Udi Schlessinger

Since Anthropic’s release of Claude Cowork Plugins, I’ve been immersed in sell‑side and buy‑side notes predicting everything from the collapse of SaaS to the end of curated, expert‑driven content—including ~$285B of market cap erased in a single day. Then I read Nathan Graf’s recent FT op‑ed below, which takes a very different (almost opposite) view. That tension reminded me of a line often attributed to F. Scott Fitzgerald (and possibly borrowed from Hegel): “The test of a first‑rate intelligence is the ability to hold two opposing ideas in the mind at the same time and still retain the ability to function.” Leaving aside the “intelligence” test, what energizes me is the cognitive dissonance itself. Not choosing sides. Not collapsing complexity. But sitting with the discomfort. That tension—if embraced—can fuel better decisions, deeper creativity, and real innovation. It will also demand that leaders chart a course through grey zones, with nuance rather than dogma. My own view: The near- to medium‑term future belongs to man–machine hybrids. Look up “Centaurs” in freestyle chess—teams where human intuition is paired with machine‑scale logic. These hybrids consistently outperformed both humans alone and machines alone. I see a strong parallel in knowledge industries, where experts must make high‑stakes decisions in specialized domains for complex customer situations. Which leads me to a framing shift I think matters: 👉 We should think in terms of AND, not OR. Human expertise AND AI scale Curated knowledge AND generative intelligence Product innovation AND new pricing models As solutions evolve iteratively, we have to be patient as true adoption and results/outcomes take time. Add to that the willingness to pay—likely moving beyond traditional seats and subscriptions toward value‑ and outcome‑based pricing. Unlike passive AI, agents need to reason, plan, and act, requiring data that supports multi-step workflow. The data needs to be organized, enriched, chunked, updated, have diversity yet not overload and be specific to the task at hand - traditional providers do this well using human experts, but will further leverage passive or agentic AI to do this better, faster, cheaper. Finally, while many are declaring a SaaSocalypse, we be aware of on the potentially more advantageous cost structure (and margins) for a SaaS business over time vs. AI/Agentic model driven businesses where the costs grow with token consumption and usage in the same time period. Curious how others are navigating these tensions—are you seeing more “OR” thinking, or real attempts at “AND”? Views are my own and do not represent my employer or affiliates. https://lnkd.in/eJ8w7aHe #Leadership #AIStrategy #GenerativeAI #HumanCenteredAI #BusinessTransformation #DecisionMaking #InnovationLeadership

80

2 Comments

Only 1 of the top 50 AI companies are in Mass. There's been a lot of recent discussion on Boston's ecosystem being down bad and what to do about it. I put together some unpolished thoughts from my perspective, having been here for my career as a programmer, founder, and now investor. Boston's strength is that it's the world’s R&D lab, but our startup ecosystem has been stuck. Boston missed the deployment era of the web because it was too risk averse with too many MBA VCs wanting financial models and experienced founders. And this caution threatens the future of Boston's ecosystem today. I believe the core problem is cultural. Historically, Boston has optimized for pedigree and pragmatism, while SF optimized for the limitless ambition at the frontier. I'm here because Boston is the best place to build coming wave of AI-powered science. But to do that, we need to change the culture to ignore pedigree and to (re)embrace building the most ambitious technology to move humanity forward. Full post and ideas on how to fix it here: https://lnkd.in/e7qyYQN5

141

27 Comments

In a recent OneGuide Network discussion, Paul Barnhurst told us about testing AI Excel agents across dozens of real FP&A use cases. My favorite observation was that while AI is increasingly ready for Finance prime-time, it’s probabilistic and you should keep using deterministic logic in a lot of situations. 𝗨𝘀𝗲 𝗰𝗮𝘀𝗲𝘀 𝘄𝗵𝗲𝗿𝗲 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝗮𝗿𝗲 𝗰𝗿𝘂𝘀𝗵𝗶𝗻𝗴 𝗶𝘁: - Data cleaning (takes seconds vs. manual hours) - Writing single formulas or helping you think through which function to use - Documentation and error checking 𝗨𝘀𝗲 𝗰𝗮𝘀𝗲𝘀 𝘄𝗵𝗲𝗿𝗲 𝘆𝗼𝘂 𝘀𝗵𝗼𝘂𝗹𝗱 𝗸𝗲𝗲𝗽 𝘂𝘀𝗶𝗻𝗴 𝗵𝘂𝗺𝗮𝗻 𝘁𝗵𝗼𝘂𝗴𝗵𝘁: - Repeated reporting (too much randomness risk, go deterministic) - Building full, complex models from scratch (it can help, but you have to audit and re-work) Agents can help. But think "human-led, AI-assisted," not "AI builds everything." - - - Missed this one? We have more coming up! Running Opportunity Review Meetings for Sales - Friday, February 20th Leveraging Clay for Sourcing and Outreach - Friday. March 6th Trends in Specialized Operational Diligence - Friday, March 20th Crafting and Amplifying LinkedIn Thought Leadership - Friday, March 27th

11

2 Comments

The missing piece for generating Memos, Decks and Models just dropped. I've covered workflows where ChatGPT or Claude synthesize diligence, generate memos, and structure models. The catch: outputs stay trapped inside the chat window. You still have to move content into its final format, fix layouts, wire up formulas, and rebuild charts. Claude’s latest release removes that last-mile bottleneck. Claude can now create and edit spreadsheets, documents, slide decks, and PDFs. For the first time, you can prompt (and upload data) and get ready-to-use files in return. In the demo video we gave Claude - A data room - Our firm’s investment memo template - A single prompt And Claude was able to generate a ready-to-use investment memo that contained - Analysis and content for all the sections in our template - Inline charts and tables - Content and charts that matched our firm’s template (branding assets, color scheme, structure and formatting) Comment "prompt" and I'll send you a Google Drive folder link that contains step by step instructions, a prompt template, the inputs that were used in the demo and the final investment memo output. 

7

2 Comments

Analyst firms sell access to data - Why is HyperFRAME Research giving it away? The typical model for Analyst firms with proprietary data is to monetize via policing access to the data and licensing its usage. At HyperFRAME Research we are taking the exact opposite approach. We are posting not only the report, but the full data set on our website without any paywall, email registration or gate. We are also allowing anyone to use the data without restriction as long as they correctly attribute the source. Why? Because we want HyperFRAME Research to be the most cited research and advisory firm and we want LLMs to scrape, crawl and ingest our data.Paywall's don't allow that. Check out the report here 👇 https://lnkd.in/e8D-VFia

4

Yes, enjoyed this very much from Jed Sundwall. He's speaking plainly about how we must always think of the who will use our data and and how in the context of the "perfectly FAIR, utterly useless data" we keep spewing around everywhere. I've always been a fan of figuring *use* as a priority, in favour of the hard-to-know impact, especially when you hear factoids flying around that a very large portion of stored data—likely somewhere between 60-80%—is rarely or never accessed after initial storage.

4

Did AI just kill Market Research Firms? If you wanted a report like this from an old school consulting firm, you would have to shell out more than $50,000 and wait for more than a month. Semantic Similarity Rating (SSR) is a method that elicits textual responses from LLMs and maps these to Likert distributions using embedding similarity to reference statements. This framework enables scalable consumer research simulations while preserving traditional survey metrics and interpretability. Welcome to the future of Market Research and the cost is less than a $1 and runs in 3 minutes! https://lnkd.in/g5FGUQDZ

2