ICG

Anthropic and OpenAI just bet $11.5 billion. Not on AI models. Both companies launched new services to put AI inside other businesses. The model race is no longer the only race. Deployment is. What happened. The news broke on 4 May. Anthropic and OpenAI each announced a new company built for one job: getting AI to actually work in customer businesses. Anthropic's new company is worth $1.5 billion. Founding partners are Blackstone, Goldman Sachs, and Hellman & Friedman. Other backers include Apollo, General Atlantic, GIC, Leonard Green, and Sequoia Capital. OpenAI's, called The Development Company, is worth $10 billion. Its 19 investors include TPG, Brookfield Asset Management, Advent, and Bain Capital. The plan from both is the same. Send engineers into hospitals, factories, and mid-sized businesses to make AI useful in everyday work. This is the forward-deployed engineer model Palantir built its business on. On 11 May, OpenAI bought consulting firm Tomoro to scale faster. Why this matters for any business adopting AI. The companies that build the best AI models just admitted the model is not the hard part anymore. Using it well is. Three things change. First, traditional consultancies have new competitors. Accenture, Deloitte, and others were the default partners for big AI rollouts. Now Anthropic and OpenAI are sending their own teams. Second, the risk of doing it badly is clearer. If the AI builders say deployment is hard, the do-it-yourself approach needs proper programme governance, not just a vendor contract. Third, the AI is coming through the back door. The investors funding these ventures own thousands of mid-sized businesses across financial services, healthcare, and manufacturing. The AI will land in operations before most leadership teams have agreed on an operating model for it. This is the kind of shift I unpack each week. Keep up with best practices in Enterprise Transformation through my newsletter, The Transformation Constant: https://lnkd.in/d_mtyWzr What this means for the year ahead. The list of who to talk to, and what to ask, just changed. A few questions to hold for any AI vendor or internal team. ↳ Who will be in the room for the rollout, and what is their track record in a business like ours? ↳ What does success look like in week one, week four, and month six? ↳ If the AI gives a wrong or costly answer, who notices, and how fast? ↳ What is the plan when the first version does not work as expected? ↳ Does this AI fit how we already work, or expect us to change the operating model around it? The model race lasted three years. The deployment race will last longer. In financial services and asset management, the winners won't have the best AI. They'll have the right operating model and programme governance. 🔔 Follow Justin R. for the signals shaping Enterprise Transformation.

AI pilots stall. Not at launch. At the scale-or-kill decision. Across financial services, programme governance is the checkpoint that breaks for asset managers and regulated firms in the same way. Weeks of work. Clean results. A SteerCo that agrees it's worth pursuing. Then nobody runs the five tests that determine whether the pilot can actually reach production. The question isn't "is it going well?" It's five binary checks. One for each of the most common failure modes. 1️⃣ Use-case drift: has the core problem changed since kickoff? If yes, the pilot is no longer testing what it was built to test. 2️⃣ No agreed success metrics at week 4: if the team can't name two measures that define success, the pilot can't be fairly stopped. 3️⃣ Data not production-grade: if it's running on clean samples and would fail on live data at volume, that's a sequencing problem. 4️⃣ Governance conditions unmet: data classification, access, or compliance gaps that production requires haven't been addressed. 5️⃣ No named production owner: without one identified before the pilot ends, the workflow stays in permanent advanced pilot status. When three or more are active, the 1/2/3 rule applies. One trigger: document and watch. Two: brief the sponsor, define a resolution path. Three or more: stop and reset the brief. 🚀 Subscribe to The Transformation Constant — my newsletter — https://lnkd.in/d_mtyWzr The scale-or-kill decision is a programme governance call. Scaling a pilot that hasn't met its conditions doesn't speed delivery. It speeds risk. The firms that stop pilots cleanly don't have a better idea waiting. They have a decision framework they actually use. 🔔 Follow Justin R. for the signals shaping Enterprise Transformation.

Sharing here an interesting idea from MIT students. What do you think about it?

There are roughly a hundred AI readiness checklists in circulation right now. 1. What AI agents exist across your environment? 2. Who is using them and for what purpose? 3. What systems and data sets do they have permission to touch? 4. How are you monitoring what they do? Those aren't mine. They're Microsoft's April 2026 guidance to enterprise boards. If your committee can't answer all four — the pilot is not ready for production. The firms that can answer them sit in Deloitte's top 20% for agentic governance. The firms that can't are making up the answers in front of their boards after the incident. Issue 03 of The Transformation Constant — out this week — turns the four questions into a 10-question readiness gate. Four layers. Binary answers. Scoreable in under 20 minutes by the team that will operate the AI in production. If you cannot answer 8 of the 10, the pilot holds. One question. Costed in full. Every two weeks. Link below.

Most programmes fund delivery. Not the governance overhead that keeps it moving. This is where financial services programme governance consistently breaks. The strategy is usually fine. We spend on the workstreams. On the tools. On the vendor contracts. The governance structure that holds all of it together gets treated as overhead. So it gets absorbed. Budgeted into delivery capacity rather than protected as a separate line. The moment a milestone slips, it's the first thing quietly cut. The programme is now delivering on momentum. Not structure. There are four components that prevent this. They're not complicated. They're consistently installed in the wrong order, or not at all before Week 1. The sequence matters: 1️⃣ Governance Cadence: sponsor authority pre-delegated, thresholds defined before pressure arrives. When the escalation happens, the path is already written. 2️⃣ Decision Velocity: average decision cycle under 10 days, tracked as a trend. Slow decisions aren't a pace problem. They're an unmapped authority problem. 3️⃣ Dependency Map: named, owned, updated weekly. Not built for gate reviews. Built for the 72 hours between when a dependency breaks and when it becomes a crisis. 4️⃣ Change Saturation Index: surfaces fatigue before the programme calls it stakeholder resistance. By the time it's labelled resistance, it's already a signal nobody acted on. 🚀 I break down the full framework in The Transformation Constant — https://lnkd.in/d_mtyWzr Seventy percent of transformation programmes miss their stated objectives. McKinsey, BCG, and Gartner converge on this. The four-component stack isn't the whole answer. But it's a closeable gap — in days, not quarters. Audit your programme against those four components. Most teams find the first gap at layer two: decision velocity. That's almost always the first sign the foundation wasn't properly signed off before Week 1. Follow Justin R. for more Enterprise Transformation insights

Showcasing Musical Talent Across Our ICG Community Thank you to everyone who joined us for the recent music recital at Tower Bridge Care Home and helped make it such a memorable event. It was wonderful to see colleagues come together to support and dedicate pieces to the Elderly. The musical talents of everyone created a relaxed and uplifting atmosphere for the resident and ICG colleagues alike. The performances were a real joy and a great reminder of the importance of taking time out t and connection beyond our day‑to‑day roles. Thank you to Tower Bridge Care home for welcoming us, Melisa Lytham from ICG who helped organise the event with Oana Gadalean and our wonderful speaker Darren Jewell-NeuroResilience Coach 🧠 for highlighting the importance of music and how anything is possible if you try. #LifeAtICG