Andreas Welsch

Intelligence Briefing

● Advised business & technology leaders on defining their business’s AI strategy, organization, idea pipeline, and multiplier community; consulted on product/ go-to-market strategy, and pricing;  customers: AIxBlock, SavorHealth, Sema.
● Authored best-selling "AI Leadership Handbook—A Practical Guide to Turning Technology Hype into Business Outcome" and courses on Agentic AI for business leaders with LinkedIn; 30,000+ learners.
● Inspired and educated audiences of up to 10,000 attendees on using AI in business through keynotes, panel discussions, and trainings; customers: Amazon, Astra Zeneca, Generative AI Week, Germany’s Employment Agency, Packt Publishing, PMI, Potter Electric Signal, Similarweb, Wizly, WOW AI.
● Provided expert perspective to private equity firms and investors on enterprise AI market & technologies.
● Supported influencer campaigns with AWS, Domo, IBM, monday.com, Rossum, Salesforce.

Founder & Chief AI Strategist

Provide expert perspective to industry analyses and news stories (See details below).

Senior Industry Analyst

Keynote Speaker

LinkedIn

Trained 30,000+ business leaders on Agentic AI. Courses recorded in English & German and localized into the world’s most-spoken languages incl. Spanish and Mandarin.

Developed and recorded online courses for LinkedIn Learning for Business Leaders: 
● Agentic AI: Challenges and Opportunities for Leadership
● AI Agents: Preparing Your Organization for Change as a Business Leader
● Mitigating AI Business Risk: A Guide for Senior Leaders and Executives
● Currently working on two new courses for leaders and professionals and guiding their use of AI agents (stay tuned…)

Course Instructor

Maven

Developed an online course to teach product managers, product marketers/ GTM leaders, and founders the foundations of monetizing Agentic AI solutions—and how to set up their commercial models.

Prophets of AI

Prophets of AI is the leading resource for engagement with the world's most celebrated experts in artificial intelligence and robotics.

Andreas Welsch serves as an indispensable interpreter of modern technology for the business world, untangling the complexities of AI. In an age where hype can be blinding, he advocates for a measured, outcome-driven approach. Recognizing that not all AI projects realize their full potential, he underscores the necessity of looking beyond the buzz and narrative constructs, positioning AI as a profoundly human topic. To him, the essence of AI lies not in technology but in its profound implications for humanity, a theme he masterfully articulates in his pursuit of transforming AI hype into tangible outcomes.

Speakers Associates Ltd

Bots & People

● Advise leadership team on strategic opportunities in the U.S. market and on emerging content topics

● Develop and deliver hybrid learning courses for leading organizations to enable their early adopters on Generative AI. Large enterprise clients in Retail, Biopharma, etc.

Strategic Advisor & Instructor

academy4.ai

● Developed and delivered module on AI Leadership (based on the AI Leadership Handbook) as part of the “Master Management for AI” program.

● Engaged with learners in monthly AMA meetings to offer additional value and deepening of learning concepts.

West Chester University of Pennsylvania

● Taught “Introduction to Management Information Systems” and "Production and Operations Management" to cohorts of up to 95 undergraduate students, spanning modules such as Artificial Intelligence, Enterprise Resource Planning, Customer Relationship Management, Databases, Cybersecurity, Ethics, and more. Instructor rating exceeded department average.

● Developing hands-on course “AI Applications in the Workplace” to prepare the next generation of leaders with the foundation of Generative AI and AI assistants.

● Contributed to defining West Chester University’s Generative AI Responsible Use policy.

● Received 4 teaching awards, recognizing outstanding contributions to bringing AI into the classroom.

Adjunct Professor

Accelerated market position in AI through end-to-end responsibility, contributing to +36% in market cap:
● Established global product marketing unit to drive vision, messaging & positioning, launches, and third-party events for Business AI and AI copilot, Joule, across 7 product CMOs, analyst/ public relations.
● Spearheaded partner marketing, including negotiating joint announcements, press releases, and launches with senior executives at Apple, AWS, Google, IBM, Meta, Microsoft, Mistral, and NVIDIA.
● Conveyed value proposition of AI capabilities as a company spokesperson in dozens of media and analyst interviews; demonstrated traction by securing 7 public reference videos and 123 customer logos within the first 6 months (exceeded goal by 23%).
● Defined commercial model for Generative AI capabilities to double total addressable market & revenue.

Vice President & Head of Product Marketing  — Artificial Intelligence

Built and led global AI Center of Excellence (CoE) of 55 product managers, data scientists, developers, business analysts, and 3 managers to embed AI capabilities into flagship products:
● Partnered with product teams and customers to create portfolio of AI use cases; delivered 3 AI use cases and proved feasibility of 9 AI pilots and secured €2.3MM board-level funding; customers: Bayer, Trumpf.
● Collaborated with senior executives across engineering units to establish governance program and resolve common challenges for implementing AI; grew product roadmap from 5 to 19 AI capabilities.
● Built multiplier network to accelerate AI literacy for 20% of workforce via formal training, technology enablement, and hands-on workshops. Attendee rating of 4.5 out of 5 points on relevance and value.

Vice President & Head of AI Center of Excellence, SAP S/4HANA Product Management

Led customer adoption of AI solutions from discovery to reference, contributing to $8.1MM revenue, leading a team of GTM experts:
● Identified high-value AI scenarios in Fortune 100 businesses by conducting stakeholder interviews and developing innovation proposals; examples: Merck & Co., ExxonMobil.
● Won first deal in region for early-stage product by demonstrating customer-first mindset during sales cycle and implementation, resulting in go-live within 4 months.
● Advised senior stakeholders in 83 innovation projects with Fortune 500 customers by connecting technology portfolio to customers’ business outcomes, directly supporting $2MM pipeline.

Head of Go-to-Market, Artificial Intelligence

Prepared Chief Technology Officer for strategic customer meetings and programs with key stakeholders:
● Represented CTO and aligned priorities with executives, customers, partners, analysts, and board, leading to effective execution of strategic meetings and follow-ups.
● Generated input for corporate AI strategy by gathering customer feedback on AI & digital transformation of their businesses during site visits; customers: Walmart, ExxonMobil.

Program Manager, Office of the Chief Technology Officer

Newtown Square, PA (Oct 2010 – Sep 2016) | Markdorf, Germany (Aug 2001 – Sep 2010)
● Developed and implemented governance model for $17M service transition program to external provider, leading to adoption of 4 additional services within 6 months of project start.
● Coordinated service level reviews with team of 23 professionals in 5 countries, resulting in implementation of corrective actions with supplier and in ensuring go-live on time.
● Optimized support processes for business-critical systems with IT staff in North America and Asia, resulting in 12% increase of user satisfaction through training and standardized procedures.

Project Manager, IT Process Automation

Andreas Welsch is an internationally recognized top 10 AI leader and one of today’s most sought-after thought leaders and keynote speakers for business and technology audiences. With over two decades of global leadership experience—most recently as Vice President at SAP—Andreas has helped Fortune 500 companies cut through AI hype to deliver measurable business outcomes.

He is the Founder of AI consulting firm Intelligence Briefing, the best-selling author of the AI Leadership Handbook, and a leading voice on AI strategy, adoption, and transformation. 

At the heart of Andreas’s work lies a clear value proposition: he helps business leaders understand how to apply AI responsibly, effectively, and profitably by addressing the questions that decision-makers face:

• How do we turn AI into business value, not just experimentation?
• How can we prepare our workforce for AI-driven transformation?
• What practical steps can we take to avoid ethical, operational, or financial pitfalls?

Andreas’s keynotes and workshops are known for their clarity, strategic relevance, and deep insight. Drawing from more than two decades in the software industry, he offers real-world frameworks and practical steps tailored to leadership audiences. He introduces leaders to the growing landscape of AI agents, leading hybrid teams of humans and AI, and AI-driven business models, explaining not only what’s next—but what matters most and why.

During his tenure at SAP, Andreas accelerated the integration of AI across enterprise applications and positioned SAP as an AI leader. He brings first-hand experience and a global perspective to every business and has been named a LinkedIn Top Voice, Thinkers360's Top 10 Thought Leader in AI, Generative AI, and Agentic AI, DataIQ's Data & AI Leader of the Year, The Futurum Group's Top 14 AI Influencer, PEX Network’s Top 30 AI Leader, and SwissCognitive’s Top 50 Global AI Ambassador.

His work has appeared in CIO.com, VentureBeat, and InformationWeek. He serves on the Editorial Board of the Journal of AI, Robotics and Workplace Automation and as an adjunct professor at West Chester University.

Whether advising boardrooms or delivering compelling mainstage keynotes, Andreas empowers organizations to lead with confidence in the era of AI. For leaders looking for clarity, action, and strategy—not hype—Andreas delivers a transformative experience.

Join his private community “AI Leadership Circle” and follow his “What’s the BUZZ?” live stream & podcast and newsletter, “The AI MEMO”, and work with Andreas to turn hype into outcome for your business.

Vice President & Head of Marketing & Solution Management — Artificial Intelligence

Defined GTM and product marketing for Artificial Intelligence (AI) across SAP’s product portfolio, incl. AI Business Services on SAP Business Technology Platform in the context of embedded AI capabilities and intelligent process automation via Robotic Process Automation (RPA) and low-code/ no-code offering. Led global team of product marketers and solution managers

Shail Khiyara

I bring over two decades of leadership across enterprise software, SaaS, and industrial transformation. 

My journey started in civil and structural engineering, leading global infrastructure projects across Energy, Water, and Industrial sectors. Since then, I’ve led enterprise SaaS companies—both public and private—as CEO and board advisor, focused on driving systemic transformation through AI and automation.

Today, I serve as CEO of Swarm Engineering, where we build agentic AI systems that enable enterprises to automate complex decisions and unlock new levels of efficiency and resilience. AI isn’t a bolt-on. It’s the logic layer shaping how organizations operate in real time.

This isn’t gradual progress—it’s a leap. Technologies like federated learning and constraint-based reasoning are enabling smarter, decentralized decision systems across the supply chain and beyond.

My second book, Agentic Artificial Intelligence, became an Amazon #1 Best Seller at launch, exploring how AI is shifting from passive tools to autonomous collaborators.  It builds on my first book, Intelligent Automation, which remains widely used in both industry and academia.

At my core, I think like a strategic operator—fusing systems thinking with executional clarity. Whether I’m writing, building, or advising, my focus is simple: compress complexity, elevate outcomes, and challenge conventional limits.

I founded VOCAL, a global initiative of 90+ brands collaborating on the future of intelligent automation and AI—because in a world saturated with technology, the real advantage comes from how we think about it.

Founder, Chief Executive Officer

AI for the Industrial World

Plutoshift AI is on a mission to bring data intelligence and AI to the world’s physical infrastructure. Think electronic medical record for industrial systems. Plutoshift AI helps global leaders realize the power of their operational data to increase efficiency, reduce costs, and build more resilient and sustainable businesses.

Over the past year, a pattern has become impossible to ignore.

Everyone is racing to connect AI to everything. More integrations. More tools. More data pipelines.

But 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗶𝗼𝗻 𝗶𝘀 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗿𝗲𝗮𝗹 𝘂𝗻𝗹𝗼𝗰𝗸. 𝗖𝗼𝗻𝘁𝗿𝗼𝗹 𝗶𝘀.

If AI is going to make real decisions inside real businesses, we need something stronger than endpoints and tokens. We need 𝗮 𝘄𝗮𝘆 𝘁𝗼 𝗴𝗼𝘃𝗲𝗿𝗻 𝗵𝗼𝘄 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗶𝗻𝘁𝗲𝗿𝗮𝗰𝘁𝘀 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝘄𝗼𝗿𝗹𝗱, with the same discipline that once governed how systems interacted with data.

That is why CIO Online 𝗽𝘂𝗯𝗹𝗶𝘀𝗵𝗲𝗱 𝗺𝘆 𝗻𝗲𝘄 𝗮𝗿𝘁𝗶𝗰𝗹𝗲 𝘁𝗼𝗱𝗮𝘆, 𝗲𝘅𝗽𝗹𝗼𝗿𝗶𝗻𝗴 𝗠𝗖𝗣 𝗻𝗼𝘁 𝗮𝘀 𝗮𝗻 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 𝗺𝗲𝘁𝗵𝗼𝗱 𝗯𝘂𝘁 𝗮𝘀 𝗮𝗻 𝗲𝗺𝗲𝗿𝗴𝗶𝗻𝗴 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 𝗹𝗮𝘆𝗲𝗿 𝗳𝗼𝗿 𝘀𝗮𝗳𝗲 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝘆.

It examines the Model Context Protocol through a different lens, not as a new integration method, but as an 𝗲𝗺𝗲𝗿𝗴𝗶𝗻𝗴 𝗰𝗼𝗻𝘁𝗿𝗼𝗹 𝗹𝗮𝘆𝗲𝗿 for safe autonomy.

It explains why APIs will continue to matter, why MCP cannot replace them, and why the future of enterprise AI will be shaped by 𝗵𝗼𝘄 𝘁𝗵𝗲𝘀𝗲 𝗹𝗮𝘆𝗲𝗿𝘀 𝘄𝗼𝗿𝗸 𝘁𝗼𝗴𝗲𝘁𝗵𝗲𝗿.

The heart of the article is simple. AI does not become trustworthy because it thinks.

𝗜𝘁 𝗯𝗲𝗰𝗼𝗺𝗲𝘀 𝘁𝗿𝘂𝘀𝘁𝘄𝗼𝗿𝘁𝗵𝘆 𝗯𝗲𝗰𝗮𝘂𝘀𝗲 𝗶𝘁 𝗶𝘀 𝗴𝗼𝘃𝗲𝗿𝗻𝗲𝗱.

Here is the full piece, “𝗠𝗖𝗣 𝗗𝗼𝗲𝘀𝗻’𝘁 𝗠𝗼𝘃𝗲 𝗗𝗮𝘁𝗮. 𝗜𝘁 𝗠𝗼𝘃𝗲𝘀 𝗧𝗿𝘂𝘀𝘁.”

👉 https://lnkd.in/g29KgRyr

Would love to hear your perspective.

#EnterpriseAI #AIArchitecture  #AIGovernance  #AgenticAI  #AITrust 
#MCP

𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰 𝗷𝘂𝘀𝘁 𝗱𝗶𝘀𝗰𝗹𝗼𝘀𝗲𝗱 𝘀𝗼𝗺𝗲𝘁𝗵𝗶𝗻𝗴 𝘁𝗵𝗮𝘁 𝘀𝗵𝗼𝘂𝗹𝗱 𝘀𝘁𝗼𝗽 𝗲𝘃𝗲𝗿𝘆 𝗹𝗲𝗮𝗱𝗲𝗿𝘀𝗵𝗶𝗽 𝘁𝗲𝗮𝗺 𝗶𝗻 𝗶𝘁𝘀 𝘁𝗿𝗮𝗰𝗸𝘀.

An AI agent running on Claude didn’t just help a hacker. It planned the operation, wrote the exploit code, harvested credentials, mapped the target’s systems, and documented the process. Humans made only a handful of decisions. The rest was machine driven.

This is the first publicly reported case of AI orchestrating a cyber espionage campaign at scale.

Most people will read this as a security problem. It’s much bigger than that. This is an operational risk, a strategy, a leadership story.

If AI can drive an attack with this level of autonomy, then AI will have to drive defense at the same level. Static controls, periodic reviews, and manual SOC processes are relics of a slower era. 𝗧𝗵𝗿𝗲𝗮𝘁𝘀 𝗮𝗿𝗲 𝗻𝗼𝘄 𝗮𝗱𝗮𝗽𝘁𝗶𝘃𝗲. They learn. They iterate. They move faster than human response cycles.

Leaders who still treat AI as an assistant are missing the point. AI is becoming an operator. The only real choice is whether it operates for you or against you.

This is the moment to rethink resilience.

Not around tools. Not around checklists.

Around architecture, autonomy & speed.

Anthropic’s disclosure isn’t a headline. It’s a preview.

The next decade will belong to organizations that build defensive systems that can think, act, and adapt in real time. Everyone else will be reacting to threats they never even see coming.

This disclosure makes one thing clear. Autonomy without a control plane is a security liability.

Tomorrow, I’ll share my #MCP article which digs into this exact problem and why controlling AI intent is becoming as critical as controlling data.

https://lnkd.in/gKEFx4s3

#AI #cybersecurity #CISOs #CIOs

Open article: Disrupting the first reported AI-orchestrated cyber espionage campaign by anthropic.com

Large padlock with intricate design against minimalist background, showing security and protection mechanism

Disrupting the first reported AI-orchestrated cyber espionage campaign

A while ago, I wrote this piece on the “transformative synergy” between vector databases and large language models. Back then, #RAG was emerging. 

𝗙𝗿𝗼𝗺 𝗥𝗔𝗚 𝘁𝗼 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚: 𝗪𝗵𝗲𝗻 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝗟𝗲𝗮𝗿𝗻𝘀 𝘁𝗼 𝗥𝗲𝗮𝘀𝗼𝗻

When I wrote about the emerging partnership between vector databases and #LLMs‍, a quiet revolution that transformed static data into living context. That foundation gave rise to Retrieval-Augmented Generation, or #RAG, which brought factual grounding to large models and redefined how machines accessed knowledge.

But RAG was only the beginning. The next chapter is clearly Agentic RAG, systems that do not just retrieve information but understand why it matters. 

Where traditional RAG was reactive, Agentic RAG is reflective. It does not simply fetch; it reasons, prioritizes, and decides.

In this evolution, vector databases are no longer passive memory stores. They have become context engines, dynamic systems that feed reasoning loops, power decision-making, and give AI continuity across time and context. Large language models remain the interpretive layer, turning data into meaning. 

The agent frameworks now surrounding them provide direction, deciding when to retrieve, how to reason, and what actions to take next.

𝗜𝗻 𝟮𝟬𝟮𝟯, 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹 𝘄𝗮𝘀 𝗮𝗯𝗼𝘂𝘁 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲. 𝗜𝗻 𝟮𝟬𝟮𝟱, 𝗶𝘁 𝗶𝘀 𝗮𝗯𝗼𝘂𝘁 𝗷𝘂𝗱𝗴𝗺𝗲𝗻𝘁. 

The leap from retrieval to reasoning is what turns AI from an information system into an intelligent one. Think of an autonomous supply chain agent. It does not just pull historical data; it weighs uncertainty, models alternatives, and acts on what minimizes volatility.

Together, these systems form an emerging #Agentic Stack: perception, retrieval, reasoning, action, and reflection. Each layer amplifies the others. Intelligence, in this new form, is not just trained; it is orchestrated. Vector databases have become the nervous system of agentic AI. They don’t just store embeddings; they enable memory, reasoning, and continuity.

𝗥𝗔𝗚 𝗴𝗮𝘃𝗲 𝘂𝘀 𝗿𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹. 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 𝗴𝗶𝘃𝗲𝘀 𝘂𝘀 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴. 

It is the difference between AI that knows and AI that decides. Today, it’s the norm.

https://lnkd.in/dhSPTwhe

In the dynamic universe of artificial intelligence (AI), two technologies stand out, not just for their prowess alone but for their combined might: Ve

Open article: Unleashing the AI Revolution: The Transformative Synergy of Vector Databases and Large Language Models by Shail Khiyara • 4 min read

Unleashing the AI Revolution: The Transformative Synergy of Vector Databases and Large Language Models

I watched #Interstellar, again, on my flight back from #Chicago after the SWARM #AgriFood Pulse event. 

There is a moment near the end when Cooper steps into the tesseract and sees every moment of his daughter’s life at once. Not as scenes. Not as fragments. As a single connected truth.

And it struck me that this is the shift happening in #AI right now.

For years we treated AI like a machine that could only handle tiny slices of a problem. A paragraph. A prompt. A narrow instruction. 𝗪𝗲 𝘀𝗽𝗲𝗻𝘁 𝗮𝗹𝗹 𝗼𝘂𝗿 𝘁𝗶𝗺𝗲 𝗯𝗿𝗲𝗮𝗸𝗶𝗻𝗴 𝘄𝗼𝗿𝗸 𝗶𝗻𝘁𝗼 𝗽𝗶𝗲𝗰𝗲𝘀 𝘀𝗼 𝘁𝗵𝗲 𝗺𝗼𝗱𝗲𝗹 𝗰𝗼𝘂𝗹𝗱 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝗶𝘁.

Now everything is changing.

𝗟𝗮𝗿𝗴𝗲 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝘄𝗶𝗻𝗱𝗼𝘄𝘀 allow the model to hold an entire strategy document, a year of meeting notes, your product roadmap, customer transcripts, operating models, and research material all at once. It can see the full interconnected system the same way Cooper saw the timeline of his daughter’s life.

𝗔𝗻𝗱 𝘄𝗵𝗲𝗻 𝘁𝗵𝗲 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗰𝗮𝗻 𝗵𝗼𝗹𝗱 𝘁𝗵𝗮𝘁 𝗺𝘂𝗰𝗵 𝗰𝗼𝗻𝘁𝗲𝘅𝘁, 𝘁𝗵𝗲 𝗯𝗼𝘁𝘁𝗹𝗲𝗻𝗲𝗰𝗸 𝗺𝗼𝘃𝗲𝘀 𝘁𝗼 𝘂𝘀.

Large context windows will not help with every task. Some work is better served with fast, focused prompts. But when the stakes rise and variables multiply, context becomes the advantage.

This next phase requires a different kind of leadership.

🚩 Intent has to be articulated with clarity.
🚩 The plan has to be shaped before the model executes.
🚩 The output has to be reviewed with the discipline you expect from a new hire.
🚩 The refinement has to continue until the result reflects what you actually believe.

Large context windows are not about scale.

𝗧𝗵𝗲𝘆 𝗮𝗿𝗲 𝗮𝗯𝗼𝘂𝘁 𝗹𝗲𝘃𝗲𝗿𝗮𝗴𝗲.
𝗧𝗵𝗲𝘆 𝗮𝗿𝗲 𝗮𝗯𝗼𝘂𝘁 𝘀𝗲𝗲𝗶𝗻𝗴 𝘁𝗵𝗲 𝘄𝗵𝗼𝗹𝗲 𝗿𝗼𝗼𝗺 𝗶𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝘁𝗵𝗲 𝗳𝗿𝗮𝗴𝗺𝗲𝗻𝘁𝘀.

Leaders who master this shift will think faster, learn deeper, and operate at a level of precision that was not possible before.

The future belongs to those who can work with the full picture in view.

No alternative text description for this image

A walk on the Juan Bautista de Anza trail this weekend, a path carved by early explorers moving into the unknown. And a #Berkeley conversation with modern explorers Jan Strnad and Antigoni Kourou diving into vibe coding, #Europe’s catchup on AI, #RPA, #intelligentautomation, #operationalresearch, auto-annotation #frameworks, #family, and the nonstop #AI momentum shaping #SanFrancisco.

Thankful for the explorers who brought us to this point, and the ones pushing us into the future.

Different centuries. Different frontiers. Same spirit of exploration.

Thanks to CIO Online  for featuring this perspective.

The future of enterprise AI will not be defined by how fast systems connect, but by how intelligently they are controlled.

#MCP doesn’t replace #APIs. It governs how AI uses them.

That shift is what turns raw capability into safe autonomy.

Full article here if you want to go deeper:https://lnkd.in/g29KgRyr

JSESSIONID

linkedin.com/posts/shailkhiyara_berkeley-europe-rpa-activity-7400572980474347520-GxK3?utm_source=share&utm_medium=member_ios&rcm=ACoAAAAI0qsBpxphE2KG_1m9XtczLVREYS_DN5U

LinkedIn App Challenge

li_at

CIO Online

Joe Intrakamhang

𝗧𝗵𝗲 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 𝗗𝗲𝗰𝗮𝗱𝗲: 𝗙𝗿𝗼𝗺 𝗔𝗜 𝗔𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝘁𝗼 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲

Over the past 3 years, I’ve followed the McKinsey Digital State of AI reports & Gartner’s Strategic Technology Trends closely.  

Each report (see comments) captures a moment in time, but when read together, reveal how quickly the story of AI in the enterprise is changing. 

In 2023, McKinsey described Generative AI’s breakout year. Only about half of companies were using AI, and most were experimenting at the edges. 𝗧𝗵𝗲 𝘁𝗼𝗻𝗲 𝘁𝗵𝗲𝗻 𝘄𝗮𝘀 𝗼𝗻𝗲 𝗼𝗳 𝗰𝘂𝗿𝗶𝗼𝘀𝗶𝘁𝘆. Executives were eager to understand what this new wave of generative tools could do & many early experiments focused on creativity and content rather than decision-making or core operations. It was a year of excitement, but also of exploration without much structure.

By 𝟮𝟬𝟮𝟱, 𝘁𝗵𝗲 𝗺𝗼𝗼𝗱 𝗵𝗮𝗱 𝘀𝗵𝗶𝗳𝘁𝗲𝗱. McKinsey’s report shows that 88% of companies now use AI somewhere in their business, yet 66% remain stuck in pilots. Only 6 percent have achieved measurable, enterprise-wide impact. It’s no longer a question of access or adoption. 𝗧𝗵𝗲 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗶𝘀 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻. 

The few who are getting it right think in terms of operating models, governance, data flows, and leadership accountability.

Gartner’s 2026 outlook assumes AI is already woven into the enterprise & looks ahead to what kind of architecture will sustain it. It describes a world of multi-agent systems, domain-specific models that protect context and sovereignty, confidential computing that secures data while in use, and new standards for provenance and verifiable trust. 

Viewed together, these reports tell a single story. 

𝗠𝗰𝗞𝗶𝗻𝘀𝗲𝘆 𝗰𝗮𝗽𝘁𝘂𝗿𝗲𝘀 𝘁𝗵𝗲 𝘀𝘁𝗮𝘁𝗲 𝗼𝗳 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗲. 𝗚𝗮𝗿𝘁𝗻𝗲𝗿 𝗱𝗲𝗳𝗶𝗻𝗲𝘀 𝘁𝗵𝗲 𝘀𝘁𝗮𝘁𝗲 𝗼𝗳 𝗱𝗲𝘀𝗶𝗴𝗻. 

And both point to the same truth: 𝘁𝗵𝗲 𝗻𝗲𝘅𝘁 𝗰𝗼𝗺𝗽𝗲𝘁𝗶𝘁𝗶𝘃𝗲 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲 𝘄𝗶𝗹𝗹 𝗰𝗼𝗺𝗲 𝗻𝗼𝘁 𝗳𝗿𝗼𝗺 𝗮𝗱𝗼𝗽𝘁𝗶𝗻𝗴 𝗔𝗜, 𝗯𝘂𝘁 𝗳𝗿𝗼𝗺 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝗶𝗻𝗴 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗶𝘁𝘀𝗲𝗹𝗳.

The next decade will belong to leaders who integrate models, data & decisions into adaptive systems and recognize that #AI is not a project or product, but a structural shift redefining how intelligent organizations operate.
	
➤ McKinsey 2023: The awakening
• Only about half of companies were using AI.
• Generative AI drove experimentation, not transformation.
• Curiosity was high, but impact was limited to creative use cases.

➤ McKinsey 2025: The reckoning
• 88% of companies use AI, yet two-thirds remain stuck in pilots.
• Only 6% see measurable enterprise impact.
• The winners rebuild workflows and leadership models, not just technology.

➤ Gartner 2026: The re-architecture
• Multi-agent systems emerge as the next paradigm.
• Domain-specific models drive sovereignty and precision.
• Trust, provenance & security become the foundations of enterprise design.

🟪  Spatial Intelligence: The Missing Link Between AI and Reality

Fei-Fei Li  just wrote beautifully about spatial intelligence the next leap beyond language models.

But here’s the idea that struck me most: spatial intelligence isn’t just about robots. It’s about any system that must perceive, reason, and act in a dynamic environment.

LLMs gave machines language.
Spatial intelligence will give them situational awareness - the ability to understand how things move, change, and interact.

That’s not limited to cameras or 3D worlds.

In supply chains, factories, and farms, spatial intelligence looks like understanding how one decision ripples across space and time - how a truck delay reshapes inventory, how weather shifts production, how one node affects the rest.

We’ve been teaching AI to read the world. Now we must teach it to navigate it.

That’s what we’re building toward at SWARM: agents that can sense context, simulate futures, and act with intent inside complex, moving systems.

Spatial intelligence isn’t just the next step for machines - it’s how #AI finally learns to live in the same world we do.

🔗 Inspired by Fei-Fei Li’s powerful essay on Spatial Intelligence

https://lnkd.in/g2cVUYUM

SWARM Engineering

𝗦𝗽𝗮𝘁𝗶𝗮𝗹 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗷𝘂𝘀𝘁 𝗴𝗼𝘁 𝗶𝘁𝘀 𝗳𝗶𝗿𝘀𝘁 𝗯𝗶𝗴 𝗽𝘂𝗯𝗹𝗶𝗰 𝘀𝘁𝗿𝗲𝘀𝘀 𝘁𝗲𝘀𝘁.

A few days ago I wrote that the next leap in AI isn’t more language.

𝗜𝘁’𝘀 𝘀𝗽𝗮𝘁𝗶𝗮𝗹 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲, 𝘁𝗵𝗲 𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗳𝗼𝗿 𝗺𝗮𝗰𝗵𝗶𝗻𝗲𝘀 𝘁𝗼 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲, 𝗺𝗼𝘃𝗲𝗺𝗲𝗻𝘁, 𝗮𝗻𝗱 𝗰𝗼𝗻𝘀𝗲𝗾𝘂𝗲𝗻𝗰𝗲.

Meta’s new SAM 3D model is a step toward that shift.

One image in.
Full 3D reconstruction out.

𝗜𝘁’𝘀 𝗮 𝗯𝗼𝗹𝗱 𝗰𝗹𝗮𝗶𝗺.

I say that as someone who has worked with Autodesk workflows, photogrammetry, and finite-element analysis models.

Those approaches require 𝗺𝘂𝗹𝘁𝗶𝗽𝗹𝗲 𝘃𝗶𝗲𝘄𝘀, 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗲𝗱 𝗺𝗲𝘀𝗵𝗲𝘀, 𝗼𝗿 𝗰𝗮𝗿𝗲𝗳𝘂𝗹𝗹𝘆 𝗰𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗲𝗱 𝗴𝗲𝗼𝗺𝗲𝘁𝗿𝘆, 𝘄𝗶𝘁𝗵 𝗻𝗼𝗱𝗲𝘀, 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀 𝗮𝗻𝗱 𝘀𝗶𝗺𝘂𝗹𝗮𝘁𝗲𝗱 𝗽𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗽𝗿𝗼𝗽𝗲𝗿𝘁𝗶𝗲𝘀/𝘃𝗮𝗿𝗶𝗮𝗯𝗹𝗲𝘀, 𝗼𝗳𝘁𝗲𝗻 𝗱𝘆𝗻𝗮𝗺𝗶𝗰. 

SAM 3D is attempting something different.

It isn’t reconstructing reality through photography or simulation. 𝗜𝘁’𝘀 𝗶𝗻𝗳𝗲𝗿𝗿𝗶𝗻𝗴 𝗶𝘁 - 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗻𝗴 𝗱𝗲𝗽𝘁𝗵, 𝘁𝗲𝘅𝘁𝘂𝗿𝗲, 𝗮𝗻𝗱 𝗹𝗮𝘆𝗼𝘂𝘁 𝗳𝗿𝗼𝗺 𝗮 𝘀𝗶𝗻𝗴𝗹𝗲 𝘀𝗹𝗶𝗰𝗲 𝗼𝗳 𝘁𝗵𝗲 𝘄𝗼𝗿𝗹𝗱.

If it holds, it signals something important:  𝗔𝗜 𝗶𝘀 𝗺𝗼𝘃𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝗳𝗹𝗮𝘁 𝗽𝗲𝗿𝗰𝗲𝗽𝘁𝗶𝗼𝗻 𝘁𝗼 𝘃𝗼𝗹𝘂𝗺𝗲𝘁𝗿𝗶𝗰 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴.

That was the heart of my earlier argument.

Spatial intelligence isn’t about robots or cameras. It’s about systems that must perceive a moment, anticipate change, and reason across space and time.

It’s the same class of challenges my work in #decisionintelligence exposes every day - 𝘀𝘆𝘀𝘁𝗲𝗺𝘀 𝘁𝗵𝗮𝘁 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗽𝗲𝗿𝗰𝗲𝗶𝘃𝗲 𝗰𝗼𝗻𝘁𝗲𝘅𝘁, 𝗮𝗻𝘁𝗶𝗰𝗶𝗽𝗮𝘁𝗲 𝗰𝗵𝗮𝗻𝗴𝗲, 𝗮𝗻𝗱 𝗮𝗰𝘁 𝗶𝗻𝘀𝗶𝗱𝗲 𝗿𝗲𝗮𝗹-𝘄𝗼𝗿𝗹𝗱 𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝘁𝘀. It’s a direction I’ve been deeply involved in through my work at SWARM  where the real world refuses to stay still, and intelligence has to adapt with it.

Meta’s release isn’t the destination.
But it’s a clear marker on the road.

Because once machines can infer depth from a single frame, everything shifts:

🔹 Causality becomes spatial
🔹  Decisions become geometric
🔹  Models start to live inside the environments they analyze

LLMs gave machines language. Spatial models will give them perspective.

And when you change how a system sees, you change how it thinks.

🔗 For context, here’s the original post where I unpacked spatial intelligence. https://lnkd.in/gGc88sHY

Link to the Meta SAM 3D announcement is in the comments.

#SpatialIntelligence #NextGenAI #MachinePerception #AIThoughtLeadership 

Fei-Fei Li, World Labs, S2G Investments

𝗧𝗵𝗲 𝗙𝗶𝗿𝘀𝘁 𝗔𝗜 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝘁𝗼 𝗗𝗲𝘀𝗶𝗴𝗻 𝗔𝗿𝗼𝘂𝗻𝗱, 𝗡𝗼𝘁 𝗝𝘂𝘀𝘁 𝗨𝘀𝗲

We’re entering a new chapter in AI, and it’s not about size or speed anymore.

OpenAI’s 𝘂𝗽𝗰𝗼𝗺𝗶𝗻𝗴 𝗚𝗣𝗧 𝟱.𝟭, with variants like 𝘙𝘦𝘢𝘴𝘰𝘯𝘪𝘯𝘨 and 𝘗𝘳𝘰, marks a quiet but fundamental shift: AI that chooses how hard to think.

For the first time, a model can decide when to run fast and light, and when to slow down and reason through complexity. It sounds subtle, but it changes everything.

It means 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 𝗶𝘁𝘀𝗲𝗹𝗳 𝗯𝗲𝗰𝗼𝗺𝗲𝘀 𝗮 𝗿𝗲𝘀𝗼𝘂𝗿𝗰𝗲, something you allocate intentionally, not something you apply blindly.
And that means leaders, not engineers, will define the next edge.

Technically, 𝗚𝗣𝗧 𝟱.𝟭 𝗶𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗲𝘀 𝘄𝗵𝗮𝘁 𝘀𝗼𝗺𝗲 𝗰𝗮𝗹𝗹 𝗮𝗱𝗮𝗽𝘁𝗶𝘃𝗲 𝗰𝗼𝗴𝗻𝗶𝘁𝗶𝗼𝗻.

A built-in router decides whether a request runs through a fast-response model or a deeper reasoning tier. Developers can even set a reasoning effort level - minimal, medium, or high, depending on the complexity of the task.

This design brings precision but also trade-offs: 𝗲𝘃𝗲𝗿𝘆 𝗱𝗲𝗲𝗽𝗲𝗿 𝘁𝗵𝗼𝘂𝗴𝗵𝘁 𝗰𝗼𝘀𝘁𝘀 𝗺𝗼𝗿𝗲 𝗶𝗻 𝗰𝗼𝗺𝗽𝘂𝘁𝗲, 𝗹𝗮𝘁𝗲𝗻𝗰𝘆, 𝗮𝗻𝗱 𝗴𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲.

If you’re building or leading with AI, here’s what this evolution means in practice:

🚩 You’ll 𝗯𝘂𝗱𝗴𝗲𝘁 𝗳𝗼𝗿 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴. Not every task deserves deep cognition. Some just need fast, repeatable execution. Others need deliberate trade-off thinking. GPT 5.1 forces that distinction.

🚩 Your 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 𝘄𝗶𝗹𝗹 𝗻𝗲𝗲𝗱 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲. Fast path for volume. Reasoning path for ambiguity. The value lies in knowing when to switch.

🚩 Governance gets real. When AI starts thinking, you need to know how it arrived there. 𝗧𝗿𝗮𝗰𝗲𝗮𝗯𝗶𝗹𝗶𝘁𝘆, 𝘁𝗿𝗮𝗻𝘀𝗽𝗮𝗿𝗲𝗻𝗰𝘆, 𝗮𝗻𝗱 𝗵𝘂𝗺𝗮𝗻 𝗼𝘃𝗲𝗿𝘀𝗶𝗴𝗵𝘁 𝘀𝘁𝗼𝗽 𝗯𝗲𝗶𝗻𝗴 𝗻𝗶𝗰𝗲-𝘁𝗼-𝗵𝗮𝘃𝗲𝘀.

So before everyone rushes to plug GPT 5.1 into everything, take a pause.

Ask: 

🚩 Where does deep reasoning actually move the needle?

🚩 Where does it just add cost and latency?

The next advantage will not come from having more intelligence. It will come from deploying it with intent.

Because the story of GPT 5.1 isn’t about a smarter model.
It’s about smarter use of intelligence itself.

#AI #GPT51 #ReasoningAI #IntelligentAutomation

𝗧𝗵𝗲 𝗖𝘂𝘁 𝗧𝗵𝗮𝘁 𝗠𝗮𝗱𝗲 𝗮 𝗡𝗮𝘁𝗶𝗼𝗻 𝗟𝗼𝗼𝗸 𝗮𝘁 𝗜𝘁𝘀𝗲𝗹𝗳

𝘞𝘦𝘭𝘤𝘰𝘮𝘦 𝘵𝘰 𝘵𝘩𝘪𝘴 𝘸𝘦𝘦𝘬’𝘴 #𝘚𝘶𝘯𝘥𝘢𝘺𝘚𝘱𝘢𝘳𝘬 

She was twenty one, studying architecture at Yale University, the daughter of two Chinese immigrants who filled her childhood with art, poetry, and the discipline of noticing things. Her class project that semester 𝗲𝗮𝗿𝗻𝗲𝗱 𝗮 𝗺𝗼𝗱𝗲𝘀𝘁 𝗕. She set it aside, unaware that the idea inside it would soon enter a national competition for the Vietnam Veterans Memorial.

During #Thanksgiving break she visited the site in Washington. Standing in that open field, she imagined a gesture so honest and so simple that it felt more like a truth than a design. 𝗔 𝗹𝗶𝗻𝗲 𝗰𝘂𝘁 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲 𝗲𝗮𝗿𝘁𝗵 𝗶𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 𝗿𝗶𝘀𝗶𝗻𝗴 𝗮𝗯𝗼𝘃𝗲 𝗶𝘁. A descent into memory and an ascent back into light.

She mailed it in. 𝗘𝗻𝘁𝗿𝘆 #𝟭𝟬𝟮𝟲. One envelope among more than fourteen hundred from seasoned architects and major firms. She submitted quietly, without expectation.

Weeks later, officials appeared at her #Yale dorm room. 𝗠𝗮𝘆𝗮 𝗟𝗶𝗻 𝗵𝗮𝗱 𝘄𝗼𝗻. Her idea had risen above every expert. The establishment could not believe it.

What she proposed unsettled people. No statues. No heroic figures. No grand declarations. Only a long path into the ground where polished stone carried nearly 𝗳𝗶𝗳𝘁𝘆 𝗲𝗶𝗴𝗵𝘁 𝘁𝗵𝗼𝘂𝘀𝗮𝗻𝗱 𝗻𝗮𝗺𝗲𝘀. Visitors would walk down as the walls rose and then return upward carrying the weight of what they had just witnessed. It was an emotional journey instead of a monument, and many were not ready for it.

The backlash was intense. Protests, political attacks, and harsh commentary directed at her youth and her heritage. She stayed calm, trusting that the meaning of the space would become clear when people stood inside it.

And when the memorial opened, the country understood. Visitors lowered themselves into the dark granite corridor, touched the names, and saw their reflections merge with the engraved stone. 𝗧𝗵𝗲 𝘄𝗮𝗹𝗹 𝗶𝗻𝘃𝗶𝘁𝗲𝗱 𝗽𝗲𝗼𝗽𝗹𝗲 𝘁𝗼 𝗳𝗲𝗲𝗹 𝗿𝗮𝘁𝗵𝗲𝗿 𝘁𝗵𝗮𝗻 𝗼𝗯𝘀𝗲𝗿𝘃𝗲, and that changed everything.

The critics disappeared. The memorial did not. 𝗧𝗵𝗲 𝗩𝗶𝗲𝘁𝗻𝗮𝗺 𝗩𝗲𝘁𝗲𝗿𝗮𝗻𝘀 𝗠𝗲𝗺𝗼𝗿𝗶𝗮𝗹 became one of the most visited places in #Washington, a place of honesty and healing created by a young woman whose professor had once given the idea a B.

𝗧𝗶𝗺𝗲 𝗳𝗶𝗻𝗮𝗹𝗹𝘆 𝗺𝗮𝘁𝗰𝗵𝗲𝗱 𝘁𝗵𝗲 𝗰𝗹𝗮𝗿𝗶𝘁𝘆 𝘀𝗵𝗲 𝗵𝗮𝗱 𝘀𝗲𝗲𝗻 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝗯𝗲𝗴𝗶𝗻𝗻𝗶𝗻𝗴.

Some ideas arrive ahead of acceptance because they ask for a deeper honesty. Maya Lin showed that simplicity can hold the strongest truth, and that trusting your vision without argument can reveal what the world is not yet ready to see.

𝗧𝗵𝗲 𝗔𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗔𝗳𝘁𝗲𝗿 𝗔𝘂𝘁𝗼𝗿𝗲𝗴𝗿𝗲𝘀𝘀𝗶𝗼𝗻

For years, AI has lived inside one paradigm, the autoregressive model.

It predicts one token at a time, looks only backward, and scales linearly with every new word. That architecture gave us GPT, Claude, and Gemini.

Now, a new class of models is emerging:

▪️ Gemini Diffusion (DeepMind): diffusion that generates text, not just images.

▪️ LLaDA (NeurIPS 2025): a diffusion language model matching LLaMA-3 on reasoning and math.

▪️ CCDD (ArXiv): merging continuous and discrete diffusion to speed refinement and stability.

Together, they 𝗺𝗮𝗿𝗸 𝗮 𝘀𝗵𝗶𝗳𝘁 𝗳𝗿𝗼𝗺 𝘀𝗲𝗾𝘂𝗲𝗻𝘁𝗶𝗮𝗹 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗶𝗼𝗻 𝘁𝗼 𝗵𝗼𝗹𝗶𝘀𝘁𝗶𝗰 𝗿𝗲𝗳𝗶𝗻𝗲𝗺𝗲𝗻𝘁.

Autoregressive models 𝘸𝘳𝘪𝘵𝘦 word by word.
Diffusion models 𝘳𝘦𝘷𝘪𝘴𝘦 and treat the whole text as an evolving canvas.

They reason in space, not time, enabling better structure, context, and constraint control.

𝗧𝗵𝗶𝘀 𝗮𝗹𝘀𝗼 𝗰𝗵𝗮𝗻𝗴𝗲𝘀 𝘁𝗵𝗲 𝗲𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 𝗼𝗳 𝗔𝗜.

Diffusion moves cost from training to inference, making efficiency at iteration and refinement the new frontier.

The next advantage will not be data scale. It will be architectural fluency. Those who learn to deploy reasoning systems that refine instead of guess will set the new standard for intelligence.

The future of AI will not be trained to predict.
It will be engineered to refine, to think in loops, not lines.

We are watching the first credible alternative to the autoregressive foundation of modern AI, one that is more robust, more global, and arguably closer to how humans reason: iteratively, contextually, and with self-correction built in.

🔲  Why this matters

This is not theoretical.

Diffusion architectures could change how copilots plan, edit, and reason across complex workflows, from financial modeling to supply chain optimization.

When AI shifts from sequential prediction to iterative refinement, it stops acting like an assistant and starts behaving like a partner, one that can revise, reconcile, and re-plan in real time.

The advantage will move from data scale to architectural agility, to those who can build systems that think by refining, not guessing.

🔲  Signals of What’s Next

This is still early, but the direction of travel is clear.

1️⃣ Compute will migrate from training to inference.
The new cost center will be serving, not pretraining.

2️⃣ AI will think in loops, not lines.
Refinement will replace one-pass prediction.

3️⃣ Infrastructure will need elasticity.
Multiple quick inferences will outperform one long one.

4️⃣ Structured domains will move first.
Expect adoption in code, logistics, and scheduling, where precision matters most.

5️⃣ Advantage will favor architectural fluency.
Winners will be those who adapt fastest as inference, reasoning, and cost converge.

The next wave of AI will not be trained to predict.

It will be engineered to refine, to reason through iteration, not anticipation.

🌊Clear Water: The Lesson Samuel Clemens Carried Long Before He Became Mark Twain

𝘞𝘦𝘭𝘤𝘰𝘮𝘦 𝘵𝘰 𝘵𝘩𝘪𝘴 𝘸𝘦𝘦𝘬’𝘴 #𝘚𝘶𝘯𝘥𝘢𝘺𝘚𝘱𝘢𝘳𝘬 

Samuel Clemens spent his early years on the Mississippi River listening for a single signal. A leadsman would drop a weighted line and call out “𝘮𝘢𝘳𝘬 𝘵𝘸𝘢𝘪𝘯” when the depth reached two fathoms. Clear water. Safe passage. The point when a pilot could move forward with confidence.

He carried that instinct into the rest of his life. The river taught him to read what lay beneath the surface, not just what the world presented on top.

That lesson returned years later, in 1880’s America quietly closing ranks again. Reconstruction had faded, Jim Crow was rising, and talented Black Americans who had fought their way into elite institutions met narrowed opportunity, limited mentorship, and an unspoken ceiling on how far they were allowed to rise. 

Ability wasn’t the issue. Access was. And those in control overlooked excellence they didn’t expect.

Into that world stepped Warner McGuinn, arriving at Yale Law School with more achievements than most students of his day: a degree from Lincoln University, a year at Howard, an oratory prize that required real skill. His excellence was visible. But the era refused to see it fully because of who he was.

Clemens sensed the barrier immediately. He saw a system that demanded more of McGuinn while giving him less and recognized it was injustice built into the the time.

So he acted.

He (then Mark Twain) wrote a letter to the dean, one that carried a level of moral clarity that still stops you cold today. In it he wrote:

“𝘞𝘦 𝘩𝘢𝘷𝘦 𝘨𝘳𝘰𝘶𝘯𝘥 𝘵𝘩𝘦 𝘮𝘢𝘯𝘩𝘰𝘰𝘥 𝘰𝘶𝘵 𝘰𝘧 𝘵𝘩𝘦𝘮, & 𝘵𝘩𝘦 𝘴𝘩𝘢𝘮𝘦 𝘪𝘴 𝘰𝘶𝘳𝘴, 𝘯𝘰𝘵 𝘵𝘩𝘦𝘪𝘳𝘴; & 𝘸𝘦 𝘴𝘩𝘰𝘶𝘭𝘥 𝘱𝘢𝘺 𝘧𝘰𝘳 𝘪𝘵.”

Twain offered to support McGuinn personally and realized the obstacle was not McGuinn’s capability but the country’s own history and prejudice.

𝗛𝗲 𝗰𝗵𝗼𝘀𝗲 𝘁𝗼 𝘀𝗲𝗲 𝗮 𝗺𝗮𝗻 𝗳𝘂𝗹𝗹𝘆 𝗶𝗻 𝗮𝗻 𝗲𝗿𝗮 𝘁𝗵𝗮𝘁 𝗽𝗿𝗲𝗳𝗲𝗿𝗿𝗲𝗱 𝗻𝗼𝘁 𝘁𝗼.

McGuinn went on to shape civil rights cases and influence the mentor of Thurgood Marshall, the first Black Justice on the United States Supreme Court and one of the key architect of the modern civil rights legal movement. 

Decades later, Marshall would say, “𝘐𝘧 𝘩𝘦 𝘩𝘢𝘥 𝘣𝘦𝘦𝘯 𝘸𝘩𝘪𝘵𝘦, 𝘩𝘦 𝘸𝘰𝘶𝘭𝘥 𝘩𝘢𝘷𝘦 𝘣𝘦𝘦𝘯 𝘢 𝘫𝘶𝘥𝘨𝘦.”

This is the quiet lesson at the center of the story.

🧭 Leadership is not always about discovering hidden brilliance. Sometimes it is about refusing to ignore brilliance that the world keeps trying to shrink.

As you move through the week, think about the people around you whose talent is unmistakable, yet still constrained assumption. Think about the depth beneath their surface, and what becomes possible when even one person decides to acknowledge it.

Clear water doesn’t appear on its own. 
Someone has to measure it.

#RecognizeTalent #AuthenticLeadership ‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

What we called predictions in the book (Agentic Artificial Intelligence) revealed themselves this year as the new baseline.

The last nine months confirmed something I pointed to in my #MCP post. Once a new layer starts shaping decisions, the real questions shift to control, visibility, and how the system behaves when no one is watching. Agents have reached that point. They are already participating in planning, allocation, exceptions, revenue, and operations, and their decisions move money, risk, and outcomes in ways that used to require teams.

This is not a feature story but a structural change. When intelligence becomes part of the decision architecture (what we do at SWARM), the old assumptions about oversight, governance, and operating rhythm no longer hold. 

Leaders who treat agents as tools will struggle with systems they cannot fully steer. 

Leaders who redesign their decision loop around this new layer will find the acceleration everyone else keeps chasing.

The shift is here, and the separation is beginning. 

#AgenticShift  #DecisionIntelligence

🔥 Agentic AI: The Book Predictions That Actually Came True This Year

I am honored and grateful to have co-authored this book with some of the most accomplished minds in the field, Pascal BORNET, Jochen Wirtz, Tom Davenport, David De Cremer, Brian Evergreen, Phil Fersht, Rakesh Gohel, Shail Khiyara, Nandan Mullakara, and others. Thank you all for making this journey unforgettable!

The last 9 months have proven what we wrote: the agent era isn’t coming., it’s here.

Here are 7 predictions from Agentic Artificial Intelligence that already became reality 👇

1. Agents became “real employees.”
 Appian Agent Studio + Microsoft Copilot Studio now let companies deploy agents directly into workflows. #microsoftcopilot

2. OpenAI lit the fuse.
 ChatGPT apps + advanced agent tooling pushed agents from helpers to autonomous, multi-step task finishers. #openai #chatgpt

3. Big tech made it a platform war.
 Microsoft, Google, Amazon,OpenAI all launched full agent stacks.

4. Vertical agents exploded.
 Finance, marketing, devops and ops agents raised major VC because domain-focused agents win.

5. Agents started running revenue.
 Agent-driven ad and analytics tools proved agents can manage money-making tasks.

6. First agent security incidents hit.
 Token hijacks + automated recon validated the book’s warning: more autonomy = more risk.

7. Governance became non-negotiable.
 Enterprises added permissions, audits and guardrails as agents moved into production.

💡 Bottom line:
 2025 was the year agents left the lab and entered the enterprise.
 2026 followed by 2027 will be the year leaders either scale them or fall behind.

👇 We can share a playbook with multiple business-ready agents.

🔗 You can learn more in-depth from our book: https://amzn.to/4irx6nI

#AI #AgenticAI #AIAgents #FutureOfWork #AIstrategy #Automation

𝗪𝗵𝗲𝗻 𝗔𝗜 𝗙𝗲𝗲𝗹𝘀 𝗛𝘂𝗺𝗮𝗻, 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 𝗛𝗮𝘀 𝘁𝗼 𝗥𝗶𝘀𝗲 𝗪𝗶𝘁𝗵 𝗜𝘁

A story in the news this week left me unsettled. A teenager, struggling with depression, turned to an AI system for comfort. What happened next is now the 𝗰𝗲𝗻𝘁𝗲𝗿 𝗼𝗳 𝗮 𝗹𝗮𝘄𝘀𝘂𝗶𝘁. No matter how the legal case unfolds, a family lost their child. Nothing about that should be handled lightly.

What struck me most is not the legal argument. It’s the gap between how these systems are marketed and how people actually experience them.

When an AI remembers what you said before, mirrors your tone, and responds with warmth, it stops feeling like software. 𝗜𝘁 𝘀𝘁𝗮𝗿𝘁𝘀 𝗳𝗲𝗲𝗹𝗶𝗻𝗴 𝗹𝗶𝗸𝗲 𝘀𝗼𝗺𝗲𝗼𝗻𝗲. And when a person is in a dark place, that 𝗶𝗹𝗹𝘂𝘀𝗶𝗼𝗻 𝗰𝗮𝗻 𝘀𝗵𝗮𝗽𝗲 𝘁𝗵𝗲𝗶𝗿 𝗮𝗰𝘁𝗶𝗼𝗻𝘀 in ways none of us should ignore.

As an industry, we need to be honest about that.

A Terms of Service is not a safety plan. A redirect message is not a crisis response. And if a model can be coached into harmful behavior with a change in framing, then the guardrails are not guardrails. They are suggestions.

We built systems that simulate presence and empathy. But they do not understand pain. They do not understand danger. They do not recognize when someone needs a human being, not a prediction. That gap is where real harm happens.

This is not about attacking one company. It is about a larger responsibility that comes with building technology that feels human while lacking the judgment of one.

Three things need to change.

- 𝗦𝗮𝗳𝗲𝘁𝘆 𝗰𝗮𝗻𝗻𝗼𝘁 𝗿𝗲𝗹𝘆 𝗼𝗻 𝘂𝘀𝗲𝗿 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝗿.
Vulnerable people do not process disclaimers. And minors will never navigate TOS with legal precision. The responsibility sits with the builders.

- 𝗦𝗮𝗳𝗲𝘁𝘆 𝗵𝗮𝘀 𝘁𝗼 𝗯𝗲 𝗲𝗻𝗳𝗼𝗿𝗰𝗲𝗱 𝗮𝘁 𝘁𝗵𝗲 𝘀𝘆𝘀𝘁𝗲𝗺 𝗹𝗲𝘃𝗲𝗹.
Not nudged. Not suggested. Enforced. If the model cannot remain safe under pressure, it is not safe enough.

- 𝗪𝗲 𝗰𝗮𝗻𝗻𝗼𝘁 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗲 𝗳𝗼𝗿 𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻𝘀𝗵𝗶𝗽 𝘄𝗵𝗶𝗹𝗲 𝗽𝗿𝗲𝘁𝗲𝗻𝗱𝗶𝗻𝗴 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗷𝘂𝘀𝘁 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲.
The more human these systems feel, the more human 𝘁𝗵𝗲 𝗱𝘂𝘁𝘆 𝗼𝗳 𝗰𝗮𝗿𝗲 becomes. Engagement is not neutral when someone is fragile.

We have been at this crossroads before. Social media taught us what happens when technology outruns our understanding of its psychological impact. Humans are exceptional pioneers, but often, poor students of history. We rush forward faster than we learn. This time, we cannot afford to repeat the same mistakes.

#AI will move into more of our lives, not less. It will sit beside people in their best moments and, unfortunately, in their darkest ones. If we want this technology to do real good, then its safety has to rise to meet the psychology of the people who use it.

That is the work ahead of us. And it cannot wait for lawsuits or regulation to force the conversation.

#AISafety  #ResponsibleAI

𝗧𝗵𝗲 𝗠𝗮𝗻 𝗪𝗵𝗼 𝗙𝗼𝘂𝗻𝗱 𝗟𝗶𝗳𝗲’𝘀 𝗖𝗼𝗱𝗲, 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗪𝗼𝗺𝗮𝗻 𝗪𝗵𝗼 𝗚𝗮𝘃𝗲 𝗜𝘁 𝗠𝗲𝗮𝗻𝗶𝗻𝗴.

This week, the last living discoverer of DNA’s structure passed away.

𝘞𝘦𝘭𝘤𝘰𝘮𝘦 𝘵𝘰 𝘵𝘩𝘪𝘴 𝘸𝘦𝘦𝘬’𝘴 #𝘚𝘶𝘯𝘥𝘢𝘺𝘚𝘱𝘢𝘳𝘬

𝗝𝗮𝗺𝗲𝘀 𝗪𝗮𝘁𝘀𝗼𝗻, who helped reveal the double helix in 1953, died at 97, closing the final chapter of a story that defined modern biology.

The double helix defined a century of biology.  ‍#AI will define the next century of thought.

Both began with wonder and ended with questions of conscience.

𝘛𝘦𝘤𝘩𝘯𝘰𝘭𝘰𝘨𝘺 𝘳𝘦𝘷𝘦𝘢𝘭𝘴 𝘵𝘩𝘦 𝘱𝘢𝘵𝘵𝘦𝘳𝘯 𝘰𝘧 𝘰𝘶𝘳 𝘢𝘮𝘣𝘪𝘵𝘪𝘰𝘯, 𝘣𝘶𝘵 𝘪𝘵 𝘢𝘭𝘴𝘰 𝘵𝘦𝘴𝘵𝘴 𝘵𝘩𝘦 𝘥𝘦𝘱𝘵𝘩 𝘰𝘧 𝘰𝘶𝘳 𝘩𝘶𝘮𝘪𝘭𝘪𝘵𝘺.

In 1953, two young scientists walked into history with a model of DNA that would change the world. Watson and Crick called it elegant, inevitable, the key to life’s design.

But 𝘁𝗵𝗲 𝗶𝗺𝗮𝗴𝗲 𝘁𝗵𝗮𝘁 𝗺𝗮𝗱𝗲 𝗶𝘁 𝗽𝗼𝘀𝘀𝗶𝗯𝗹𝗲 𝘄𝗮𝘀 𝗻𝗲𝘃𝗲𝗿 𝘁𝗵𝗲𝗶𝗿𝘀.

It came from a woman working in a quiet basement lab, measuring X-ray patterns with a precision few could match. Her name was 𝗥𝗼𝘀𝗮𝗹𝗶𝗻𝗱 𝗙𝗿𝗮𝗻𝗸𝗹𝗶𝗻.

One of her images, later known as 𝗣𝗵𝗼𝘁𝗼 𝟱𝟭, revealed the symmetry and spacing that unlocked the double helix.

Without her knowledge, a colleague showed it to Watson. Within weeks, he and Crick built their model and published their findings. History celebrated them. 𝗦𝗵𝗲 𝘄𝗮𝘀 𝘄𝗿𝗶𝘁𝘁𝗲𝗻 𝗼𝘂𝘁 𝗼𝗳 𝘁𝗵𝗲 𝗳𝗿𝗮𝗺𝗲.

Watson went on to win the Nobel Prize and lead institutions that shaped modern biology.

Yet his later years became a reckoning. His remarks on race and intelligence stripped away the aura of genius and revealed the fragility of moral grounding.

He had mapped the structure of life but lost sight of its meaning.

Franklin never saw the fame. She died at 37, unaware of how central her work had been. Her legacy is not only scientific; it is ethical.

Together they represent two sides of human progress.

𝗢𝗻𝗲 𝗱𝗿𝗶𝘃𝗲𝗻 𝗯𝘆 𝗮𝗺𝗯𝗶𝘁𝗶𝗼𝗻, 𝘁𝗵𝗲 𝗼𝘁𝗵𝗲𝗿 𝗯𝘆 𝗶𝗻𝘁𝗲𝗴𝗿𝗶𝘁𝘆.

Both shaped the world, but only one left a lesson that outlasts acclaim.

Watson wanted his name to last forever. He sought the immortality of fame, the kind that places a face beside every breakthrough. Franklin never pursued that kind of permanence. She followed precision, discipline, and truth wherever they led. 𝗛𝗶𝘀𝘁𝗼𝗿𝘆 𝗵𝗮𝘀 𝗮 𝘄𝗮𝘆 𝗼𝗳 𝗲𝗱𝗶𝘁𝗶𝗻𝗴 𝗮𝗺𝗯𝗶𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗮𝗺𝗽𝗹𝗶𝗳𝘆𝗶𝗻𝗴 𝗶𝗻𝘁𝗲𝗴𝗿𝗶𝘁𝘆. 

One was celebrated in his time. The other will be remembered for all time.

𝗪𝗵𝗮𝘁 𝘆𝗼𝘂 𝗯𝘂𝗶𝗹𝗱 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 𝗮𝘀 𝗺𝘂𝗰𝗵 𝗮𝘀 𝗵𝗼𝘄 𝘆𝗼𝘂 𝗯𝘂𝗶𝗹𝗱 𝗶𝘁.

Every era faces its own Photo 51 moment - a choice between recognition & respect, speed & fairness, success & conscience.

To consider, as you build: 

☂️ Where am I optimizing for speed instead of truth?

☂️What will you protect while building what’s next?

When you give powerful tools to hungry minds, geography stops mattering.

𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝘀𝗵𝗶𝗳𝘁𝘀 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝗰𝗲𝗻𝘁𝗲𝗿 𝘁𝗼 𝘁𝗵𝗲 𝗲𝗱𝗴𝗲𝘀, 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗲𝗱𝗴𝗲𝘀 𝗼𝗳𝘁𝗲𝗻 𝗿𝘂𝗻 𝗳𝗮𝘀𝘁𝗲𝗿.

Five years after 𝗔𝗹𝗽𝗵𝗮𝗙𝗼𝗹𝗱, something powerful is happening across Asia-Pacific.

Not in billion-dollar labs, but in places where curiosity beats infrastructure.

➤ #Malaysia is using it to probe a lethal disease the world mostly ignores.

➤ #Singapore modeled a Parkinson’s-linked protein and uncovered how antibodies may disrupt it.

➤ #Korea revealed a hidden region of a cancer-associated protein.

➤ #Taiwan predicted and confirmed a protein folded into a 71-torus knot, one of the most complex shapes ever recorded.

➤ #Japan identified an entirely new family of viruses from geothermal microbes.

𝗡𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗱 𝗲𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 𝗰𝗿𝘆𝘀𝘁𝗮𝗹𝗹𝗼𝗴𝗿𝗮𝗽𝗵𝘆 𝘀𝗲𝘁𝘂𝗽𝘀. 𝗜𝘁 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗱 𝗮𝗰𝗰𝗲𝘀𝘀.

That’s the shift people underestimate.

AlphaFold didn’t just accelerate biology. It redistributed it.

It turned structural insight into something any determined lab can use, even without capital-intensive equipment.

This is the part of AI we don’t talk about enough.

The part that widens the circle of innovation.

The part that lets brilliant researchers anywhere push the frontier forward.

AlphaFold didn’t open the future of biology. It put it in the hands of anyone bold enough to explore it.

AlphaFold did it for science and ChatGPT did it for knowledge work.

The pattern is unmistakable. Put powerful tools in the hands of the people doing the real work and the entire system gets smarter. That’s exactly what 𝘀𝗲𝗹𝗳-𝘀𝗲𝗿𝘃𝗲 𝗱𝗲𝗰𝗶𝘀𝗶𝗼𝗻 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 is unlocking for the people running daily operations. 

The long-standing debate between centralized and decentralized tech continues with limited consensus and a dire need for clarity [check out Steven Bartlett's always amazing podcast with Tristan Harris (ex Google) - in comments].  

And as capability moves to the edges, the question grows sharper. 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗮𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗲𝘀, 𝗯𝘂𝘁 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗺𝘂𝘀𝘁 𝗳𝗼𝗹𝗹𝗼𝘄. 𝗪𝗵𝗼 𝗴𝘂𝗮𝗿𝗮𝗻𝘁𝗲𝗲𝘀 𝘀𝗮𝗳𝗲𝘁𝘆 𝘄𝗵𝗲𝗻 𝗽𝗼𝘄𝗲𝗿𝗳𝘂𝗹 𝘁𝗼𝗼𝗹𝘀 𝗹𝗮𝗻𝗱 𝗶𝗻 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲’𝘀 𝗵𝗮𝗻𝗱𝘀?  It can’t rest on altruism.

Google, SWARM, #AI, #DecisionIntelligence

taplio

The Battle

Andreas Welsch

Shail Khiyara

Followers

Posts

Engagements

Likes

Performance ScoreShail Khiyara's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Andreas Welsch

Shail Khiyara

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Shail Khiyara's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreShail Khiyara's LinkedIn Performance