –– achieving what I call “superagency.” Ultimately, the surge in productivity will guide business leaders to a realization: the right move isn’t to do the same work with fewer people but to create even greater value by leveraging more employees with new AI-driven superpowers. Our goal should be clear. Build technology that works with us, not for us—tools that extend what it means to be human, not replace it. Because when we amplify human ingenuity, the possibilities are infinite. 

We're moving away from charging for *access* to software and toward a model of charging for the *work delivered* by a combination of software and AI agents. Let’s dive into what’s happening and what it means for you ⤵️ 1. The rise of disruptive AI pricing models Tech companies are realizing they can't solely rely on seat-based subscriptions in an age of AI, automation and APIs where value is disconnected with how many people are logging in. Perhaps Salesforce going all-in on Agentforce (and charging $2 per conversation) was the push the industry needed. Each product category has its own flavor of disruptive pricing. - Legal AI products might charge for a demand package generated by AI or an AI-generated summary. - Creator AI products might charge for the content that gets produced such as a video generation or amount of video created. - GTM products might charge for specific tasks completed or workflows executed by the AI. 2. Selling work, not necessarily success As a customer, I wish I only had to pay for software when it delivered results. But the reality is that true success-based billing won’t work for the vast majority of today’s products. Most products should charge for work output instead. The issue is attribution. You want the customer to get a fantastic outcome — and you want them to recognize that your product powered that outcome. As soon as you start charging for success, the customer begins to rethink the results. 3. Goodbye ARR as we know it? Shifting to these newer value-based pricing models isn't a simple pricing change you can just announce in a press release. It's a business model evolution that looks a lot like the shift from on-prem to SaaS in the first place. These new AI pricing models might mean greater volatility in both usage and spend. Variable margin profiles across products and customers. Seasonal revenue fluctuations. The potential for project-based, non-recurring use cases. Put simply, annual recurring revenue (ARR) continues to get dethroned. — Full post in today’s Growth Unhinged newsletter: https://lnkd.in/ea5eTrVD Things are about to get interesting 🍿 #ai #pricing #saas

Most Retrieval-Augmented Generation (RAG) pipelines today stop at a single task — retrieve, generate, and respond. That model works, but it’s 𝗻𝗼𝘁 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁. It doesn’t adapt, retain memory, or coordinate reasoning across multiple tools. That’s where 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜 𝗥𝗔𝗚 changes the game. 𝗔 𝗦𝗺𝗮𝗿𝘁𝗲𝗿 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗳𝗼𝗿 𝗔𝗱𝗮𝗽𝘁𝗶𝘃𝗲 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 In a traditional RAG setup, the LLM acts as a passive generator. In an 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 system, it becomes an 𝗮𝗰𝘁𝗶𝘃𝗲 𝗽𝗿𝗼𝗯𝗹𝗲𝗺-𝘀𝗼𝗹𝘃𝗲𝗿 — supported by a network of specialized components that collaborate like an intelligent team. Here’s how it works: 𝗔𝗴𝗲𝗻𝘁 𝗢𝗿𝗰𝗵𝗲𝘀𝘁𝗿𝗮𝘁𝗼𝗿 — The decision-maker that interprets user intent and routes requests to the right tools or agents. It’s the core logic layer that turns a static flow into an adaptive system. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿 — Maintains awareness across turns, retaining relevant context and passing it to the LLM. This eliminates “context resets” and improves answer consistency over time. 𝗠𝗲𝗺𝗼𝗿𝘆 𝗟𝗮𝘆𝗲𝗿 — Divided into Short-Term (session-based) and Long-Term (persistent or vector-based) memory, it allows the system to 𝗹𝗲𝗮𝗿𝗻 𝗳𝗿𝗼𝗺 𝗲𝘅𝗽𝗲𝗿𝗶𝗲𝗻𝗰𝗲. Every interaction strengthens the model’s knowledge base. 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗟𝗮𝘆𝗲𝗿 — The foundation. It combines similarity search, embeddings, and multi-granular document segmentation (sentence, paragraph, recursive) for precision retrieval. 𝗧𝗼𝗼𝗹 𝗟𝗮𝘆𝗲𝗿 — Includes the Search Tool, Vector Store Tool, and Code Interpreter Tool — each acting as a functional agent that executes specialized tasks and returns structured outputs. 𝗙𝗲𝗲𝗱𝗯𝗮𝗰𝗸 𝗟𝗼𝗼𝗽 — Every user response feeds insights back into the vector store, creating a continuous learning and improvement cycle. 𝗪𝗵𝘆 𝗜𝘁 𝗠𝗮𝘁𝘁𝗲𝗿𝘀 Agentic RAG transforms an LLM from a passive responder into a 𝗰𝗼𝗴𝗻𝗶𝘁𝗶𝘃𝗲 𝗲𝗻𝗴𝗶𝗻𝗲 capable of reasoning, memory, and self-optimization. This shift isn’t just technical — it’s strategic It defines how AI systems will evolve inside organizations: from one-off assistants to adaptive agents that understand context, learn continuously, and execute with autonomy.

We’ve all heard about AI’s potential to boost productivity. But what truly matters to me is whether it’s making work better for the people who show up every day. At Cisco, our People Intelligence team, in collaboration with IT, has been exploring this very topic, and the findings are fascinating. Here are five key insights from our research that leaders should take seriously: 1. Leaders are key to adoption. At Cisco, employees are 2x more likely to use AI if their direct leader uses it. 2. Generic AI training doesn’t work. Role-specific, practical training accelerates AI use. 3. Confidence gaps exist among senior leaders. Directors at Cisco often feel less confident with AI than mid-level employees, underscoring the need for tailored support at all levels. 4. Employee autonomy fuels adoption. Hybrid work environments are powerful accelerators for AI adoption, while mandates can hinder it. Employees who voluntarily go to the office are more likely to use AI, while those who are required to work on-site have lower adoption. 5. AI use is linked to employee well-being, but the relationship is complex, with both benefits and trade-offs that require thoughtful navigation. This is just the beginning. Next, we’re looking at how AI is transforming the way teams operate. For now, one thing is clear, employees who use AI aren’t just more productive. They’re also more engaged, better aligned with company strategy, and empowered to focus on meaningful work. #AIAdoption #EmployeeExperience #FutureOfWork

AI isn’t just a tool—it’s a new kind of labor. Yet most organizations are still approaching generative AI like software procurement.I have created an AI Labor Playbook to help organizations succeed with AI. In most high-performing organizations, every team feels under-resourced. The backlog always outpaces the headcount, and internal investment decisions often turn into zero-sum battles. But a new paradigm is emerging: with generative AI, organizations can now tap into scalable, on-demand AI labor. Instead of waiting weeks for another team to prioritize their request, a marketing analyst can generate a dashboard, a team lead can optimize SEO content, and a product manager can do deep analysis of customer feedback—all within minutes. This shift doesn’t just ease pressure; it unlocks the ability for individuals to prototype, experiment, and execute without waiting for permission, budget, or another team’s capacity.But here’s the catch: AI labor doesn’t manage itself. It’s fast, flexible, and powerful—but only when people know how to lead it. That means identifying the right opportunities, crafting useful prompts, and supervising the results with care. The biggest bottleneck isn’t the AI—it’s the fact that most people were never trained to work this way. We didn’t grow up collaborating with systems that can simulate a customer persona, generate code, or produce market analysis in natural language. Directing AI labor is a new skill—part communication, part systems thinking—and building that skill across your organization is essential if you want to scale the value of AI beyond a few power users or tools.I’m sharing this new guide that reframes how organizations should think about AI adoption. It discusses the concept of "AI as labor"—and lays out a model for building an internal AI labor market, empowering your people to lead AI work, and avoiding the costly traps of vendor lock-in and fragmented tools.

Deloitte will issue a partial refund to the Australian Federal Government after admitting that AI had been used in the creation of a $440,000 report littered with errors, including three non-existent academic references and a fabricated quote from a Federal Court judgment. It’s a headline that lands right at the heart of the transformation happening across the Professional Services industry. AI can now analyse summarise and even simulate insight at scale. But what it can’t yet do is know when to stop. When to question. When to say: ‘this doesn’t seem right.’ For all the talk of automation, this story is a reminder that the value of human advisors isn’t disappearing, it’s being redefined. AI will take thousands of junior roles in Professional Services firms, but senior judgement and quality assurance remain the last line of defence. In an era where a single model can generate a thousand pages of analysis almost instantly, the firms that win won’t just be those that move faster, they’ll be the ones that still think deeply. And this is exactly where the next generation of Partners will earn their keep. The best will become orchestrators, pairing technology’s scale with human discernment. They’ll know how to deploy AI inside the workflow without losing control of the narrative. Because when the model does the work, but the human still owns the outcome, that’s where trust, and value, still live.

Humans keep inventing new ways to "measure" how busy software teams are, and then mistaking them for how good the work is. Lines of code counted how much got typed, never whether it should exist. Story points were meant to size effort, then quietly turned into a quota. Pull requests tell you something changed, not whether the change was worth making. Each one started as a proxy for value and ended up as the thing itself, optimized until it stopped meaning anything. Now we keep seeing "tokens consumed" reported like an accomplishment. It is the strangest one yet, because it is a cost. A team burning nine billion tokens has told you how much compute it spent, not whether anything got better, shipped, or used (imagine bragging about your electricity bill.) The moment a metric becomes a target, it stops describing reality and starts distorting it. AI did not break this. It just attached a meter to it. The only number that has ever mattered is whether the work in front of a user got better. That one is still annoyingly hard to count. ↓ 𝗜𝗳 𝗔𝗜 𝗶𝘀 𝗽𝗮𝗿𝘁 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗷𝗼𝗯, 𝗺𝘆 𝗳𝗿𝗲𝗲 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿 𝗶𝘀 𝗳𝗼𝗿 𝘆𝗼𝘂. 𝗜 𝘀𝗵𝗮𝗿𝗲 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀, 𝘁𝗼𝗼𝗹𝘀, 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁’𝘀 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝗳𝗶𝗲𝗹𝗱: https://lnkd.in/dbf74Y9E

One of the biggest challenges I'm seeing right now with enterprise AI? Lack of context. The fix everyone should be talking about in 2026: knowledge graphs. In enterprise AI, context is everything. Without it, even the most advanced LLMs produce generic output — which can fall short in real-world business environments. That's why the knowledge graph is so central to enterprise AI. Think of it as a collection of all the institutional knowledge a company has amassed — the company-specific insights, policies, best practices and protocol that make a business special and valuable. It's this knowledge graph that transforms AI from vague and often unhelpful to a customized tool that unlocks real efficiency for a business. With enterprise software delivery, knowledge graphs are especially important. Our Knowledge Graph captures everything that makes your software environment unique — from services to security policies, compliance, infrastructure and more. It provides a a living, dynamic digital twin of an organization’s entire software development and delivery ecosystem. Ultimately, it's this knowledge graph that gives AI the context to safely and efficiently automate complex software tasks. #BigIdeas2026