vencortex

Alpha generation needs to be aligned end to end - Across buy, build and exit. Decision governance ensures every choice along the investment lifecycle is aligned with your context constraints and goals as investor, thus bringing the required repeatability, transparency and discipline. David Sambur takeaways for the software industry: - Software valuations and growth assumptions are being reset as higher rates and AI challenge long-standing private equity underwriting models. - A significant portion of prior value creation in software was driven by multiple expansion rather than operational improvement. -Realized outcomes, including DPI, may diverge from prior marks as exits test valuation assumptions and fund rankings. The next cycle may favor disciplined underwriting, cash flow generation and assets with more durable and controllable value drivers. vencortex #hybridintelligence #decisionintelligence https://lnkd.in/d8MjWhXe

Alpha generation needs to be aligned end to end - Across buy, build and exit. Decision governance ensures every choice along the investment lifecycle is aligned with your context constraints and goals as investor, thus bringing the required repeatability, transparency and discipline. David Sambur takeaways for the software industry: - Software valuations and growth assumptions are being reset as higher rates and AI challenge long-standing private equity underwriting models. - A significant portion of prior value creation in software was driven by multiple expansion rather than operational improvement. -Realized outcomes, including DPI, may diverge from prior marks as exits test valuation assumptions and fund rankings. The next cycle may favor disciplined underwriting, cash flow generation and assets with more durable and controllable value drivers. vencortex #hybridintelligence #decisionintelligence https://lnkd.in/d8MjWhXe

Active investors must realize they aren't paid for reading faster. Speed of consumption is a poor proxy for alpha. For the active investor, deploying LLMs to automate mediocre processes will not yield superior results. It’s about seeing things no one else does, predicting probabilistic scenarios of the future, and then making the optimal choice given your constraints, strategy, and goalsconstantly updating as new information flows in. It’s about optimizing investment decisions. #hybridintelligence #decisionintelligence #ActiveInvesting vencortex Dr. Lorien Pratt Neil Brown Caroline Hedges CFA Charles Diebel

For years the investment industry believed the edge was information. Better data. Faster news. More research. Everyone was chasing the same thing. Access. AI is about to make that game irrelevant. Information is becoming abundant. Filings can be summarized instantly. Earnings calls broken down in seconds. Entire research notes generated on demand. But this misses the real point. Markets were never about information. They were about interpretation. Two investors can read the same 10-K and reach completely different conclusions. One sells. One buys. The document didn’t create the edge. The judgment did. AI will remove a lot of routine work across many industries. That much is obvious. But it will also expose something that has always been true. The people whose value was having information will struggle. The people whose value is knowing what matters will become more valuable. In other words, information is becoming a commodity. Judgment isn’t.

During a lifetime you will have 10-20 big winners and hundreds or thousands that didn’t live up to your expectations. Long-term investing success is capturing those 10-20 monsters while also understanding 95% of what you own will deserve to be sold.

The Three Models of Asian Shareholder Activism "The modern governance playbook requires a tailored approach, one that recognizes that activism in Tokyo is motivated by fundamentally different reasons from activism in Seoul or Singapore. "Japan - Because political defenses are fading, the battlefield has shifted entirely to the financial. Investors are focused on balance sheet efficiency and capital allocation. South Korea - The primary engine of this change is the government’s “Corporate Value-Up Program,” which has effectively deputized investors to enforce national policy. Singapore - In Singapore, the rules of activist engagement are dictated by simple math. Unlike in the United States, many listed companies here are majority-owned by a single family or a state-linked entity, meaning the “insiders” often hold more than 50% of the votes. Because of this structural “blocker,” activists have moved away from general popularity contests. Instead, a distinct new trend has emerged where investors target specific moments of failure — operational crises, debt defaults or lowball offers — where the board loses the moral authority to lead and the “insider” math stops protecting them." #governancearbitrage #asiangovernance #corporategovernance #boardsofdirectors #activistinvesting https://lnkd.in/grTWFbtr

When will the “asbestos abatement” phase for LLMs begin in anger? So much rectifying to be done due to the pervasive, toxic sprawl of vibe coded LLM-based software applications that got shoved into automated business decision execution flows without common sense (aka pre-LLM) design, monitoring or governance. But there are a variety of approaches needed depending on how deep the rot has taken hold in your organization, and how much “reality” we’re willing to take: - Misagentification: incorrect or premature deployment of an LLM based AI agent resulting from flawed design, configuration, or risk scoping that activates unintended autonomous authority - Disagentification: distorted, misrepresented, or poorly governed deployment of an LLM based AI agent that creates inaccurate assumptions about its autonomy, control boundaries, or accountability - Malagentification: deployment of an LLM based AI agent that produces harmful, unsafe, or systemically damaging outcomes regardless of original intent - Deagentification: decision governance driven reversal or reduction of autonomous authority in a deployed LLM based AI agent without necessarily removing it from the architecture - Exagentification: architectural extraction of a deployed LLM based AI agent from live decision flows, removing it from runtime execution and orchestration pathways Can the “black box” even be shut anymore? #LLM #AISafety #DecisionGovernance

Public Company Boards: Prepared For Increasing Company Transformations? A New Level of Urgency For A Change In Governance Model "This is from an interview of Chris Hohn, activist investor and Christian Sinding, CEO of EQT, a Swedish private equity firm with $227 billion under management. The part of the interview below is Christian Sinding responding to the interviewer’s question, “So how do you do that? You identify a company that you want to buy, pay a price that you're happy to pay, and then what's the playbook to create the value?” Underlying Sinding’s response is the mindset and approach to business that is desperately needed on the part of public company boards:" #governancearbitrage #valuecreation #fullpotential #corporategovernance https://lnkd.in/gXRqZeaV

We have a new report out called: "Bayes and Base Rates: How History Can Guide Our Assessment of the Future." *The ongoing massive investments in artificial intelligence (AI) aim to satisfy a huge increase in anticipated demand, which in turn has led some firms to offer rosy growth forecasts. *To assess these forecasts, investors may apply Bayes' Theorem by starting with an initial belief and updating it as new results appear. Base rates are a sensible start for initial beliefs.  *The base rates for U.S. public companies over 75 years suggest OpenAI and Oracle Cloud have a low probability of meeting their five-year revenue projections. Offsetting this are data showing rapid diffusion of AI, which signals major demand and short-term growth.  *A large database of projects shows less than 10 percent are completed on time and on budget, which should temper overly optimistic expectations for the buildout of AI infrastructure. *Companies sometimes pursue a preemptive strategy in which they announce big capacity commitments to deter competitors and entrants from investing. Link in comments

One big realization in 2025: I can get a lot of research funding at MIT to rigorously prove that LLMs cannot—and will not—solve most real-world time series and tabular data problems. Is it a good research question - sure. Is it productive for our society - probably not. For those who know my work: I’ve spent years on tabular, relational, multi-table, and time series data. In our lab, we build models on real enterprise data and help organizations extract value from it. In April 2024, we ran a serious test: can LLMs natively solve a time series task? We evaluated GPT (via OpenAI API) and Mistral. We were thorough. The result: some promise, but it fundamentally did not work. We published the paper and the rationale. We also followed up with a venturebeat article about it. Since then, almost everyone I meet at MIT asks the same question: Can LLMs or generative AI be used for this? After enough conversations, I found telling myself and jokingly said that to an exec from an enterprise: “I could probably make a lot of money just proving LLMs won’t work for your problem.” To be clear, I don’t think LLM companies themselves want to frame the technology this way. There is enormous value in language, and the language market alone is massive. That’s where LLMs shine. But for learning models for tabular + time series data today, they:  • incur ~20-100× more time and costs  • provide much lower accuracy  • require heavy pre- and post-processing  • and do not scale to industrial datasets In 2026, we’ll do this one step at a time: demystify what LLMs can do—and what they can’t and quantify the productivity loss in going after them. #generativeAI #LLMs #tabulardata