I think the managed security services model is changing. A few things pushed the market here. The first one is AI. It changed the tools, but it also changed the processes.
One thing AI made easier is building security tools. And that pushed the market in two directions.
First, we have more security vendors than ever, because building is cheaper and easier. In the past, your half-baked security idea would end up as an open source project. Now everyone tries to monetize it. Like anything in life, this has good and bad sides. In the past, when you needed a security tool, you usually settled for the one that covered around 70-80% of your requirements. Now you have so much choice that you can get close to 90%.
Second, many teams try to build it themselves and match 100% of their requirements. The problem with building it yourself: AI made it easier to build stuff, but you still need to come up with the architecture and the inner workings. Plus MAINTENANCE. Right now, tools you built yourself are not easy to maintain. Yes, you can create agents that do stuff for you, but in my experience, when something goes wrong, they break things even faster. So yes, building is easy. Maintaining and scaling is hard.
Product Updates
Security changed. Services didn't. Until now.
Security teams shouldn’t have to choose between the speed and scale of AI and the depth and judgment of human expertise.
Daylight combines an AI-native security platform with expert security engineers to detect, investigate, and respond to threats up to 10X faster than traditional managed security providers.
Built around each customer’s environment, Daylight adapts to existing systems, workflows, and priorities. Choose fully managed, co-managed, or forward-deployed security operations to scale your team’s capabilities without adding more tools, headcount, or operational overhead.
MSSPs and MDRs face a hard problem: serving dozens of customers, each with a different tech stack.
Orryx AI solves it with a multi-tenant, technology-agnostic platform for AI-powered incident orchestration and management.
Built on 15+ years of SOC experience, Orryx combines automation, knowledge-as-code and advanced reasoning to cut the cost per investigation while making every investigation faster and more reliable, which is critical for any AI SOC.
And it delivers consistent results across multiple LLMs. That puts control back in your hands: no model lock-in, and a private intelligence layer that protects your proprietary knowledge and data from competitors and frontier LLMs
There is a good debate on this. Frank Wang argues in The changing buy vs. build calculus for security that security needs to build again, because AI raised the bar for what is worth buying. Ross Haleliuk pushed back in There’s only one kind of tool security teams should be building with AI. His take: the cost of building is going down, but the cost of ownership is not, so most teams should build the glue and leave the heavy security tools to vendors.
Both agree on one thing: if you have a strong engineering team, you should build. I’m not so sure. Even a strong engineering team might not have the know-how to scale the tool and keep the token and cloud bill under control. Building it is one skill. Running it at scale without burning money is another.
If a product works but burns tokens (AI spend) and compute (cloud spend) like crazy, it is not really a solution. You will spend a ton of time optimizing for cost.
Funny story. In most orgs I worked at, every in-house tool my teams built was great until someone from IT came screaming about the cloud bill. And because another department was paying, nobody optimized until someone screamed.
I’m not saying all internal tools end up like this. I just want to show what can go wrong.
From “we monitor for you” to “we build for you”

OK, that was a long intro. What I want to discuss today is why MDR is changing from monitoring for you to building for you. Some call it the Palantir model, and they are not wrong. But I think it’s a slightly modified version. The best results come when you provide both forward deployed engineers and analysts.
Let’s give it a name first, because why not. I will call it Managed Security Operations , or won’t be surprised to see something like Managed Preemptive Security.
So what are the core offerings of this model?
The platform
You need a platform to build and run things on. So you sell the platform your security operations run on. It needs to be an integrated system of:
SIEM (or decoupled SIEM capabilities: federated search and maybe a detection engine)
Context, the company brain: your CMDB (asset and identity inventory) and your business intelligence
Case management
Agent builder
Automation builder (a combo of the newer agentic SOAR capabilities)
What I think is different from the old MDR model is that this covers more than SOC and triage. Vulnerability management, DLP, IAM, AppSec, GRC. Any security function, managed.
The services
You have two service deliverables:
Engineers who build agentic solutions and workflows for you. You can take it over from there, or let them keep running it next to your team.
Traditional MDR-style services: building and improving detections, doing triage and investigations, and running response and remediation where needed.
Ross says the glue needs internal context that no vendor can fully copy. He is right. That’s why the engineers have to be forward deployed. They sit with your team and learn how your environment works before they build anything. And Frank has a good way to describe the value: you are buying a slice of specialized talent you would not hire full time.
One more thing I think is key: doing QA on the agentic workflows, so you get a continuous loop. Even MDRs advertise this continuous loop, but I rarely saw it implemented. Usually it went like this: you get a suspicious or malicious alert, it gets escalated to the customer, and only when the customer complains about an alert flood does someone fix it.
Where this goes next
I think this can grow even further. Imagine something like Scribe Optimize (using this as example have no commercial relationship to promote or endorse the product) running constantly and telling you which other processes can be automated or improved. Then you move into proactive security. And after that, of course, fully managed preemptive security.
Check out our SecOps Market Landscape tracker and evaluation frameworks



Discussion about this post