Anthony Lye

New release from Packt Publishing… “Building Business-Ready Generative #AI Systems — Build Human-Centered Generative AI Systems with Context-Aware Agents, Memory, and LLMs for the Enterprise” See it at https://amzn.to/3Jdcio5 𝓚𝓮𝔂 𝓕𝓮𝓪𝓽𝓾𝓻𝓮𝓼: 🔵Build an adaptive, context-aware AI controller with advanced memory strategies 🟢Enhance GenAISys with multi-domain, multimodal reasoning capabilities and Chain of Thought (CoT) 🟠Seamlessly integrate cutting-edge OpenAI and DeepSeek models as you see fit

14

3 Comments

In a recent Financial Times piece, written by Thoma Bravo's Holden Spaht, he calls out the risks many AI start-ups are taking, which really resonated with me. He points out that narrow, single-purpose tools may deliver flashes of efficiency, but they struggle to connect back to the broader business systems that actually drive sustained value. At RealPage, Inc., we’ve taken that insight to heart with Lumina — embedded, agentic AI that can reason and act across workflows, built into the core of our platform rather than bolted on as a point solution. Our AI agents don’t work in isolation; they collaborate with human teams, sharing work, context, and accountability directly inside the systems operators already use. That same deep integration allows resident-facing AI to coordinate seamlessly with property operations and back-office teams, ensuring actions taken at the front line are reflected across the business. True advantage comes when AI augments the full operating system of a business, helps operators manage risk, and connects outcomes across functions. A standalone tool may optimize a slice of the workflow; a platform-optimized end-to-end solution elevates decision-making, governance, and accountability at scale. That’s why investment in trust and risk management — from data integrity to regulatory and policy compliance and operational safeguards — isn’t an afterthought for us. It’s a differentiator. AI only delivers value when customers can rely on it safely and predictably, across their entire portfolio. Integrating intelligence, not fragmenting it, is how we’re helping operators unlock real, durable performance.

200

6 Comments

LLMs can ace the bar exam—yet still spin fairy-tales when you ask why. Models first solve then invent the reason. A 2024 causal-mediation audit of 12 frontier models showed that in most runs the answer was unaffected by the chain-of-thought the model just wrote. In other words, it solved the problem first and invented a rationale later. Relying on that single rationale is no different from accepting a memo with the signature page missing. High-stakes workflows need stronger guardrails. Here’s the engineering pattern that is quietly becoming the norm across audit-sensitive teams: ⸻ 1 | Retrieval-first, Reasoning-second Ground the model before it thinks. Pull every plausible candidate record from your ERP with deterministic SQL and turn the model’s job into ranking a finite list. Retrieval-Augmented Generation (RAG) has already cut hallucinations by anchoring answers to real documents and citations. 2 | „Four-eyes” architecture Split generation and verification. Agent #1 proposes; Agent #2 interrogates the reasoning, cross-checks retrieved evidence, and either signs off or sends it back. The four-eyes principle has kept banks and insurers honest for decades. Recent research shows multi-agent setups without an explicit verifier still fail up to 60 % of the time—so design that checker in from day one. 3 | Full-stack logging & replay Persist every prompt, retrieval call, ranking decision, and veto. Come audit season you can replay the entire pipeline, not merely wave at a black-box answer. ⸻ Take-home for CIOs/CFOs: Explanations that aren’t causally linked to outcomes are compliance liabilities. The safe path is retrieval-anchored inputs, dual control of reasoning, and traceability that an external auditor can step through click-by-click. (We follow the same pattern inside Hypatos’ agentic automation stack, but the blueprint is open for anyone who values audit-grade AI.)

11

AI is indeed powerful, and the potential for efficiency gains is significant. MCPs represent a valuable opportunity for GTM workflows. However, the recent buzz around Claude Code has led to some misleading narratives in the GTM space. Many posts claim, “I replaced my entire sales team with an AI agent” or “MCPs made my SDRs obsolete,” but these statements are not entirely accurate or wise. Here’s a clearer picture of the current landscape: - Perplexity suggests we should abandon MCPs and revert to APIs. - Manus (Meta) is advancing with its version of Cowork, which appears to be more developed. - OpenAI is reportedly shifting its focus toward enterprise solutions, with the upcoming GPT 5.4 model being its best yet. - Additionally, credible sources indicate that Claude is facing challenges with compute costs at current usage levels. As someone who builds with Claude daily, I appreciate the product and have developed MCP servers, skills, and workflows around it. Yet, I also recognize the rapid shifts in this landscape. You can create remarkable solutions now, and I encourage you to do so. However, it's essential to pause, study, and observe the developments before making drastic changes to existing systems. The pace of innovation is swift; stay focused and avoid being overly disruptive.

12

1 Comment

OpenAI may be setting a trend with its announcement about consulting practice with a minimum price tag of $10M per engagement.   The reason OpenAI is doing this is revenue – consulting margins (40-60%) are much higher than margins for APIs and cloud services. OpenAI needs to grow to justify such a high valuation as they are currently losing $5 billion annually (despite $10 billion in revenue).🚀   But, jumping from an R&D environment into consulting is easier said than done... OpenAI engineers are extremely good at R&D but they are not business process experts. Customer implementations and working on specific customer business processes require very different skills.   Organizations don’t just want powerful AI tools but partners who can help them navigate the complex process of digital transformation.   ➡ The focus must be on revenue, not on AI tool metrics.  ➡ AI should be used to augment teams, so people can work on revenue. ➡ AI should be invisible to the end user. ➡ Consultants should communicate in the currency of business outcomes.   In our AI consulting practice, we have been focused on diligently measuring every outcome in the business currency of our customers. We know that getting the most value of any technology depends on how well it is implemented, not on vanity metrics used to compare various LLMs. To get the most value out of AI, a company must commit to becoming AI first. Anything less means leaving money on the table.   #AI #AIconsulting #OpenAI #AIstrategy

47

8 Comments

Launching a High-Impact GTM Strategy for a Next-Gen Cloud & HPC Provider Earlier this quarter, we helped a fast-growing hyperscale cloud company aiming to enter the global enterprise market. They needed clarity on three things: • Which industries offer the strongest adoption potential • How to differentiate against major hyperscalers • What truly drives enterprise buying decisions in cloud & HPC Our Approach 1️⃣ Market Intelligence & Competitive Benchmarking We analyzed hyperscale trends, growth segments (AI/ML, biotech, fintech), pricing benchmarks, and competitor gaps across security, performance, and support. 2️⃣ Deep Buyer Insight & Product Fit Validation Through enterprise buyer research, we uncovered priorities, blockers, and workload needs, clarifying where the client could win with confidence. 3️⃣ End-to-End GTM Strategy We built a complete GTM blueprint: positioning, messaging, ICP prioritization, launch channels, and partner ecosystem recommendations. 4️⃣ Investor & Leadership Enablement We refined the pitch narrative with market opportunity sizing, differentiation strategy, and a data-backed GTM motion. Outcomes - Clear roadmap for enterprise expansion - Sharper differentiation against leading hyperscalers - Faster enterprise lead generation through refined ICPs - Stronger investor confidence - Reduced trial-and-error in GTM execution Helping companies launch with clarity, not assumptions, at the core of what we do at Strategy Here. If you're building or refining your GTM strategy and want an approach rooted in real insights, happy to connect. --- For more case studies and insights, visit: www.strategyhere.com #gotomarket #gtmstrategy #marketintelligence #competitiveresearch #cloudcomputing #hpc #enterprisetech #b2bsaas #strategyconsulting #customerinsights #techstrategy #businessgrowth

1

TechCrunch and Greenfield Partners just released their top 60 AI Disruptors list. Which means it’s time for a data driven analysis of their selections 😁 **Link to the full in the bottom of the post.** Most recurring investors? Impressive lineup - Lightspeed, Andreessen Horowitz, Kleiner Perkins, Index Ventures, Accel. Each with at least 5 portcos. Company sizes by headcount: 1️⃣ Small (<20) - 30%, 18 companies.   2️⃣ Medium (21 - 75) - 55%, 33 companies. 3️⃣ Large (76+) - 15%, 9 companies. Primary focus is on early-growth companies around A-Series or with large seeds pocketed in recent months. Fastest growing companies YoY in employees hiring for each category: 🥇 Cogent Security (Small). 🥇 Profound (Medium). 🥇 Nexthop AI (Large). All growing by triple or four digits in headcount %. 💰 Total funding to-date average across the list is $58M.  💰 The latest round average is $37M. Since the sample is pretty wide spread geographically, I would argue these companies can provide good benchmarks if sliced and diced to their respective sizes. 💵 Exodigo leads the list with the most funding to-date, totaling $225M+ raised, followed by Torq with $190M, both Israeli companies. Geographic origination: 🇺🇸 50%, 30 companies - U.S. 🇮🇱 26.6%, 16 companies - Israel. 🇪🇺 23.4, 14 companies - Europe. Hiring trends in the last 6-months: 8.3%, 5 companies actually either reduced their headcount or stayed stale. 23.4%, 14 portrayed ‘modest’ (depending on their size) headcount growth of 4%-20%, impressive. 68.3%, 41 are skyrocketing their workforce by 21%-400%, truly incredible! At least 6 companies haven’t disclosed new funding in the last 18-months meaning there should be opportunities available for investors in the coming quarter. 21 companies are led by at least one repeat founder, a strong over representation. Want to know who? Perform your own analysis with Dealigence? Get the full list here ➡ https://lnkd.in/dE-Za9Yy Full release link in the first comment.

59

2 Comments

(-) AI and hyperscale trends are playing an outsized role in influencing enterprise/on-prem OEM input costs. While 2018 saw a significant cloud spending cycle (the strongest Y/Y growth until 2024), the intensity of AI investments from the world's leading hyperscalers and neoclouds is very different from the last memory cycle, which is accelerating the tightness in memory markets that well surpasses 2016-2018 (Exhibit 45). HBM is in significant demand because of AI, but this has pulled supply away from DDR5 (DRAM), while shortages of HDDs are causing significant spikes in cloud eSSD demand. Said differently, investments in AI from the world's leading hyperscalers are influencing memory pricing for PC, server, smartphones, where demand trends are more muted on an absolute basis and vs. 2016-2018, when on-prem enterprise hardware spending was stronger than it is today.

1

Most supplier collaboration tools work. Until they hit real scale. Portals, viewers, and review platforms reduce friction. But they still assume collaboration happens after the model has been exported. As supplier networks grow — especially across multi-tier global programs common in aerospace and defense — coordination layers multiply. More checkpoints. More oversight. More manual alignment. That’s not collaboration maturing. That’s complexity leaking out of the system of record. #Manufacturing #PLM #SupplierCollaboration #DigitalThread

10

SaaS gross margins are under pressure, and it's bad news for valuations. In a recent survey of SaaS CFOs by Cloud Capital, 89% reported that rising cloud costs negatively impacted gross margins. In some cases, new AI development efforts have increased cloud costs. For others, traditional workloads are costing more because of AI-driven pricing pressure. Either way, declines in gross margin are putting significant pressure on SaaS valuations. Let's unpack exactly why and how much. The first step in understanding the role of gross margins in SaaS valuations is to forget SaaS valuations before 2021. Before 2021, the space was new, growing quickly, and investors assumed companies could transition to profitability in the future. This was rational at the time, but when growth slowed, valuations naturally shifted from multiples of revenue to multiples of profit. During the revenue-multiple era, if your company had a "normal" gross margin, a few points one way or the other was not material in its valuation. Now, however, because gross margins directly impact profits, they also directly impact valuations, and in a substantial way. Let's look at current public SaaS data to demonstrate. The median gross profit margin of public SaaS companies is currently 78%, and the median operating profit is 18%. Hypothetically, if gross margins were to drop by five percentage points for a billion-dollar company, operating profit would fall from $180 million to $130 million. The decline in profits would be destructive to the company's valuation for three reasons. First is the simple math with a constant PE multiple of 30 (close to the current median). $180 million x 30 = $5.4 billion $130 million x 30 = $4.0 billion. So, with a constant multiple, the valuation decreases by 25% given a 5-point decline in gross margins. Second, since investors focus more on future earnings than on current earnings, and this company's earnings are shrinking, its P/E ratio is likely to drop dramatically below 30. Third, COGS matter more to investors than operating expenses. Given the same level of operating profit, high-gross-margin companies are worth more. Why? COGS are variable expenses that are hard to cut without immediate negative consequences. New methods and efficiencies must be found to reduce COGS, and that's more difficult than cutting operating expenses to drive profits. This example is a gross simplification, but it's still important to understand the basic math SaaS companies face. If a business is making big bets on AI, it might be able to explain away the gross margin decline, but only if it can back up its assertion with data isolating core gross margins from development costs. (more on this in future posts) As a core driver of profits, gross margins weigh heavily on valuations, and they are headed in the wrong direction. Maintaining or improving gross margins in this environment needs to be on every CFO's radar. Article and survey results link in the comments.

18

8 Comments

AI capital is flowing at unprecedented scale, but the real differentiator is whether it converts into revenue. Too often, capital stalls in the execution layer. RevOps discipline, ICP clarity, and GTM acceleration are what unlock growth velocity. At Celeris we bridge that gap through 𝗖𝗮𝗽𝗶𝘁𝗮𝗹 𝗔𝗰𝘁𝗶𝘃𝗮𝘁𝗶𝗼𝗻 and 𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗔𝗰𝗰𝗲𝗹𝗲𝗿𝗮𝘁𝗶𝗼𝗻, ensuring funding translates into ARR growth instead of stalling in workflows. Capital is available. Execution is the bottleneck. Leadership and discipline make the difference.

8

1 Comment

The time for merely experimenting with AI in your sales organization is over. A recent McKinsey survey revealed that only 21% of organizations have enabled generative AI for B2B buying and selling, with just 22% having piloted specific use cases. But that's rapidly changing in 2025 as companies scale their AI initiatives to increase sales productivity. As sales leaders, we need a pragmatic framework for AI integration. Here's what we are seeing working with our customers: 1. Focus on ROI, Not Just Implementation Organizations must take a strong look at the true ROI of their AI initiatives. For AI capabilities to add real value, sales organizations need to ensure features like opportunity scoring and deal recommendations are relevant and rooted in data. Additionally, there is a TON of experimentation going on and periodically organizations need to step back and see which initiatives are really moving the needle. Start by identifying specific metrics you'll track to measure AI's impact, whether that's win rates, deal velocity, or forecasting accuracy. 2. Transform Your Data Foundation First Data quality isn't just important—it's everything. The most successful B2B players are leveraging AI to analyze extensive data to suggest white space markets, identify new verticals, and prioritize opportunities where they have competitive advantage. McKinsey & Company Before adding fancy AI tools, ensure your CRM data is clean, complete, and connected. This is a big one! 3. For reps, focus on Augmentation. For Revops, focus on Automation The best AI implementations don't replace your salespeople—they elevate them. While AI offers incredible enhancements in sales, it's not about replacing human roles but augmenting them. Some skills, like empathy and authenticity, remain irreplaceable in B2B sales interactions. Revops is the place where there is a tremendous potential for automation. 99% of organizations are just scratching the surface of what is possible here. 4. Implement AI Governance Early By 2025, 35% of chief revenue officers will resource a centralized "GenAI Operations" team as part of their go-to-market organization. This isn't just about compliance—it's about creating guardrails that encourage innovation while ensuring consistent, ethical use across your team. 5. Start with High-Impact Use Cases Begin with applications that directly impact revenue: - Hyper-targeted account selection and prioritization - Deal risk identification and intervention - Next-best-action recommendations - Personalized content generation for outreach - Conversation intelligence for call coaching The winners in 2025 won't be those who avoided AI, nor those who blindly embraced every shiny new tool. The winners will be leaders who strategically integrated AI into their sales motion while keeping the human element at the center. What AI applications are you prioritizing in your sales organization this year? #salesleadership #AI

11

2 Comments

🚨 Caution with off the shelf LLMs in business critical environments Salesforce’s new benchmark study on large language model agents is a needed reality check. More than 50 percent failure rates and frequent hallucinations were reported, even from top tier models like GPT-4 and Claude 3 Opus. 🔍 Highlights: • Most LLM agents could not reliably complete complex tasks • Hallucinations and dead ends were common • Performance varied significantly based on architecture and context This is why we focus on building AI in controlled and customized environments, especially in regulated industries such as healthcare and logistics. The risks of automating flawed processes or relying on hallucinated data are too high to ignore. Benchmarks provide useful insight, but real world success depends on structure, oversight, and alignment with operational workflows. #AI #LLM #DigitalTransformation #SupplyChain #Healthcare #Governance #Salesforce https://lnkd.in/eX_RD8AK

7

Mergers, acquisitions, and divestitures are among the most complex and high-risk activities any organization can undertake. TBM can be a powerful tool to architect strategic and measurable value.

26

Excellent insights from Orlando Bravo. At Hyena Enterprises Global (H•E•G), we’re seeing the same shift firsthand - AI isn’t replacing software; it’s transforming how it’s conceived, integrated, and scaled. Through our F4ORGE architecture, large language models are redefining development and decision-making pipelines, embedding intelligence directly into the fabric of enterprise systems. This convergence marks not the end of software, but the dawn of a new, adaptive era - one where AI and software evolve as one. #AI #ArtificialIntelligence #SaaS #Innovation #F4ORGE #HyenaEnterprisesGlobal #Defense #DefenseIndustrialBase #EWD #Workforce #WorkforceDevelopment #FutureWorkforce #FutureOfWork

1

When the model makers become channel builders, pay attention. Anthropic just committed $100 million to a partner ecosystem. Certifications. Co-investment. Dedicated engineers embedded in partner deals. Sales playbooks shared openly with the firms helping enterprises deploy Claude. That infrastructure exists because frontier models cannot reach the enterprise alone. The distance between a model demo and a production deployment is enormous, filled by people who understand compliance, change management, integration, and industry context. Partners do that work. Anthropic just backed that reality with capital — and every frontier AI model company will follow. The model wars are becoming ecosystem wars. And ecosystem wars are won by the teams that can actually operate across them. That is the part most people are not talking about yet. Every new partner program that launches — every certification, every co-sell motion, every marketplace listing — adds complexity that lands on the desks of partner managers who are already stretched thin. The infrastructure to manage that complexity has to operate across company boundaries, not just inside your own walls. That is a fundamentally different problem than anything the partnership category has faced before. What comes next is a world where every major AI company has a mature partner ecosystem, where multi-model, multi-partner deals are the norm, and where orchestrating across that complexity is a genuine competitive advantage. We are at the very beginning of that world. At WorkSpan, helping partner teams navigate what comes next — and turn ecosystem complexity into revenue — is exactly what we exist to do. The partners who move now will shape how it gets built. https://lnkd.in/gX3d_-kd

80

6 Comments