Rippletide

Your best people should not have to onboard every new AI agent. Your team has spent years learning which deadlines cannot slip, when an exception is worth making, and what a good decision looks like. Consider a replacement order. The system calls it “standard.” The customer's production line is down. An experienced operator knows why that order goes first and why faster shipping may be worth the cost. The valuable lesson isn't “always rush replacement orders.” It's knowing when the situation justifies it. Now switch models. Add another agent. Replace a software vendor. Why should the company have to learn that lesson again? If its experience lives only in one person's head, the agents keep asking. If it lives only inside one agent, replacing that agent means rebuilding the knowledge. We're building Rippletide around a different principle: the company should own its business judgment, and different agents should be able to use it. That includes knowing when an old decision no longer fits the facts. Upgrade the intelligence. Keep the experience. If you replaced your main AI tool tomorrow, what would your team have to teach all over again?

At Datadog Summit San Francisco today 👋 Observability has made huge progress in helping us understand what is happening. The next frontier is what happens after the signal. AI agents can increasingly investigate incidents, reason about possible remediation, and even take action. That is incredibly promising. But getting from a technically plausible action to the right business outcome requires something more: business validation, policies, context and clear authority over what an agent is allowed to do. That is exactly the space we are exploring at Rippletide: helping agents move from detecting and recommending to resolving, safely and within the rules of the business. Lots to learn today. And a very exciting space to build in. If you’re at the Summit, come say hi 👋 #DatadogSummit #AIagents #Observability #AIOps

I paused my AI assistant. I had become its project manager. I'm building an AI company. I still had to do it. The assistant could write, plan, create visuals, and keep producing. But I was still correcting the same things, checking whether work was actually finished, and reminding it which decisions we had already made. I had delegated the output. I had kept the management job. We count the seconds AI saves on a task. We rarely count the attention it takes to keep the work on track. Another correction. Another check. Another “we already agreed on this.” For a founder, that isn't a small cost. It's time taken from customers, hiring, and decisions nobody else can make. I don't need a more convincing acknowledgment. I need the next action to reflect what we decided. This hasn't made me less ambitious about agents. It has changed what impresses me. The productivity metric I care about isn't how much my AI produces. It's how much work I no longer have to manage. What's one task you genuinely stopped supervising after giving it to AI?

We’re hiring a Chief Product Officer in San Francisco. 🚀 This is a high-impact, company-building role for an exceptional product leader who can turn deep technology into category-defining product. You’ll work directly with the founders, own product end-to-end, and help shape how Rippletide scales. We’re looking for someone sharp, technical, customer-obsessed, and comfortable making hard calls. Know someone exceptional? Send them our way. Details in the comments.

What should stop an AI agent from taking the wrong action? A sandbox protects the environment. A decision layer protects the decision. As Blaxel put it: “A sandbox lets an agent act. A decision layer helps it act correctly.” We shared how Blaxel + Rippletide work together to combine both. Read our take here and click the link in comment to read their article ⬇️

We’re growing the Rippletide team in the US 🇺🇸 We’re hiring in San Francisco across sales, solutions, marketing, partnerships, and operations. We’re still a small team, so these are high-ownership roles. You’ll work closely with the founders, spend time with customers, help shape the product, and have a real impact on how we build the US business. Open roles: • Founding Enterprise Account Executive • Founding Solutions Engineer • Marketing Manager • Business Development Manager, Partnerships • Operations Manager We’re looking for people who enjoy early-stage environments, move quickly, and want to help build the Action Runtime for AI agents in production. Interested, or know someone who could be a great fit? Explore the open roles and apply with the link in comment!

Legally Blonde for me. Legally bound for the agents. At #Dreamforce to see Reese Witherspoon, so the pink Rippletide cap felt mandatory. But there’s a serious point behind the joke. AI agents are getting more autonomy every week. The question is no longer whether they can act. It’s whether they can act without breaking the rules. That’s what we’re building at Rippletide: enforceable policies for AI agents. More autonomy. Same rules.

Fin reports 76% average resolution. Here’s our offer: Fin + Rippletide. 90%+ autonomous resolution. Or the project is on us. The toughest support cases require judgment. Should this customer get a refund? Does their contract change the answer? What have we already promised them? Getting those decisions right is what lets an agent finish the job. Rippletide brings business judgment to your agents, checking proposed actions against your policies and the customer’s situation before they happen. You keep Fin. We add Rippletide. We put our fee on the outcome. We agree on the scope and success criteria upfront. We deliver the project. If we don’t reach 90% autonomous resolution, you don’t pay us. Your real cases. Your business rules. Our money on the line. Running Fin? Bring me the cases it still hands back to your team. I’m at Dreamforce. DM me and let’s meet.

Your AI can build an app. Getting it to use the right logo? Apparently, that’s the hard part. “Use this logo.” “Yes, you can update the CRM.” “We literally covered this yesterday.” Working with AI shouldn’t feel like onboarding the same intern every morning. We’re building Rippletide to make your corrections stick. So your agents learn how you work, follow your preferences, and ask when your judgment actually matters. Less repeating yourself. More getting things done. We’re in San Francisco. If you’ve ever told your AI “I already told you,” let’s grab coffee. ☕

Rippletide did not start with an approval workflow. It started with a harder question: Where does a company’s decision actually live? Not in one policy. Not in one database row. Not in the model. Take a supplier exception. The answer may depend on the supplier, the contract, the amount, the current approval, a temporary exception, and the person who owns the risk today. Each fact can be retrieved. The decision still has to be assembled. For two years, Rippletide worked on the underlying problem: how to connect the action, current context, business conditions, evidence, and exceptions into one explicit decision. Technically, that meant a hypergraph decision database, automatic ontologies, and neuro-symbolic reasoning. In plain English: connect the right business facts to the exact action before it happens. That research now powers the Action Runtime. The agent proposes what to do. Rippletide makes the company’s judgment explicit for this case, now. The moat is not another model watching the first model. It is making operating judgment computable at the moment of action.