Verve Group Se Class A · OMXSTO:VER
Verve Group Se Class A — Investor Day (June 2026)
136 slides from Verve Group Se Class A's investor day presentation, June 2026
Slide 1: Verve Capital Markets Day 2026
June 2026 Verve Capital Markets Day 2026
Slide 2: Disclaimer
Disclaimer THE INFORMATION CONTAINED IN THIS PRESENTATION IS STRICTLY CONFIDENTIAL. ACCORDINGLY, THE INFORMATION INCLUDED HEREIN MAY NOT BE REFERRED TO, QUOTED OR OTHERWISE DISCLOSED BY YOU, NEITHER DIRECTLY OR INDIRECTLY NOR WHOLLY OR PARTLY. BY REVIEWING THIS INFORMATION, YOU ARE ACKNOWLEDGING THE
Slide 3: The Mobile Advertising Intelligence Ecosystem
The Mobile Advertising Intelligence Ecosystem The open internet has a powerful alternative to walled gardens. And it lives in your pocket. While the walled gardens capture attention, the open internet captures intent. We turn the world's most underleveraged signals into actionable intelligence, givi
Slide 4: Capital Markets Day Speakers
Capital Markets Day Speakers Moderator Guest Speaker Ingo Middelmenne Dr. Usama Fayyad Head of Investor Relations Professor at Northeastern University Verve Speakers Remco Westermann Christian Duus Raimund Bau Mishel Alon CEO CFO President, Verve Retail Media CBO 4
Slide 5: Capital Markets Day Overview
Capital Markets Day Overview Business Updates 9:00 Introduction 9:10 Verve’s Equity Story & Commercial Update 10:00 Financial Update 10:30 Q&A Session 1 11:00 Break Expert Sessions 11:30 Keynote by Dr. Usama Fayyad 12:00 Closed Loop Retail Media: The Compounding Outcome Engine 12:30 From Signals to
Slide 6: Verve's Equity Story & Commercial Update
Verve’s Equity Story & Commercial Update
Slide 7: The Mobile Open Internet Market
The Mobile Open Internet Market
Slide 8: Mobile: The Only Advertising Channel with a 1:1 Consumer Relationship
Mobile: The Only Advertising Channel with a 1:1 Consumer Relationship Mobile carries the largest advertising TAM globally and continues to grow as the primary access point to the consumer. ● Always-On Consumer Gateway ● Unmatched Signal Depth ● Closed-Loop Outcomes The Only Channel the Consumer The
Slide 9: The Mobile Market Represents a Huge $68 Billion Opportunity
The Mobile Market Represents a Huge $68 Billion Opportunity A $68B market today, growing at 11% CAGR through 2030. That is before accounting for the structural shift already accelerating toward the open internet. Proxy A Proxy A Proxy B Total Global Mobile Ad Spend¹ Excluding Mobile Search² Mobile N
Slide 10: LLMs Negatively Impact Web & Search While Driving Mobile In-App Traffic
LLMs Negatively Impact Web & Search While Driving Mobile In-App Traffic AI is redirecting the open web. The users with the highest intent are landing in apps, and mobile is where those signals are richest. ● SparkToro & Datos ● Gartner Group ● Sensor Tower ● Reuters Institute Nearly 60% of searches
Slide 11: The Open Internet is the Most Undervalued Medium in Advertising
The Open Internet is the Most Undervalued Medium in Advertising Ad budgets remain heavily skewed toward walled gardens despite the open internet commanding most consumer attention and time. ● Walled Gardens ● Open Internet ● Ad Spend per User Hour Third-party Mobile Apps (Open Internet) $0.07 80% 69
Slide 12: The Open Internet's Challenges Are Verve's Opportunities
The Open Internet’s Challenges Are Verve’s Opportunities The fragmentation, data quality, and measurement gaps that limit in-app advertising today are precisely where Verve competes. Scaled Access to Cost Efficiency & Quality & Depth Precision Targeting Measurement & Mobile Inventory Transparency of
Slide 13: How Verve Is Built To Win The Open Internet
How Verve Is Built To Win The Open Internet
Slide 14: Verve Brings Walled Garden Performance to the Open Internet
Verve Brings Walled Garden Performance to the Open Internet Five capabilities, fully integrated, that give advertisers the scale, signal depth and closed-loop accountability they have only ever found inside walled gardens. Scaled Access to Vertically Integrated Depth of Targeting Corporate Mobile In
Slide 15: Scale Matters: Verve Built a Top 15 AdTech Platform in Under 6 Years
Current Status Scale Matters: Verve Built a Top 15 AdTech Platform in Under 6 Years Leveraging Direct SDK Integrations to Reach 2.5 Billion Global Consumers via First-Party Signal Intelligence. €579M 4,086 1,135 +65,000 LTM Revenue, Q1‘26 Software Clients >$100k Clients Apps +2.5B +1.2T ~1,000 End C
Slide 16: Scale Matters: Reinforcing Scale in Core Markets, Verticals and Channels
What's Next Scale Matters: Reinforcing Scale in Core Markets, Verticals and Channels Starting from a position of strength in mobile, expanding across channels and deepening in the markets where we already win. Growing The Sales Force to Cover More of the 4,000 Agencies Deepening Advertiser Access Ex
Slide 17: Vertically Integrated: From Acquisitions to One Platform
Current Status Vertically Integrated: From 18 Acquisitions to One Platform Verve connects advertiser demand and publisher supply through a vertically integrated platform. DSP, SSP, and data operating as one. ● Advertiser 1.2T ● Publisher / Consumer Ad Impressions LTM Q1 2026 $ 0.9 “AI-driven matchin
Slide 18: Vertically Integrated: Building a Best-in-Class Ad Intelligence Platform
What's Next Vertically Integrated: Building a Best-in-Class Ad Intelligence Platform One fully integrated stack by Q2 2026, Focus on AI-driven matching across every layer from bid optimization to audience prediction. Our Growth Focus The Final Integration Routing More Demand Moving to an AI Driven M
Slide 19: Data Depth & Curation: Five Tiers of Signals, One Intelligence Platform
Current Status Data Depth & Curation: Five Tiers of Signals, One Intelligence Platform From deterministic identifiers to conversational intent, converging into one intelligence platform. The broadest tier by reach, but increasingly Tier 1 Legacy Identifiers Cookies & Opt-in IDFA/GAID restricted as i
Slide 20: Data Depth & Curation: Maximizing Data to Reinforce Outcomes
What's Next Data Depth & Curation: Maximizing Data to Reinforce Outcomes The more signals we integrate, the more value we compound for every advertiser and publisher on our platform. Our Growth Focus Optimize SDK & Exchange Data Expand Signal Access Sharpen the Prediction Model Scale SDK opt-in reac
Slide 21: Targeting & Measuring: Outcomes are What Matters
Targeting & Measuring: Outcomes are What Matters An AI feedback loop converting our rich data signals scale into high-precision targeting and yield. ● The Foundation Has Shifted ● Verve Operates Across Both Environments ● Measurement Across Both Environments o GDPR, CPRA and ATT permanently redrew o
Slide 22: Targeting & Measuring: Next Level - Establishing Sector-Focused Solutions
What's Next Targeting & Measuring: Next Level - Establishing Sector-Focused Solutions Translating sector-specific signal intelligence into real-world business outcomes. Our Growth Focus Closed Loops Attribute Digital Spend Directly To Real World Outcomes 1 - Signal capture Each sector is different;
Slide 23: Corporate Development & Balanced Growth
Corporate Development & Balanced Growth Organic growth, operational discipline, and AI-driven efficiency form the foundation for the next phase of scaling Verve. Focused Investment in Focus on Internal Emphasis on Core Growth Areas Efficiencies Organic Growth Targeted organic investments only, focus
Slide 24: Verve Brings Walled Garden Performance to the Open Internet at Scale
Verve Brings Walled Garden Performance to the Open Internet at Scale Five deeply integrated capabilities that together deliver what the open internet has always needed: walled garden performance without the closed ecosystem. Scaled Access to Vertically Integrated Ad- Depth of Data and Targeting and
Slide 25: Financial Update
Financial Update Christian Duus, CFO
Slide 26: Sustained Long-Term Growth Performance
Sustained Long-Term Growth Performance Proven ability to convert investments into sustainable revenue growth and attractive returns over time Revenue Growth Adj. EBITDA Growth +28% +19% €m €m 705 36% CAGR 160 32% CAGR 133 134 551 437 95 93 3031 322 71 252 140 29 2020 2021 2022 2023 2024 2025 2026E 2
Slide 27: Four Structural Drivers Enabling High Organic Growth
Four Structural Drivers Enabling High Organic Growth Strong market tailwinds complemented by Verve's expanding portfolio of growth initiatives and strategic differentiators 4 • Network effects enhance • Data moat enhanced with intent • Economies of scale lead to Platform Platform Verve’s relevance t
Slide 28: First Quarter Financial Highlights
First Quarter Financial Highlights Strong organic growth of 6.4% YoY and operating cash flow generation, profit level impacted by strategic investment phase. Significant FX headwinds of 9.6% in the quarter. Q1 2026 Revenues and Adj. EBITDA1 Q1 2026 Sources of Growth +3.7% LfL Revenue Growth +5.3% Or
Slide 29: Solid Customer and Impression Growth Coupled with Structural Retention
Solid Customer and Impression Growth Coupled with Structural Retention We are very successful at landing clients; and increasingly focusing on scaling these into ‘large software clients’ >$100K. Ad Impressions (billions) Number of Software Clients3 +25% Total 310 311 4,086 288 3,734 248 259 3,495 2,
Slide 30: Structural Improvement in Gross Profit Margin Carries Through to Q1
Structural Improvement in Gross Profit Margin Carries Through to Q1 Unified supply side platform is more efficient and is benefiting from dynamic margin management, combined with lower hosting costs. Gross Profit Margin Like-for-Like1 +2.7%-pts Q1’26 vs Q1’25 Gross Profit Margin Gross Profit = Reven
Slide 31: Commercial Expansion Progressing Well and Positioned to Capture Future Growth
Commercial Expansion Progressing Well and Positioned to Capture Future Growth €10m investment in sales talent and coverage is laying the foundation for accelerated growth across brands and agencies Upfront Investment Progress to Date €10m OPEX investment in 2026 117 salespeople in Q1 2026 Investment
Slide 32: Strong Cash Flow Generation Despite Periodic Working Capital Timing Effects
Strong Cash Flow Generation Despite Periodic Working Capital Timing Effects Underlying cash conversion at 70-80%, with Q1'26 working capital normalization restoring LTM performance to historical levels. Operating Cash Flow1 Cash Conversion (OCF/Adj. EBITDA) €m 137 116 116 105 94 103% 69 86% 79% 57 7
Slide 33: Strong Revenue Growth While Maintaining Working Capital Discipline
Financial Update Strong Revenue Growth While Maintaining Working Capital Discipline Our receivables securitization program helps mitigate working capital build-up from growth. Development in Trade Receivables and Trade Payables 2024-2026 Structural 45-day timing gap creates ongoing funding need 194
Slide 34: Verve's €100m Receivables Securitization Program
Financial Update Verve‘s €100m Receivables Securitization Program Converting receivables into strategic growth capital to address NWC needs Offering highly attractive payment terms for clients in an Emerging Channel: Demand Side pay after 90 days Liquidity + Scalability "Liquidity as a Moat" Supply
Slide 35: Growth Planned Without Heavy Capital Requirements
Growth Planned Without Heavy Capital Requirements Organic growth primary investment priority: low maintenance CAPEX, scalable expansion CAPEX, and disciplined M&A CAPEX Development CAPEX Characteristics & Allocation Priorities 1621 Acquisitions: Hereof only 9.9m in deferred payments remaining Low Re
Slide 36: Equal Focus on Executing Growth Opportunities and Reducing Leverage
Equal Focus on Executing Growth Opportunities and Reducing Leverage Strong cash flow generation and disciplined liquidity management to reduce net debt and support 2.0x–2.5x leverage target Adjusted Leverage Ratio Interest Coverage Ratio 1 2 €m €m 147 146 2 446 448 142 1 421 134 351 295 95 3.1x 3.1x
Slide 37: Bond Market Access Has Enabled Growth at Increasingly Attractive Terms
Bond Market Access Has Enabled Growth at Increasingly Attractive Terms Verve has demonstrated repeated refinancing and tap execution capability well in advance of maturity 55% of our total outstanding bonds are hedged €12.5m in yearly interest savings 2027 and onwards 550 500 50 465 65 350 80 225 17
Slide 38: Guidance
Vv Guidance FY2026
Slide 39: Four Structural Drivers Enabling High Organic Growth
Four Structural Drivers Enabling High Organic Growth 2026 Revenue Expectation: 7-9% growth from underlying market growth, plus minimum 5%-pts from market share gain 4 • Network effects enhance • Data moat enhanced with intent • Economies of scale lead to Platform Platform Verve’s relevance towards d
Slide 40: Guidance for 2026 Reaffirmed
Guidance for 2026 Reaffirmed H1 profitability impacted by strategic upfront sales team build up FY 2025 FY 2025 FY 2026 Mid-point Actuals LfL Guidance YoY Uplift Revenue (like-for-like) (in €m) 551 602 680 – 730 +17% Adj. EBITDA (in €m) 134 134 145 – 175 +19% As exact timing of 'sales-productivity i
Slide 41: Strategic Further Corporate Development
Strategic Further Corporate Development Driving alignment of corporate structure with international footprint and US-centered AdTech environment Our Future Agenda Moving to Ireland as EU common-law domicile The relocation to Aligns Verve’s constitutional framework with international technology and I
Slide 42: CFO quote: Verve is entering its next growth phase
"Verve is entering its next growth phase and confidently accelerating growth through targeted investments where returns are highest ... … while focusing on preserving the capital discipline, cash flow profile, and financial strength that will underpin long-term shareholder value." Christian Duus, CF
Slide 43: Q&A Session
Vv Q&A Session 1
Slide 44: Coffee Break
Vv Coffee Break See you at 11:30
Slide 45: Capital Markets Day Part 2
Vv Capital Markets Day Part 2
Slide 46: Keynote
Keynote Dr. Usama Fayyad, Northeastern University
Slide 47: AI Transformation in Marketing & Advertising
Al Transformation in Marketing & Advertising: Recent Developments, & Strategic Considerations Usama Fayyad, Ph.D. Chairman & Founder - Open Insights SVP, Inaugural Director - The Institute for Experiential AI at Northeastern University June 10, 2026 – Boston June 16, 2026 – New York v3 Verve Group C
Slide 48: Have we not heard enough about AI?
Have we not heard enough about AI? The hype is unavoidable, and the expectations are ● overblown ● mostly confused… and very confusing ● In many cases underestimating the true value 48 Copyright, U. Fayyad © 2026
Slide 49: The Hype versus the Reality?
The Hype versus the Reality? What is the reality for 2026? ● AI Technology is “simple” yet very powerful in the knowledge economy ● The AI does not “understand” or “analyze” ○ simulation via auto-complete is amazingly effective ○ a very useful and flexible information retrieval modality ○ What chatb
Slide 50: How our Brain identifies Missing Values
BRAIN GAMES How our Brain identifies Missing Values 50 Copyright, U. Fayyad © 2026
Slide 51: Example in Banking and Fintech: Missing Data
Example in Banking and Fintech: Missing Data Can we use AI - to “complete” our view of the customer in KYC? While we have may data sources, they are disparate and incomplete Can we have a “brain” that can work effectively with incomplete views? 51 Copyright, U. Fayyad © 2026
Slide 52: Example in Marketing & Advertising: Missing Data
Example in Marketing & Advertising: Missing Data Data on content, Partial data from mobility, intent, publishers and Apps business/publisher data Detailed data - Similar consumers, Brands, stores, environmental transactions factors, etc. Can we use AI - to “complete” our view of the consumer? Can we
Slide 53: The Last 4 Years of AI
The Last 4 Years of AI 2023: Fascination ● amusement and curiosity 2024: Denial ● Not really relevant to our business ● This is a passing trend… 2025: Delusion ● We are using ChatGPT, and therefore we are adopters of AI ● A few real business uses emerge 2026: Panic: AI is Real and We are very much b
Slide 54: Data: Some of the Big Challenges for AI
Data: Some of the Big Challenges for AI Successful AI is totally dependent on ML/Data Science, hence need good training data: Data remains a huge challenge for most organizations Good training data is extremely expensive to get … reliable labelling even more expensive Just collecting and managing ra
Slide 55: Secret 3: Making AI Work
SECRET 3: Making AI Work (from talk on Secrets of Making AI Work) 3 Collect All Data at Granularity - for Algorithms: Capture “Events, Outcomes, and Context” at as fine a granularity as possible Focus on Structured as well as Unstructured data Majority of Data in any organization is Unstructured (90
Slide 56: How do we make this Data work for the business?
Must Capture Data at High Granularity But most businesses are not equipped to effectively manage data as an asset How do we make this Data work for the business? New economy of Interactions is rich with unstructured data in fact, 90% of Data in any organization is UNSTRUCTURED (Gartner) Without prop
Slide 57: Bridging the Great Divide in the World of AI and Data
Bridging the Great Divide in the World of AI and Data Copyright, U. Fayyad © 2026
Slide 58: AI Haves
AI Haves Companies Actively driving innovation 58 Copyright, U. Fayyad © 2026
Slide 59: Have Nots and AI Haves
Have Nots Companies observing AI AI Haves evolution from distance Companies Are you here or on the Actively driving other side ? innovation Copyright, U. Fayyad © 2026
Slide 60: What AI Haves do right: Data as an Asset
What “AI Haves” do right ? - They use Data as an Asset Data Understanding Recognizing data's importance Events Capturing every bit of Data Context Understanding the context of data Right Action Driving effective outcomes with data Outcomes Measuring the results of actions 60 Copyright, U. Fayyad © 2
Slide 61: What AI Haves do right: Human in the Loop
What “AI Haves” do right ? - They implement Human in the Loop Recognize Human Role Human intervention is the most valuable asset Achieve AI Guidance Success ‘Human in the loop’ AI Integrate for Google, Open AI, Implementation Human Input Amazon, Tesla, and all companies that make AI work Corrections
Slide 62: What AI Have Nots do?
What AI “Have Nots” do ? They continue to dig through tons of structured datasets for manual xls and BI reports 62 Copyright, U. Fayyad © 2026
Slide 63: What can we do about Data?
What can we do about Data? Capitalize on Data as the Essential and Lasting Asset… Enable the capture, management, and retrieval of data that captures operations and interventions… Some actions a. AI Data Strategy: data capture, storage, and management b. Enable the capture data from operations syste
Slide 64: Secret 4: Making AI Work
SECRET 4: Making AI Work (from talk on Secrets of Making AI Work) 4 Capture ALL data from EVERY human intervention when, why, desired outcome, and context with clear permission and disclosure In most organization this data is never Captured Leakage of IP into Data Exhaust – a true waste of the most
Slide 65: Customer Interactions and Intent Architecture
Customer Interactions and Intent Architecture 65 65 Copyright, U. Fayyad © 2026
Slide 66: Customer's view, your view, and reality
Customer’s view, your view, and reality… How Your Organization example: a bank and its customers Views Customers? Day to Day Life Moments Buying a home Commute Small Business Savings Account Insurance Utility bills Getting married Policy Health Renewals Insurance Health Buying a car Life Auto Insura
Slide 67: Implications for Marketing?
Implications for Marketing? ? Understanding Intent drives Relevance Relevance drives the highest value: in advertising and in marketing - less interruptive, more engaging, value to all How do we do this while preserving privacy? How do we do this Responsibly? what are incentives for consumers to sha
Slide 68: A Fun Case Study in Marketing
From GenAI to JenAI? A Fun Case Study in Marketing… leads to deep issues in AI Safety… Same Tech for Deep Fakes An Interesting tool for marketing? Consider Virgin Cruises Marketing Campaign… 68 Copyright, U. Fayyad © 2026
Slide 69: From GenAI to JenAI?
From GenAI to JenAI? Same Tech for Deep Fakes An Interesting tool for marketing? umf-JenAI-virgin-voyages-video.mp4 Consider Virgin Cruises Marketing Campaign… users can craft custom video invites from JLo herself… ● Implemented by a startup: DeepLocal - before the chatGPT hype 69 Copyright, U. Fayy
Slide 70: Is JenAI really using Generative AI?
is JenAI really using Generative AI? Personalization at scale: ● Using generative voice: match user speech to JLo’s speech pattern and style based on training. ● Using generative video: match movement, pose, gestures to JLo’s avatar Results - invites from JLo generated videos: ● Sold 1000 tickets in
Slide 71: Amusing Applications?
Amusing Applications? DALL-E, Open AI Fiction is Easy to Recognize But reality gets much more complicated (e.g., fake news) Sophisticated mashup or plagiarism? 71 Copyright, U. Fayyad © 2026
Slide 72: But Reality Gets Much More Complicated
But Reality Gets Much More Complicated + Videos! 72 Copyright, U. Fayyad © 2026
Slide 73: Gen AI enables Rapid Bad Uses
Gen AI enables Rapid Bad Uses • Fraud • Impersonation • Social Engineering • Cybercrime EU AI Act • Human-likeness - June 2024 (without transparency) 73 Copyright, U. Fayyad © 2026
Slide 74: Responsible AI as an Area of Differentiation
Responsible AI as an Area of Differentiation in Marketing/Branding/Advertising Practice - How to practice AI responsibly? - How to Assess the Risks of Using AI? - An opportunity to lead - A strategic advantage - The more platforms & regulation tighten restrictions on targeting; the bigger the opport
Slide 75: On Generative AI in Marketing
On Generative AI in Marketing Panel at SxSW-2025 in Austin: “Enough with the Delving” AI will disrupt and transform Marketing, Branding, and CRM • Generating content will be accelerated dramatically • Creative will be produced faster, cheaper, more personalized • AI will change the way we do work in
Slide 76: AI in Advertising - Skechers
AI in Advertising - Skechers Skechers draw backlash for full-page ad in Vogue that reeks of AI. ‘You actually didn’t save any money because now I hate you’ Chloe Berger Sat, December 14, 2024 at 1:36 AM EST 4 min read ● Skechers recently came under fire for allegations of AI use in the December issu
Slide 77: Key Themes to Remember
Key Themes to Remember ● Working AI is all Machine Learning today ○ Data, Data, Data ● Smarter not larger models ○ What can we do to find alternatives to the ever larger LLMs ● Talent and upskilling the workforce is a necessity – Not Optional! ○ Experiential, Experiential, Experiential ○ We need to
Slide 78: Some Notes on the AI Opportunities in Marketing
Some Notes on the AI Opportunities in Marketing 78 Copyright, U. Fayyad © 2026
Slide 79: Changes and Confusion Bring Great Opportunities (1)
Changes and Confusion Bring Great Opportunities (1) 1. Shifting consumer search patterns ○ No one understands AI Search What happens in a world of: “there is only one answer to a query” ? ○ Intent from chatbot conversations ○ Signals from pseudonymized chats 2. Changing media ○ back to TV but now ad
Slide 80: Changes and Confusion Bring Great Opportunities (2)
Changes and Confusion Bring Great Opportunities (2) 3. The new content landscape ○ Content criticality ○ Retail Media for branding and for conversions 4. Mobile In-App advertising ○ Evolving to leverage new signals ○ Privacy and ID-less targeting is an opportunity not a blocker! 5. The Science of An
Slide 81: An old case study in cookie-less ID
An old case study in cookie-less ID ● in 2015 I was on board of eXelate (later acquired by Nielsen), and they had an existential crisis as an ad network starting 2014 ○ Cookies disappeared ○ Apple iOS8 started blocking cookies (which formed basis for later total blocking of ad IDs ● solution - ID vi
Slide 82: A recent case study in ID-less Targeting/Tracking
A recent case study in ID-less Targeting/Tracking ● Jan 2026 - LinkedIn worked with Verve Group to assess ID-less technology ● The results were powerful ○ Cost per Install (CPI): Reduced by 38%. ○ Cost per App Activation: Reduced by 39%. ○ Post-Install Tracking: Achieved over 90% event visibility po
Slide 83: Changes and Confusion Bring Great Opportunities (3)
Changes and Confusion Bring Great Opportunities (3) 6. Data is the Strategic Moat ○ Without data no AI will work ○ Zero-party and First-party data is the cleanest approach 7. User opt-in for data collection withstands all privacy policy ○ Google et al can get opt-in in exchange for value to user ○ H
Slide 84: Some Notes on the Evolution of GenAI Usage
Some Notes on the Evolution of GenAI Usage 84 Copyright, U. Fayyad © 2026
Slide 85: Major Developments over the Last Year (1)
Major Developments over the Last Year (1) 1. Models getting larger (not necessarily better): from LLMs to GLMs 2. The RAG Hack to compensate for expensive re-training 3. Agentic AI everywhere (especially workflow automation) 4. Computer programming is cracked (sort of) - Vibe coding and the SAS-poca
Slide 86: Major Developments over the Last Year (2)
Major Developments over the Last Year (2) Are larger models better? ● Why are these LLMs getting bigger? ○ GPT3 (175B) → GPT 3.5 (~ 0.5T) → GPT 4 (~ 1.5T) ○ Rumors: GPT4.5 is an order of magnitude larger. Around 12 trillion parameters total ○ Like GPT4, claim is GPT4.5 data snapshot is from 2023… WH
Slide 87: Consider the LLM Quandary
Consider the LLM Quandary ● You build one (very large) know-it-all model (LLM) ○ 90-100 languages ○ Training on every major field of science ● What happens when I update the model with new data - re- training (fine tuning) Example: Add ability to build financial statements in Japanese ○ E.g. German
Slide 88: Major Developments over the Last Year (2)
Major Developments over the Last Year (2) Larger LLM’s are not necessarily better Smaller LLMs (SLM) Better Together (SLM + Knowledge Graphs) Private LLMs on the rise - Narrow specialization of SLMs 88 Copyright, U. Fayyad © 2026
Slide 89: Takeaway Lessons
Takeaway Lessons Ask not: “What is the largest model we can afford?” Ask: “What is the smallest LLM we can get away with?” What is our angle to do a lot more with less? Answers from our experience… a. Prior knowledge: knowledge graphs, network models, other “synthesized” knowledge bases b. Enable co
Slide 90: Small Language Model (SLM)
Small Language Model (SLM) Ability to incorporate Ability to run on local Ability to Retrain Most corrections and the Model machines, mobile Imprtantly interventions by users devices, and the “edge” during the application 90 Copyright, U. Fayyad © 2026
Slide 91: Major Developments over the Last Year (3)
Major Developments over the Last Year (3) Robustness and mitigating LLM errors RAG (Retrieval Augmented Generation) Advanced RAG - pre- and post-retrieval processing Modular RAG - Hybrid Search, Recursive Retrieval and Querying, Step-back approach, Sub- queries, Hypothetical Document Embeddings Grap
Slide 92: 2026 AI Predictions
(from AI Prediction for 2026 document) 2026 AI Predictions 3 The build-up of unsustainable “Technical Debt” created by LLM prompts, vibe coding, and RAG pipelines. ○ 2026 will witness an inflection: the mess that has been building up as AI custom solutions are developed will be recognized as serious
Slide 93: Major Developments over the Last Year (4)
Major Developments over the Last Year (4) Many developments on technical fronts… LM Agents Student-teacher LLM models Synthetic data Research to understand what happens inside the “black box” 93 Copyright, U. Fayyad © 2026
Slide 94: What About Agentic AI?
What About Agentic AI? How is Agentic AI Different from Why all the hype? Generative AI? o Thousands of startups claiming “agentic AI” o Much venture capital and corporate funds available Prompt Response Agentic AI Text Text Memory Workflow Gen. AI Tools Agentic AI Focuses on Creating Focuses on Doi
Slide 95: So what are the options for agentic AI?
So what are the options for “agentic AI”? ● Use the frontier models ○ Ask LLMs to create and manage agents ○ Not efficient or scalable, not private ● Leverage specialized “narrow” SLMs as adaptive agents ○ RAG is a good start – but SLMs are a must ■ Micro-Rag steps guide “C-o-T” type reasoning ■ Spa
Slide 96: Case-Study - Cyber Security Hacking
Case-Study - Cyber Security Hacking TECHNOLOGY https://newatlas.com/technology/gpt4-autonomously-hack-zero-day-security-flaws/?utm_source=flipboard&utm_content=other GPT-4 autonomously hacks zero-day security flaws with 53% success rate By Joe Salas June 08, 2024 Planning Agent GPT-4 was able to exp
Slide 97: Mythos – Agentic AI for Cybersecurity
Mythos – Agentic AI for Cybersecurity ● Mythos sits atop Claude Opus 4.7 ○ Clause Opus 4.7 had strong cybersecurity capabilities but was deliberately limited with guard rails ○ Mythos is the specialized model atop ● Anthropic claims Mythos is too powerful to allow public access ○ Claimed to be espec
Slide 98: Agentic AI Fail - Anthropic Project Vend Experiment
Agentic AI Fail - Anthropic - Project Vend Experiment EXPERIMENT PREMISE Anthropic’s Agent (Claudius) was given responsibility over a vending machine Set up to sell the goods and generate a profit. It was give access to the resources needed to research and to restock the machine THE GOOD Sourcing Pr
Slide 99: Analogy: Human Agents vs AI Agents
Analogy: Human Agents vs AI Agents Human Agents AI Agents Both Managed And Governed by a Line Manager 99 Copyright, U. Fayyad © 2026 99
Slide 100: Agentic AI: Lessons Learnt
Agentic AI: Lessons Learnt Sandbox & Governance & Ethical Permissions Hygiene Oversight Production Protection Policy Definition Compliance Rules Limited Agent Scope Human in the Loop Realistic Pilot Phases Human Review Checkpoints Constrained Tasks Value Proof Gradual Ramp-Ups Successful Agentic AI
Slide 101: Secret 7: Nurture Talent and build AI Culture
(from talk on Secrets of Making AI Work) SECRET 7: in Making AI Work 7 Nurture Talent and build AI Culture Talent and CULTURE are critical, employee & executive education are a must Culture eats Strategy for Breakfast Establish a Data and AI organization with ongoing upskilling and strong talent man
Slide 102: What is our AI Talent Strategy?
What is our AI Talent Strategy? • Levels of Training: a. AI Literacy & Awareness we offer programs in all these b. AI business usage ● For b, c, d: we believe best approach c. AI technical usage is Experiential Education: Learn d. AI Technical in-house Expertise while applying the new technology to
Slide 103: Some Questions
Some Questions Should we do anything on AI/GenAI now or wait and see? What happens if we just observe and learn? Do we use public LLMs or should the team build private SLMs? Can we leverage Open Source/Freeware to start? How do I make Agentic AI practically work for me? Why is everyone saying larger
Slide 104: Some Words of Wisdom to Reflect On
Some Words of Wisdom to Reflect On - December 2025 “The dogmas of the quiet past, are inadequate to the stormy present. The occasion is piled high with difficulty, and we must rise — with the occasion. As our case is new, so we must think anew, and act anew. We must disenthrall ourselves, and then w
Slide 105: What is going to happen in 2026?
exported_qrcode_image.jpg What is going to happen in 2026? Copy of the 2026 AI predictions memo? My AI Predictions Made at start of 2026 https://drive.google.com/file/d/1CdGPIjX_OPhoUmxB7p- FW2tUVC_Kx7Bz/view?usp=sharing 105 Copyright, U. Fayyad © 2026
Slide 106: Main takeaway themes for this talk
Main takeaway themes for this talk 1. Data: The critical role of Data as the perennial missing component and data capture and availability always being an “afterthought” • Many new sources of data - wearables, nearables, health, IOT, etc… • multi-modalities are much easier to utilize now (beyond aud
Slide 107: Summary/Concluding Thoughts
Summary/Concluding Thoughts 1. AI is an enterprise imperative – challenging to make work – But is a big factor in competitiveness in the knowledge economy 2. HOWEVER: No Data ⇒ No working AI Human intervention a must Capture your IP: events, outcomes, context continuous correction of algorithmic err
Slide 108: Summary/Concluding Thoughts
Summary/Concluding Thoughts The biggest danger to AI today? The gap between how amazing the demo is and how disappointing the implementation results… 1. Generative AI offers a means for accelerating work, but not fully automating it 2. Generative AI can help reduce robotic, repetitive, and manually
Slide 109: Thank you! Any Questions?
Thank you! AI.northeastern.edu www.open-insights.com Any Questions? IEAI-NU UFayyad USAMA M. FAYYAD @Experiential_ai usama@open-insights.com @UsamaF Assistant: Rana@open-insights.com Institute for Experiential AI (email to request a copy of these slides) exported_qrcode_image.jpg Legends of Data & A
Slide 110: Usama Fayyad, Chairman & Founder, Open Insights
Usama Fayyad, Chairman & Founder, Open Insights at Northeastern University: SVP, AI & Data Strategy, Sr. Advisor to President, & Professor of CS Inaugural Director, Institute for Experiential AI (EAI) at Northeastern University Education Large Orgs Goal: Make AI and Data usable, useful, manageable -
Slide 111: Closed Loop Retail Media: The Compounding Outcome Engine
Closed Loop Retail Media: The Compounding Outcome Engine Raimund Bau, President Verve Retail Media
Slide 112: Retail Media has Overtaken Linear TV
RETAIL MEDIA Retail Media has Global Retail Media overtakes Global Linear TV Overtaken Linear TV Retail Media reaches 197 billion dollars in 2026, ahead of linear TV at 139 billion, and now makes up 16 percent of all advertising worldwide. It is the fastest-growing major advertising channel. Source:
Slide 113: Retail Media is Growing, but Three Gaps Hold it Back
CHALLENGE Retail Media is Growing, but Three Gaps Hold it Back Spend is online, the sale No cross-retailer The full funnel is not is in-store campaigns connected Most purchases happen in physical stores, but Every retailer markets only its own For most retail media campaigns and retail media network
Slide 114: Our Sofa-to-Store Approach
SHOPPER JOURNEY We are Addressing These Shortcomings with our Sofa-to-Store Approach Combining all media to achieve one comprehensive end-to-end journey with actual sales impact for the brands. We use AI for targeting and sensors. 114 114
Slide 115: We Took This Concept to Scale in Less Than 10 Months
SCALING We Took This Concept to Scale in Less Than 10 months. And Built Germany’s Largest Cross Retailer Network. >5,000 >17,000 >46 M 9% 200+ In-store Screens CPG, drugstores, Gross Contacts Store Sales Lift CPG Brands live pharmacies per week on scientifically as active in grocery stores, in Germa
Slide 116: The Size of the Network Dwarfs Existing Single Retailer Networks
NETWORK Stores Cross-retail Independent Pharmacy Cinema The Size of the Verve RM 17,000+ Yes Yes 9,000 830 Network Dwarfs REWE Connect 3,700+ No No No No Existing Single Retailer Networks. Lidl Connect 3,200 No No No No dm Media 2,000+ No No No No Kaufland Media 780 No No No No Source: Verve Retail
Slide 117: Our Add2Wallet Technology Enables Sofa-to-Store
ADD2WALLET Our Add2Wallet Technology Enables Us to Play Sofa-to-Store With or Without In-Store Screens 1 2 3 4 5 117
Slide 118: AI Turns Receipt and Redemption Data into Geospatial Targeting
DATA LOOP AI Turns Receipt and Redemption The outcome loop Data into Geospatial Targeting for the Upper Funnel Cross-retailer receipt data feeds an AI layer that reads category dynamics down to the postal-code level, showing exactly where a brand is gaining or losing share. Media spend then defends
Slide 119: We Use Receipt Data and Redemption Data to Understand Where to Target
GEO TARGETING We Use Receipt Data and Redemption Data from Stores to Understand Where to Target to Achieve the Most Impact for Brands 119
Slide 120: Across Categories We Deliver Sales Lifts for Major Brands
SALES LIFT Across Categories We Deliver Sales Lifts for Brand Lift Major Brands Soft drinks Spezi +184% With one flight, Spezi, a regional brand, Dishwasher care Finish +76% moved ahead of Mezzo Mix, a Coca-Cola Company brand, with a plus 184 percent sales lift. This exemplifies the impact of our Fr
Slide 121: Why This is Hard to Copy: Each Layer Builds the Next
USP LAYERS Why This is Hard to Copy: Each Layer Builds the Next 1 - Physical in-store presence 2 - Cross-retailer receipt data 3 - Proprietary AI models 4 - Independent uplift measurement 17,000-plus stores, 9,000 That presence produces a The data set trains AI models The loop closes with Difference
Slide 122: Quote: Building the largest retail media network in Germany
PREDICTION "In less than one year we built the largest retail media network in Germany. Now we are building the same in the United States, powered by receipt data, geospatial AI, in-store sensors, and sofa-to-store." Raimund Bau, CEO Verve Retail Media 122
Slide 123: From Signals to Outcomes: The New Rules of Predictive Targeting
From Signals to Outcomes: The New Rules of Predictive Targeting Mishel Alon, CBO and President of Verve For Advertisers
Slide 124: Intent is no longer just typed into a search box
CONSUMER INTENT Intent is no longer just typed into a search box. Earlier signals. Richer signals. More 50% 38% predictive signals. of search queries are using AI-powered search U.S consumers used generative AI for online shopping ChatGPT user How to block external noise Beyond silence: AI-powered A
Slide 125: The old targeting model is breaking
MARKET SHIFT The old Dependent on persistent identifiers and static audiences. Proprietary search LLM, and zero party intent data for direct activation. targeting model is breaking ID-based model 125
Slide 126: Capturing and activating on consumer intent is growing increasingly difficult
VERVE’S SOLUTION NYC to LAX flights Capturing and activating on Luxury shopping consumer intent is growing increasingly difficult for advertisers Without the right partner, data signals are just noise. Five star hotels 126
Slide 127: The truth? Most platforms are limited
1 2 7 The truth? Most platforms are limited 127
Slide 128: Capture high-intent shoppers
CPG EXAMPLE Capture high-intent shoppers Reach shoppers from first inspiration to final checkout – before they’ve committed to a grocery list or brand. LLM Chat Prompts “What are high-protein vegan snacks?” Semantic technology at work Client’s shopper Schema Online search behavior “Organic supermark
Slide 129: Advertisers aren't short on dashboards
INTELLIGENCE Advertisers What advertisers What they struggle Delay Intent = aren't short on can answer today to answer Wasted budget dashboards • What did we spend? • Where is demand forming? • Stale budget • What converted? • How does my brand show up • Over-targeted users in AI conversations? They
Slide 130: From signal to outcome, in one workflow
PRODUCT THESIS From signal to Read Understand outcome, the signals the opportunity in one workflow Search Polling, On Device LLM prompts Campaign performance Engagement POS & Retail data ATOM Activate Optimize the audience toward outcomes 130
Slide 131: The new rules of predictive targeting
SUMMARY The new rules of predictive targeting Intent starts here Activation must be immediate Search is no longer the only window The value is turning insights into action into consumer intent. quickly. Identity is changing AI changes the workflow The market is moving away from Campaigns become guid
Slide 132: Q&A Session
Vv Q&A Session 2
Slide 133: Closing Remarks
Vv Closing Remarks
Slide 134: Continued Focus on Our Mission: Let's Make Media Better
Continued Focus on Our Mission: Let's Make Media Better Verve’s Strategic Positioning The trusted alternative to walled gardens that delivers business outcomes at scale. Scaled Access to Vertically Integrated Depth of Targeting Corporate Mobile Inventory Ad-Intelligence Data & Data & Measuring Devel
Slide 135: CEO quote: Our strong momentum is set to continue
“Our strong momentum is set to continue. We are a disciplined growth company with a focus on value creation: anchored in organic investment, supported by stronger margins, AI-driven efficiencies, and first returns from our strategic investment cycle. We are doubling down on our strengths to accelera
Slide 136: Got Questions?
Got Questions? Ingo Middelmenne ingo.middelmenne@verve.com © 2026 Verve Group, Inc.