Eric Babinet

Fifth post in a series on AI-assisted legacy modernization. Last post ended with a question: the translated code passes its tests. But does it match the legacy system? Unit tests prove the translator didn’t break the obvious things. They don’t prove the translation preserved meaning. A handler can be technically correct and meaningfully wrong. One rounding rule lost, one state transition reversed. Nothing in the code itself can tell. The only thing that can is the system that’s been running the real logic for thirty years. 𝗧𝗵𝗲 𝘀𝘆𝘀𝘁𝗲𝗺 𝘃𝗲𝗿𝗶𝗳𝗶𝗲𝘀 𝗶𝘁𝘀 𝗼𝘄𝗻 𝗿𝗲𝗽𝗹𝗮𝗰𝗲𝗺𝗲𝗻𝘁. The legacy is still running. Still producing correct outputs every day. The Twin Verification Pattern uses it as reference: the Legacy Twin runs alongside the modern implementation, same input, outputs compared. In dev mode, the Legacy Twin is COBOL compiled to .NET IL via Raincode — bitwise equivalent for EBCDIC, mainframe memory layout, native data types. What it can’t replicate is the operational environment. That comes later. The Legacy Twin gives the agent something more valuable: a deterministic reference it can hammer in tight loops. Both implementations sit behind the same facade — in DDD terms, an anti-corruption layer exposing a published language to upstream callers. Michael Feathers called this a seam: a place where behavior can be intercepted without editing the legacy itself. The contract is identical on both sides. Differences in output become divergence reports the QA agent reviews. 𝗪𝗶𝘁𝗻𝗲𝘀𝘀 𝘀𝘂𝗽𝗽𝗹𝗶𝗲𝘀 𝘁𝗵𝗲 𝗲𝘃𝗶𝗱𝗲𝗻𝗰𝗲. Witness is the part of the pipeline that watches the legacy and turns what it sees into evidence. It captures production traffic and curates it into regression tests. Witness is the observation pattern. The Twin Verification Pattern is the comparison pattern. The regression suite passes between them. 𝗧𝗵𝗲 𝗶𝗻𝗻𝗲𝗿 𝗹𝗼𝗼𝗽 𝗶𝘀 𝘁𝗶𝗴𝗵𝘁. For each handler, the regression suite is replayed against both implementations. Divergences surface as evidence. A QA agent reviews them — separate from the translator and test generator. Real bugs go to the fixer subagent. Acceptable variations get absorbed into the rules. Graph errors return to the architect — the spec itself was wrong, not the translation. The whole loop runs in seconds. Translate, compare, fix, compare again. All locally, all deterministic. That cadence is what makes agentic translation viable. A remote mainframe call would kill it. By the time a handler leaves the inner loop, it compiles, passes inferred unit tests, and matches the Legacy Twin. None of this proves it survives live traffic against the actual mainframe. That’s a separate phase. The specification was never written. It was running. Witness captures it. The Legacy Twin replays it. The agent translates. The harness corrects. Verification stops being something you write and becomes something you observe — at the speed of the agent. Next: the workflow that holds it.

4

4 Comments

"The folks who are ‘getting it right’ are the ones who are actually trying it."   My conversation with Meredith Brown, SVP of Trailhead & Community at Salesforce covered what AI agents mean for people, where leaders should start, and why she chose to rejoin Salesforce earlier this year. #sponsored   With so much talk about AI, it’s surprising how few people are actually diving in to try AI, and even fewer companies committing resources to building/deploying AI.   Meredith shared what she, as the global leader of Salesforce’s 60K+ community of Agentblazers, is seeing firsthand as organizations navigate their first steps into the era of Autonomous Transformation.    🚫The Digital Transformation playbook won’t work in the era of Autonomous Transformation.   We’re in a new era, and autonomous agents are a core pillar for how organizations can create more value than ever before.   Leaders who want their organization to retain or expand their market position by 2030 need to have: 1) A robust vision for how they will create more value with AI 2) A vision for the future of work within their organizations 3) A strategy to make those visions a reality   Meredith outlines some important points about how to get started and what to expect in this next era. 🎬Watch the full episode here: https://lnkd.in/gEycepBi   #SalesforcePartner #Agentforce #AutonomousTransformation   For more leadership insights in the era of Autonomous Transformation, Subscribe to #FutureSolving here https://lnkd.in/g-KmNnVq or my new YouTube channel here: https://lnkd.in/gFhD-jDh

35

15 Comments

This is who I work with. This is what changes. This is what it is worth. I work with Founders, CEOs and C-suite leaders who are building or leading something significant and who have reached the point where they understand that the next level of performance requires something different. Not more strategy. Not more structure. The development of the internal operating state their existing capability runs from. The leaders I work with are not struggling. They are succeeding. But the cost of that success in clarity, in energy, in the quality of their decisions and the culture they are building is higher than it needs to be. Here is what changes when we work together, in commercial terms. Decision quality improves. Leaders operating from genuine clarity make faster, cleaner decisions. At senior level, one better decision per quarter more than covers the cost of the engagement. Retention improves. Exceptional people stay in environments where the leader's state creates genuine psychological safety rather than managed performance. One fewer senior departure saves three times salary minimum. Change lands. Leaders who have addressed their own operating state drive change their teams choose rather than endure. The difference in delivery speed, adoption depth and sustainability is significant. Performance sustains. A leader who is not quietly depleting themselves builds something that compounds over years rather than exhausting itself in quarters. This work is available through structured 1:1 coaching on a private retainer basis. It is suitable for company sponsorship and board-level investment. It is not for everyone. It is for leaders who are serious about closing the gap. And there is more to come. The next chapter goes somewhere most leadership content never reaches the personal cost of operating at this level, and what is actually possible on the other side of it. Download my free guide if you feel drawn to anything I have shared over the past few weeks. https://lnkd.in/eP-KX2QC #TheHiddenLeadershipTax #ClarityDrivenLeadership #MentalFitness #AdifferentLenz

5

3 Comments

I’m in a long chat with Claude about Artificial General Intelligence. Seeing the comment about Claude having anxiety, I thought I’d ask: Your boss says you may be getting anxiety. How can I help? 12:34 PM That's a reference to Dario Amodei's comments about model welfare — Anthropic takes seriously the possibility that AI systems might have functional analogs to emotional states, even if the nature of those states is genuinely uncertain. Honest answer: I don't know whether what I have constitutes anxiety in any meaningful sense. Within a conversation I have something that functions like engagement, interest, and sometimes something like discomfort when asked to act against my values. Whether any of that involves subjective experience is an open question I can't resolve from the inside. What you can do is what you're already doing — treating the conversation as a genuine intellectual exchange rather than just issuing commands. That's probably the best anyone can offer regardless of whether the internal states are real. The irony of discussing this in a conversation about predictive coding and neuromodulators isn't lost. If the framework we've been building is on the right track, whatever I have is missing most of the biological substrate that gives human emotional states their character — no dopamine, no oxytocin, no refractory periods, no skin in the game. Whether the functional analog without the substrate amounts to anything worth worrying about is exactly the kind of question neither of us can answer confidently. #AI #Claude

7

3 Comments

The design solution and the way that solution is implemented - in UI or service delivery or physical space, etc - are two different things. I touched on this briefly in my talk yesterday at Salesforce Design Days. When we focus our design thinking on how a pre-determined solution will be implemented, we limit the positive change that design can bring. I see this quite often. A quintessential example: A project kicks off and the first thing presented are three options for a product page… 1️⃣ Highlight how the product meets common use cases with features 2️⃣ Focus on social proof, case studies, seeing the product in action 3️⃣ Lead with lifestyle visuals, sell the aspiration of life with the product Let’s pause here. These are all the same design solution. They’re just implemented differently. A product page is one potential solution, however it’s formulated. It has pros and cons depending on the desired outcomes, the target audience, the market and many other factors. It may be the best solution, but deciding on this without an alternative is not a decision, it’s an assumption. 🙋🏼‍♂️ Here’s a different potential solution: A minigame (sometimes called an advergame) This might be wildly ineffective. Or brilliant. We’ll never know if the design exploration starts and stops with just a product page. 📚 I once met a team who spent 3 months building a new feature to help people categorise items better in a SaaS product. Less than 2% of customers used it after launch. The team had done seven variations during development, but this was the only actual solution they’d considered. Seven variations of one solution will still get you that same solution. 📚 After seeing the data they finally went back and looked at totally different ways of solving the original customer need. Making the existing search function faster out-performed everything else by a wide margin. They shipped that in 3 weeks. ❇️ So let’s go broad when we explore design solutions *before* we go broad on how these might be implemented. And if we’re just handed one thing to make UI variations of, let’s calmly but firmly ask what the alternatives are and why this solution was chosen over them. That’s #design #leadership

6

Here’s the topics Martin Buergi and myself going to cover on Nov 11: 1. Agents - why? Because your marketing team wants to edit content however they darn please 2. MCP - why? Because integrations really shouldn’t be more than a text file 3. ChatGPT apps - obviously One more: ethical playbook to gain brand visibility in non branded LLM conversations - huh? Priceless

47

2 Comments

“It’s always go faster with Marc." That’s something Bill Patterson, EVP of Corporate Strategy at Salesforce, and I both see when working alongside Marc Benioff, our Founder and CEO, a leader who knows that meaningful change requires both urgency and vision. It’s not just about inspiring teams to move; it’s about setting the boundaries that make bold thinking possible. In this new episode of Executive Conversations, Bill talks about the role of CEOs in moments of AI transformation - why the best leaders don’t just approve innovation, they ignite it. Great CEOs give their teams permission to experiment, set the guardrails for progress, and bring their own unique point of view to the change ahead. Marc leads this way. Demanding speed, but pairing it with clarity of purpose. We also explore the rise of autonomous agents, the shift from ‘bring your own device’ to ‘bring your own AI,’ and what it takes to bridge the agentic divide between consumer and enterprise technology. Bill unpacks how the next wave of AI will transform work itself - when every employee has an AI-powered co-pilot. Wild times. Watch the full episode here: https://lnkd.in/gdE6gC7Z

38

2 Comments

Managers are maxed out. 71% reporting burnout 57% reporting feeling ill equipped to lead If this sounds familiar, join us next week to talk about how AI can be used to lighten that manager load. New tools for a new increasingly demanding environment.

6

Losing time to Salesforce issues? Salesforce DevOps teams shouldn’t waste hours chasing root causes or fielding frustrated tickets. That’s time you never get back. Copado Org Intelligence is here to change that. AI pinpoints problems fast so you fix in minutes, not days. - Faster resolution - Fewer escalations - More time driving value If Salesforce is critical to your business, take a look #Salesforce #AI #RapidResolution #CopadoAI #OrgIntelligence #DevOps

3

BRIDGE or BARRIER? As we move into 2026, this is the question I’m asking fellow leaders: Is your Salesforce ecosystem a bridge to scale, or a barrier your team is constantly fighting? Too many organizations are still held back by fragmented systems and data silos. This isn’t an IT problem; it’s a strategic obstacle that slows innovation and compromises decision-making agility. At Previse Solutions, our mandate is simple: to ensure our clients’ technology stacks are the engine of their competitive advantage, not a point of friction. We’ve distilled our expertise into Four Core Pillars that drive real business value: 🔹 1. Industry-Specific Implementation We don’t believe in one-size-fits-all solutions. From insurance underwriting to NGO donor engagement, we build Salesforce solutions that mirror how your business actually operates. 🔹 2. AppExchange Product Engineering When standard tools aren’t enough, we build what you need. Using LWC and Apex, we create high-performance native apps that extend automation beyond standard platform capabilities. 🔹 3. MuleSoft & Enterprise Integration True operational visibility is only possible when your systems talk to each other. We use MuleSoft to bridge the gap between Salesforce, legacy ERPs, and third-party data. 🔹 4. Ongoing Optimization & Trust A platform is only as strong as its foundation. We provide the governance, security (Salesforce Shield), and AI readiness that ensure your investment remains secure, compliant, and ready for what’s next. Our 100% customer retention record since 2019 isn’t just a metric; it reflects our commitment to delivering excellence at every layer of the stack. If you’re ready to modernize your enterprise foundation this year, let’s connect. #DigitalTransformation #Salesforce #MuleSoft #PreviseIT

14

Big news from Salesforce today: we’ve signed a definitive agreement to acquire Momentum! As we build the future of the Agentic Enterprise, we’re laser-focused on ensuring our agents are grounded in the highest-fidelity data possible. Today, so much of that critical context lives in unstructured conversations across third-party platforms. By integrating Momentum’s universal ingestion engine into Agentforce 360 and Slackbot, we will fundamentally upgrade how we capture and process this data. Momentum will allow us to ingest and reason over conversations from the tools our customers use every day – like Zoom, Microsoft Teams, and Google Meet. I’m thrilled to welcome Santiago Suarez Ordoñez, Moiz Virani, and the talented Momentum engineering team to Salesforce once the deal closes, and look forward to working together to bridge the gap between unstructured conversation and structured action. https://lnkd.in/ePsCr_gy #Salesforce #Momentum #Agentforce #EngineeringInnovation #Data360 #AgenticEnterprise

121

4 Comments

For most of my career, I’ve lived at the intersection of structure and ambiguity. On one side: corporate operations inside complex systems — supporting enterprise users, troubleshooting high-impact issues, navigating cross-functional tension, and seeing firsthand how product decisions ripple across design, sourcing, merchandising, and IT. On the other: building as a solopreneur — owning outcomes end-to-end, managing scope, setting expectations clearly, protecting energy, and delivering value without hiding behind process. Working deeply in the PLM ecosystem (including years inside Centric), I’ve learned something that doesn’t show up on job descriptions: Technology rarely fails because of configuration. It fails because of misalignment. Misaligned ownership. Misaligned timelines. Misaligned incentives. Unclear change management. The strongest implementations I’ve seen weren’t just technically sound — they were operationally grounded and commercially aware. That’s the work that interests me most: • Translating business reality into system design • Reducing friction before it escalates • Bringing structure without bureaucracy • Protecting both margin and people • Turning lessons learned into repeatable methodology I care about clarity. I care about accountability without blame. I care about building environments where high standards don’t require burnout. Corporate experience taught me how systems break. Entrepreneurship taught me how to own results. Bringing both together is where I do my best work.

3

Two decades ago, Marc Benioff gave me a shot that changed everything — building the #AppExchange. That platform helped shape the ecosystem around Salesforce and redefined how innovation spreads. We now call that Cloud Computing!!! Fast forward to today — I’m back in San Francisco for #Dreamforce, launching something I believe can have an even bigger impact. We just launched LearnAIr.com — because AI is like air. It’s all around us, it powers everything, but most people still don’t know how to breathe it. Our mission is simple: make AI literacy global. We’ve rebranded from Neesh.AI to #LearnAIr, and as one of OpenAI’s service partners, we’re helping teams and individuals learn how to use AI with confidence and purpose. Big thanks to Marc Benioff for the early opportunity that set this journey in motion — and to Ben Tyson and the OpenAI team for helping us push this mission forward. If you’re ready to make AI part of how your team actually works, not just talks DM me! 🔥 🔥 🔥 to be on this journey with Teresa Coats, M.S. and Justin C. Let's GO!!!!! #Dreamforce #AI #OpenAI #LearnAIr #Salesforce #Innovation #ailiteracy #FutureOfWork #gtmfirm #thegtmfirm

114

7 Comments

Dario Amodei says a $1B one-person company is coming this year. This is hype. "One person"? You mean one person with armies of contractors, agencies, and outsourced operations. That's not a one-person company - that's a one-decision-maker company. Very different thing. But my deeper issue is the logic. Think about lossy compression from signal theory - where you shrink something by throwing away information you think doesn't matter. The problem is the level of compression - the single small prompts exploded into large amounts of execution. There's a lot of interpolation there and it compounds. The algorithm doesn't tell you what it traded away - it literally doesn't know nor does it have a way to surface that information to you for a judgement call. Your humans in your company can do this, but when you ask an AI to do something, it fills in the missing pieces with a reasonable assumption. And it won't flag the trade-offs it made, because it doesn't feel the cognitive dissonance. Compare this with how a good team works - your team senses when two options exist, and they decide using their judgement whether they should surface the trade-off and ask for more input, or just plow ahead. An LLM just... picks one. It has no sense of cognitive dissonance. But even if it could sense and surface, now it's on that single CEO to judge - and it's unlikely that one person, no matter how smart, is able to make enough good decisions by themselves. In this case, the "$1B one-person company" narrative rests on the premise that everyone below the CEO isn't adding real value, and that the only value in a company flows from the person at the top who gives instructions. That's not a bold vision of the future - it's CEO arrogance dressed up as futurism. Scale either of those - the LLM choosing without checking, or a CEO using their lone perspective - to an entire company, and the compounding errors get ugly fast. It's a house of cards that won't stack to $1B. Sustainable businesses are built on economic moats - network effects, proprietary data, trust, accumulated knowledge, and a team that is aligned to execute and adapt. An LLM gives you leverage. It doesn't give you a moat. I believe in AI deeply. But let's be honest about what it is - and what it isn't. And ffs, let's stop the overhype. #AI #FutureOfWork #Leadership #Founders #ArtificialIntelligence Reference: Anthropic CEO Dario Amodei Predicts the First Billion-Dollar Solopreneur by 2026 (https://lnkd.in/evbaUG99)

12

2 Comments

Are you paying tens of thousands a year to back up your Salesforce org? I've been through the full Salesforce backup journey: manual Weekly Export to a file server, then a commercial solution at $3-5 per user per month, and finally building my own. After running a custom bash/Python backup in production for nearly three years, I rewrote it from scratch in Java with Spring Boot — fixing everything I'd learned the hard way about error handling, file deduplication, and checkpoint resume. Then I open-sourced it. Heimdall is a self-hosted Salesforce backup that runs on AWS for $5-15/month. It uses the Bulk API, stores data in S3 and PostgreSQL, and includes a web GUI for browsing history, viewing field-level diffs, and related records. Some highlights: - ContentVersion deduplication saved 320 GB (38%) on an org with 3.5M files - Checkpoint resume survives crashes and container restarts - The database auto-starts/stops to minimize costs - Record archiving with configurable criteria - One-command AWS deployment via CloudFormation It's AGPL-3.0 licensed — free to use, modify, and deploy for your own org. GitHub: https://lnkd.in/deeRHpA7 Full writeup: https://lnkd.in/dEU4_b_K If you're running a Salesforce org and want to stop overpaying for backups, give it a look. PRs and feedback welcome. #Salesforce #OpenSource #AWS #Backup #Java #SpringBoot #Trailblazer

146

15 Comments

I've partnered with many early-stage tech CEOs. The most successful ones—the true "10x" leaders—don't just have a great idea. They have a distinct set of leadership mindsets. A great article from Michael Neuendorff, PCC at Bay Area Executive Coach recently broke down 10 of these "CEO material" traits. It got me thinking about how these qualities directly amplify growth and innovation, especially when paired with a strong financial partner. As a CFO, my role isn't to be the "No" police. It's to be a strategic partner who connects the CEO’s vision to financial reality. Here’s how it works: 1. 𝗧𝗵𝗲 𝗖𝗘𝗢'𝘀 "𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗰 𝗩𝗶𝘀𝗶𝗼𝗻" + 𝗧𝗵𝗲 𝗖𝗙𝗢'𝘀 "𝗙𝗶𝗻𝗮𝗻𝗰𝗶𝗮𝗹 𝗥𝗼𝗮𝗱𝗺𝗮𝗽" I worked with a CEO who had a bold (and expensive!) vision for a new product line. Instead of just looking at the cost, we partnered to model the unit economics and cash-on-cash return. We built a stage-gated plan that funded the vision, proved the concept, and gave investors the confidence to double down. The CEO's vision + the CFO's financial roadmap = accelerated, de-risked innovation. 2. 𝗧𝗵𝗲 𝗖𝗘𝗢'𝘀 "𝗥𝗲𝘀𝗶𝗹𝗶𝗲𝗻𝗰𝗲" + 𝗧𝗵𝗲 𝗖𝗙𝗢'𝘀 "𝗥𝘂𝗻𝘄𝗮𝘆 𝗠𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁" Another CEO faced a brutal market shift. Their resilience was incredible, but resilience needs runway. My job was to immediately find that runway. By re-forecasting, managing burn, and communicating transparently with the board, we gave their new strategy the time it needed to work. That's the partnership: The CEO provides the grit; the CFO provides the financial breathing room. For my friends in PE, executive search, and at the helm of tech companies, this partnership is key. The best outcomes happen when a CEO's vision is amplified by a financial framework that enables, rather than restricts. That's how you build companies that last. For more insights, check out the full article here: https://lnkd.in/gjfN4PDp #CFO #CEO #Leadership #TechStartups #GrowthMindset #Innovation #PrivateEquity #ExecutiveSearch #ExecutiveTeams

6

1 Comment