על אודות
פעילות
9K עוקבים
-
Daniel Shimoni שיתף את זהThis is how fast the shift happened: Less than a year ago, I spoke with dozens of data leaders about their challenges with AI. Over 80% of them said they wouldn't trust AI to write the SQL or Python code their teams wrote. In general, they had serious reservations about anything AI gave them back. That feedback was a big part of why we decided to build the Context Warehouse, to make AI trustworthy at scale. Fast forward to today, less than 12 months later. If I asked those same people, "Do you still expect your team to write code without using AI?" I don't think I'd get a single yes. The trust issue didn't get solved. It got outrun. It's up to the industry now to build the tools that make AI workable and reliable.
-
Daniel Shimoni שיתף את זהMapping context for your agents looks like a project with an end. You know the questions people ask. You mostly know where the data lives. So you pick the systems that should be most relevant, map what you can, and figure you'll be done in a quarter. What you're actually doing is predicting what an agent will need. Predictions only hold in a sandbox. Data shifts, teams reorganize, someone redefines what counts as an active customer, and the map you built in March is quietly wrong by June. So either your agents give confident, expensive, and wrong answers. Or you spend every week from now on maintaining the map instead of building what your customers ask for. To make AI work for your org, you've created a task you can never complete. I don't think the answer is mapping faster. Scope has to be decided when the question arrives, not six months before it.
-
Daniel Shimoni פרסם/פרסמה את זהSince founding Modus, I've done more outbound calls than I can wrap my head around. What I didn't expect was how much of it feels like product work. Before Modus I spent years thinking about the person on the other side of the screen. What are they trying to get done? What's in their way? Why would they trust this thing enough to change how they work? Outbound is similar. Someone gets a message from me. They don't know me. Why should they care, if all I want is something from them? So I've been trying to approach it the way I'd approach a user. Be curious about what's actually hard for them, and respect that their time is the scarcest thing they have. The calls that actually go somewhere usually start from an honest guess about someone's week. ’Something’s troubling you, I might be able to help, let's talk.’ The first step of a great product or GTM effort is deeply caring about people. There’s too much noise out there for people who want to sell.
-
Daniel Shimoni שיתף את זהSuper exciting, but even more than that - super interesting. If you're trying to figure out how to build a company brain, this is your chance to learn from some of the best people working on this problem today.Daniel Shimoni שיתף את זהRegistration is officially LIVE! Grab your spot for our meetup with Sola Security and Microsoft for Startups. Seats are limited👇 https://lnkd.in/dkUzs6tt See you on Aug 31st! 🧑🎓 Tomer Mesika | Leon Goldberg | Meital Shamia | Roey Zalta | Opher Hofshi #Context #CompanyBrain #ContextWarehouse
-
Daniel Shimoni שיתף את זהIt’s NEVER buy vs build. It’s always buy vs build+maintain. Every team that "built context internally" priced it according to how much it’ll cost to build and forgot the maintenance. Maintenance isn't a just as marginal addition to your business. It’s huge. Because businesses aren’t stale. They change: Sources shift Schemas drift Definitions evolve Business KPI are added or removed The person who built the context layer leaves. It’s part of any business. But when this happens after you spend a small fortune building a company brain or a context layer - it can go stale or just become irrelevant within months. The economics of internal build are pretty clear: 1-20 engineers costing at least $200K-$4M a year in headcount costs. These engineers will not be making your users happy by working on your business roadmap, and they’ll have a forever piece of work to maintain. The build is a project. Keeping it up to date is a forever-job you didn’t budget for. The real comparison of ‘build vs buy’ isn't your engineers vs. a vendor. It's your engineers, forever, vs. a vendor. In 99% of the cases, from an economic standpoint, building a context layer yourself is not the right call.
-
Daniel Shimoni שיתף את זהA wrong answer isn't the worst answer you can get from your AI agents. An answer that's 75% correct is way worse. You’ll ignore obviously wrong answers. A 75% correct answer.. you might be fooled to trust that one. It will make enough sense. And it throws you to ‘almost accurate’ data sources. But the conclusion you’ll draw might be completely wrong and send you down a rabbit hole of tokens, hours, and useless work - or worse, it might lead you to a domino effect of wrong decisions based on an initial wrong business context. This is AI hell right there. It’s almost right. It’s partially right. It’s mostly right. It’ll never be right enough for large scale or delicate processes. No new version is going to change that. Only a Context Warehouse will.
-
Daniel Shimoni שיתף את זהHave you ever had a moment when someone else explains your market and product like he’s been in the same room with you for months? Well, we couldn’t ask for a better moment of clarity than this - Nikesh Arora, CEO of Palo Alto Networks, explains why context is going to be the main differentiator, not model improvements. Context is going to get the main spotlight on the AI stage during the next 5 years. And any company who wants to go all-in on AI has got to invest in a context solution, cause no model upgrade will solve this. Thank you for that, Nikesh. This is just a snippet for a very insightful conversation with Harry Stebbings, Jason M. Lemkin, and Rory O'Driscoll on the 20VC podcast. Full episode here: https://lnkd.in/dbxRTqtt
-
Daniel Shimoni שיתף את זה"We're building our company brain" If you're in tech, I bet you've heard that one at least a few times during the last 6 months. This is absolutely the right move if done correctly. Problem is, it could also become an extremely expensive mistake. Here's why: https://lnkd.in/de9r_iWbWhy the Company Brain You're Building for Your CEO Might Be Your Biggest MistakeWhy the Company Brain You're Building for Your CEO Might Be Your Biggest Mistake
-
Daniel Shimoni שיתף את זהMany startups begin with an insight. Ours began with frustration. A huge one. This time last year I was still a VP of Product at a company with tons of data and a lot of questions about that data. I loved how data-driven I had to be to succeed in that role. But it was also a struggle. A big one. See, as a VP Product, I was responsible for several company-level OKRs. And like many product leaders, I was expected to have answers at my fingertips to any question about anything that happened in any of our products or features. I can't count the number of times I heard the question "What happened to retention yesterday?" You’d think that for a company with so much data, finding the answer would be easy. It wasn’t. It was always a rabbit hole. I’d look at data from Tableau. Snowflake. Google Analytics. Salesforce. Amplitude. And then realize that to get the answer, I’d need one or two analysts spending a full day figuring it out. That experience stayed with me. The frustration. The feeling of not being fully on top of what I was responsible for. I remember that vividly. Modern companies have more data than ever before. And the bitter irony is that all this data still doesn't get us clear answers to simple business questions. No human can manually connect all the data scattered across dozens of systems. Later, when Tomer and I compared notes, we realized we had both been living the same problem from different angles. Not a data problem. There's almost too much of that. A context problem. The information exists. It just lives in too many places, disconnected from the questions people are actually trying to answer. And the AI everyone expects to fix this hits the same wall. It reads the same systems, without the context that would make any of it mean something. Some people call the missing piece a company brain. Others call it a context layer. We call it a Context Warehouse. That gap between what a company can access and what it actually understands is what we've spent the last 6 months closing. That’s how modus was born.
-
Daniel Shimoni אהבתי את זהYour ranking system probably doesn’t have preferences. It has positions. Ask an LLM which of two vendors better fits a requirement, which of two accounts is the stronger opportunity, which of two draft responses goes to the customer — and a meaningful share of what comes back is determined by which one you listed first, not by judgment about the options. The part that should worry product teams is that this isn’t the model shrugging when it can’t tell the difference. New work in PNAS Nexus separates those two cases and finds genuine reversals: the model has a preference, and presentation order overrides it. The direction also flips with quality. Strong options get a first-position advantage, weak ones get a last-position advantage, so the instability is largest exactly where your candidates are closest and the call is hardest. Randomizing order just randomizes who benefits. Running every permutation and taking the majority returns a clean tie precisely when the preference is fragile, so the distortion reads as indifference and passes review. And in B2B this compounds, because ranking rarely happens once — an account gets scored, the top slice gets re-ranked for routing, the survivors get compared again for messaging. Order effects at each stage aren’t independent errors that wash out; they’re a systematic tilt applied repeatedly to a shrinking set. The useful inversion is to stop treating instability as noise: if a choice flips when you swap the order, that’s a signal the options are close, which is a better trigger for human review than any confidence score the model will hand you. One confident answer tells you nothing about whether the preference behind it is real. #AIProduct #LLMEvaluation #ResponsibleAI #DecisionSystems #EnterpriseAIFragile preferences: A deep dive into order effects in large language modelsFragile preferences: A deep dive into order effects in large language models
-
Daniel Shimoni אהבתי את זהSo excited to be hosting Insight Partners ScaleUp:AI conference with some great guests and friends Save the date October 6th in NYC - also join us virtuallyDaniel Shimoni אהבתי את זהExcited to announce the first round of speakers for ScaleUp:AI 2026 — October 6 in NYC. Joining us this year: • Bobby Yerramilli-Rao — CSO & CVP, Microsoft • Ron Gabrisko — CRO, Databricks • Sarah Yager — Human Rights & Responsible Deployment Lead, OpenAI • Sebastian Mallaby — Senior Fellow, Council on Foreign Relations; Author, The Infinity Machine • Sonali Basak — Chief Investment Strategist, iCapital • Tasso Argyros — VP Engineering, Databricks • Gen. (Ret.) Tim Haugh — Former Commander, U.S. Cyber Command & Director, NSA From enterprise AI and national security to responsible deployment — more speakers to come. 🔗 https://lnkd.in/eTdpYRM4
-
Daniel Shimoni אהבתי את זהDaniel Shimoni אהבתי את זהAI is making all marketing orgs sound the same 😔 But that doesn't mean we should shy away from using AI. My bet is that the teams that stand out are the ones with a curated, thoughtful context layer. Your messaging, your positioning, your point of view, your taste, your unique ideas, your sales strategy. Things you used to think of as just documentation...become your unique edge in a noisy market. A team with a highly thoughtful and curated sales enablement library can point AI at it and get a very useful outbound plan. A team without one gets a hallucinated deck full of "5 ways to accelerate revenue." A team with a mature design system and curated brand story can point AI at it and ship strong landing page in hours. A team without one ships generic AI-slop marketing pages that every reader can smell from a mile away. Which means the AI unlock for most GTM teams isn't really an AI project. It's treating the stuff you already have (messaging, positioning, enablement, design, your perspective on the market) as infrastructure instead of just knowledge or documentation. Curating it carefully with human judgement. Keeping it authoritative and grounded in what you uniquely believe. And wiring AI to draw from it. It becomes infrastructure. Which is way more powerful than documentation. Are other GTM folks thinking this way, does anyone have opinions on the most important piece of context to "curate"?
-
Daniel Shimoni אהבתי את זהDaniel Shimoni אהבתי את זהThe irony of AI: The more you use it to produce more "stuff" the more people expect you to produce more "stuff". We're quickly heading into a cycle of limitless output. More output might feel good initially, but it might destroy everything. No one is producing more time while producing more "stuff" even if they've got agentic employees that are supposedly processing "stuff" and streamlining "things." I love what Cory Doctorow wrote recently about sending someone unverified AI output, "...is an attempt to coerce a stranger into unpaid labor on your behalf." Even if they're not a stranger, you're still outsourcing the actual work (read: "thinking") to the recipient. If there was ever a time to hang our hats on "quality over quantity" it's now.
-
Daniel Shimoni אהבתי את זהDaniel Shimoni אהבתי את זהYour AI Agent Has a Governance Problem What are the BIGGEST GOVERNANCE PROBLEMS AI agents are exposing in your organization ??? Is it context ... accountability ... behavior change ... governance reality ... people … or something I completely missed ? #datagovernance #noninvasivedatagovernance #NIDG #datacatalyst #changemanagement #datafluency #AIagent #AIagentgovernance
ניסיון חינוך
ניסיון בהתנדבות
-
Community Volunteer
Mechinat Ha’Negev Program for Leadership (Sde Boker, Israel)
- 1 שנה
ילדים
• Volunteered at the ‘Chaim’ Organization as a tutor for children recovering from cancer.
• Volunteered at the ‘Etgarim’ Organization as an assistant instructor of outdoor activities for children with disabilities.
שפות
-
English
מיומנות שפת אם או דו-לשונית
-
Hebrew
מיומנות שפת אם או דו-לשונית
המלצות התקבל
2 אנשים המליצו Daniel
הצטרפו עכשיו כדי נוףפרופילים דומים אחרים
גלה פוסטים נוספים
-
CHAMBER OF COMMERCE ISRAEL - UKRAINE
5 Israeli Startups to Watch Ahead As “Startup Nation”, Israel consistently launches more cutting-edge startups per capita than the rest of the world. Here are 5 Israeli startups that are riding the next tech wave. - Base44, the AI-powered no-code app builder, was launched in 2025 by Maor Shlomo. The startup transforms a text prompt into a fully functional web app or game. Base44 had over 100,000 users and partnerships with Israeli tech leaders like eToro and Similarweb. In 2025, Base44 was acquired in 2025 for $80 million by Wix. - Decart restyles streams, webcam feeds, and gameplay with AI. The startup was founded in 2023. In 2025, Decart raised $100 million in Series B funding at a $3.1 billion valuation. Dean Leitersdorf co-founded the startup in 2023. - VFR developed a dynamic ad player that powered by real-time user data and places ads where it makes sense. In 2025, VFR secured $10 million at a $150 million valuation. - Styletech is the AI-powered platform that makes a single image runway-ready in seconds. Brands can mix and match looks, models, and campaigns at the click of a button to cut both costs and complexity. - Prompt Security built the firewall to keep generative AI safe and under control in real-time. Founded in 2023 by Itamar Golan and Lior Drihem, it was acquired by SentinelOne. Photo: Prompt Security Source: https://lnkd.in/dXY23bZ7
-
Chaim Oren
𝐒𝐚𝐥𝐞𝐬𝐟𝐨𝐫𝐜𝐞 𝐜𝐮𝐭 𝐬𝐮𝐩𝐩𝐨𝐫𝐭 𝐫𝐨𝐥𝐞𝐬 𝐚𝐬 𝐀𝐈 𝐭𝐨𝐨𝐤 𝐨𝐯𝐞𝐫, 𝐲𝐞𝐭 𝐫𝐞𝐯𝐞𝐧𝐮𝐞 𝐦𝐢𝐬𝐬𝐞𝐝. 𝐓𝐡𝐚𝐭 𝐢𝐬 𝐭𝐡𝐞 𝐥𝐞𝐬𝐬𝐨𝐧. Question for Israeli businesses: Will your customers stay satisfied when AI handles the first mile, and the last mile, or only one of them? When I face these challenges, I turn to my friend Chris Dyer, a culture expert, tech CEO, and three-time bestselling author. Here is what he told me works, based on his experience with NASA - National Aeronautics and Space Administration, Johnson & Johnson, Citi, and Berkshire Hathaway. AI can lower cost to serve, shorten wait times, and scale 24x7. It does not guarantee growth or loyalty. Salesforce says AI agents now handle a large share of support, and about four thousand roles were removed, while the revenue outlook still softened. Learn the right takeaway: productivity is not valued by default. What to do now? • Classify work, automate repetitive tickets, reserve humans for high emotion and high value moments • Track customer outcomes, FCR, CSAT, churn, downgrade rate, time to human handoff • Design the handoff, clear escalation paths, response time targets, and ownership by name • Tighten knowledge, weekly refresh of playbooks and prompts, retire stale answers • Stress test ethics, privacy, disclosure that an agent is an agent, easy opt-out for a person If you would like Chris to work directly with your firm in Israel, please message or email me at: orengroup1@gmail.com. #AI #CustomerExperience #Salesforce #Automation #BusinessStrategy##StartupNation#IsraelTech