SoS Logo

From Cybersecurity Platforms to Systems

A subtle but significant revolution is lost in the noise about platformization: AI is moving us beyond platforms to systems.

Nov 16
Concepts
from-cybersecurity-platforms-to-systems

If you want to start a riot in a room full of cybersecurity people, just yell out “PLATFORMIZATION!”

Platforms have become one of the cybersecurity industry’s most volatile topics, thanks in part to overuse of the term by marketing teams and a general tendency among practitioners to resist anything centralized and corporate-sounding.

Part of the controversy is semantics. We need a better definition of “cybersecurity platform.” That’s one objective for this article.

Another part is scope. It’s a lot more accurate to call something a cybersecurity platform when we’re specific about which security domains it covers.

The idea of a single cybersecurity platform sounds ridiculous because it is. I’ve said it before, and I’ll say it again: there will never be one cybersecurity platform that does everything.¹ Cybersecurity covers too many domains, problems, and workflows for that.

Just because we don’t have one single platform doesn’t mean we can’t have any platforms, though.

We already have single-domain platforms in identity, endpoint, network, cloud, security operations, email security, and more. They may not cover an entire domain, but enough to clear the bar of being a platform.

Multi-domain platforms arguably exist, too. At minimum, we’re well on our way to them. I’m not here to debate that today, though.

Lost in all the platformization noise is a subtle but significant revolution: AI is moving us beyond platforms to systems.

A system is more than just software. It connects people, agents, and tools to pursue objectives, coordinate action, and learn from results. Outcomes, not outputs.

To understand the shift from platforms to systems, we have to look at how we got to this point in the first place. Let’s talk about that next.

How we got here: products, suites, and platforms

Products

Historically, new security problems have been met by companies with focused products to solve them.² That was true when cybersecurity was a nascent industry in the 90s, and it’s still true today.

Most companies start by building one product around a clear job, finding customers, and expanding from there. That’s a sensible way to start, and often a good way to solve a problem that larger incumbents haven’t addressed yet.

The cybersecurity industry's early product categories are good examples of this pattern:

It’s a similar story with other market leaders today:

We could keep going with examples, but you get the idea. These companies (and many others) started with a focused product in a single category before expanding and eventually growing into large companies.

These were simpler times. The main question for security leaders was, roughly, “how do I stop each new threat?” Identify a threat, buy a product, repeat. We didn’t give much consideration yet to things like cost savings, visibility, speed, or many other issues that are front and center today.

Products are relatively narrow solutions by definition. Buying best-of-breed security products is an approach we hear a lot in the industry, both then and now.

A security team might want the depth of a focused product, especially when a problem is novel or urgent.

Sometimes that’s fine, but treating too many problems as “novel” is how we end up with stats like the average enterprise having 80+ security products.⁴

With every new product comes the work of running it alongside everything else: integrating data, managing policies and configurations, teaching security teams another workflow, and maintaining everything over time.

The products might be excellent, but the operating model can still be exhausting. A best-of-breed approach can work, but it gets unmanageable quickly in a domain with a litany of problems that change all the time.

Enterprises with large security teams can often absorb this level of operational overhead. A lot of other organizations can’t.

The implication isn’t just the number of security products the team has to carry. It’s just as bad, and maybe even worse, to have advanced security products that are underutilized.

The commercial model gets exhausting, too. Forty security vendors means 40 procurement processes to run, 40 bake-offs, 40 contracts to sign, 40 salespeople to manage, 40 numbers to call for support, 40 admin dashboards to open, and so on.

This is fun the first five or ten times you do it. After that, it’s a massive headache.

In cybersecurity, the commercial part was easier to fix than the technical and operational part. The incentives aligned, too. Companies wanted to get bigger, and security leaders wanted to consolidate vendors. That’s how we got to suites.

Suites

Suites were the next step in the cybersecurity industry's commercial maturity. A suite is a group of products sold together, regardless of whether they share a technical foundation or not.

This helped to address some of the commercial angst security leaders had about procuring and operating cybersecurity products. On the industry side, suites drove our first generation of larger scale cybersecurity companies.

By the early 2000s, assembling broader security suites had become an established strategy. Cybersecurity companies were bringing several products under one roof by building them, buying them, or both.

In 2005, Forrester produced a report evaluating client security suites as a market category. By 2008, McAfee reported $1.6 billion in revenue across corporate and consumer operations. Symantec reported $5.9 billion (including adjacent areas like storage, backup, consumer, and services).

Calling this milestone “commercial maturity” was intentional on my part, though. Suites were more of a business and commercial construct than a technical one.

Buying a suite of security products might make the procurement process easier, but then you’re right back to the technical and operational challenges that individual products and best-of-breed architectures have.

In fairness, we were in a different era of technology in the 2000s. Desktop operating systems and client-server architectures still reigned supreme. The Web 2.0 era was just getting started around the middle of the decade. Cloud computing didn’t start until the latter half. I had a BlackBerry, my first smartphone.

I bring up this trip down memory lane only because the frontier of technology itself wasn’t ready for cybersecurity platforms yet. It wasn’t realistic for a cybersecurity company to build a broad platform in that era even if they wanted to — at least not in the way we think about platforms today.

There were early signs, though. Early efforts gave us a few clues about what a broader platform could become:

  • McAfee's ePolicy Orchestrator had common policy management and administration for parts of its portfolio.
  • Symantec had several product-level integrations, including endpoint protection and compliance in one agent.
  • Trend Micro described Smart Protection Network in 2008 as joint client-cloud infrastructure intended to support multiple offerings.
  • Check Point's Software Blades offered modular security applications on a common gateway foundation.

Again, these examples (and other ones you could pick from the same era) were early. I wouldn’t say or imply they were failures — just iterations on what would later become real platforms once the market and underlying technologies caught up. And catch up they did.

Platforms

The 2010s brought cybersecurity’s platform era. These were formative years, both for cybersecurity platforms and the industry in general.

Our companies grew up. There were 31 cybersecurity companies who went public during the decade, including many of today’s most well-known companies.⁵

Our technology grew up, too. This era was the start of a larger shift from broad product portfolios to more integrated platforms. Many of our companies had grown enough to fund the R&D to build and buy their way to broader offerings.

Advancements like software frameworks, better coding tools, cloud infrastructure, SaaS, and more made it a lot more practical for everyone, including cybersecurity companies, to build with reusable components and common foundations.This doesn’t mean everything is nice and tidy, though. We’re still navigating the transition from products to suites to platforms today.

Where this left us: a landscape of products, suites, quasi-platforms, and a few real ones

As much as I love the promise of platforms and the progress we’ve made to get there, I’ve seen first-hand how far we still have to go.

A decade of consulting in large enterprises will do that to you. I’m still optimistic, but I’m also appropriately grounded and pragmatic about where we’re at today.

All of this feels intense at times because these eras are a recurring cycle, not a linear progression. We didn’t move cleanly from products to suites to platforms. They’re more like stages, and different categories of cybersecurity occupy different stages at the same time.

We’re always going to have products. New generations of focused products keep showing up because our problems keep changing. There are a thousand financing rounds per year in cybersecurity.³ We invest a lot of capital in building new companies.

New threats, technologies, and circumstances create opportunities while other categories consolidate as they mature. Some focused companies become suites or platforms, but they don’t have to.

Suites aren’t going anywhere, either. We still have them today, and sometimes simpler procurement and potential cost savings are enough.⁶

You could also make a case that flex models are a modern take on suites. A customer can buy a subset of modules from a larger company and change what it consumes over time. Flexible commercial terms and a wider catalog of products can coexist with a platform.

We’ve built and bought our way to cybersecurity platforms, but multiple products under one logo doesn’t guarantee their data, policies, interfaces, and code work together. Someone still has to do the integration. That someone had usually been the customer, or the systems integrator the customer hired. Sometimes that was me.

I suspect this gap in expectations is a big part of the negativity around platforms. If a security leader’s vision of a cybersecurity platform is “everything just works together” (or some version along that spectrum), of course it’s going to be maddening if they get more sharp edges than expected.

So what, exactly, should we expect a cybersecurity platform to provide?

What a platform provides

There is something I’ve always been baffled about from the ongoing debate around platforms in cybersecurity: we already have reliable definitions of a technology platform.

A lot of it is academic research, but it’s well known and produced by some of the top institutions in the world. There isn’t much room for debate.

In the MIT version, Gawer and Cusumano define a product platform as shared assets from which a company can develop related products. The Harvard Business School version (Baldwin and Woodard) adds an architectural view: a relatively stable core, components that can vary, and defined interfaces between them.

That’s the academic version, which (as academics do) is intentionally abstract. But the research is a well-grounded and necessary starting point given all the conjecture about platforms in cybersecurity.

Here’s a less academic rendition: A platform is more than a collection of products. It gives different capabilities a foundation they can reuse. Reusability is the distinction between a collection of products and a platform.

In a cybersecurity platform, that foundation might include shared data models, services that build context or apply policy, and common APIs. One module in the platform can use the same context or service as another without rebuilding it.

Single-domain platforms put this model to work within identity, endpoint, or another major domain. Again, I don’t think it’s a stretch at all to say this has already been done many times over.

Multi-domain platforms have to carry meaningful parts of the foundation across domains. That’s hard to do, but an overly rigid definition can make the job seem more impossible than it really is.

I don’t think it’s fair to say a multi-domain platform has to be 100% integrated across every user interface, data model, policy, API, or other components. An all-or-nothing scope that requires or expects every single cybersecurity product category to be included isn’t realistic, either. I’ve talked with a lot of cybersecurity CEOs, and none of them have ever told me that’s the goal.

In more realistic and practical terms, I would look for shared or consistent representations of the data, a coherent user interface, integrated APIs for core functions, and controls and policy work together when it makes sense. The quality and usefulness of the platform’s integrations and overall architecture matters more than full uniformity for its own sake.

For me, the idea of a platform is almost easier to understand as an anti-definition. I learned the hard way what a platform isn’t. I saw the cost of poor integration myself. When components don’t work together by default, connecting them takes a ton of time, and the result still falls short of customer expectations more than I’d like to admit.

A big indicator of a platform is that it moves more of the integration work from customers (and their SIs) upstream to the product foundation. One R&D investment can benefit many customers instead of every customer needing to make one-off investments on their own.

A good foundation doesn’t eliminate implementation work, but it can give security teams time back and make the products easier to use together. Call this the customer experience test of platform value.

This might be controversial, but third-party technology partnerships and integrations are fair game for platforms, too. I don’t believe the boundary of a platform has to stop with the company who built it. This is the cybersecurity version of the open vs. closed platform debate.

In cybersecurity, the amount of variance in customer environments often forces their hand into wanting choice and interoperability. It’s hard to adopt a first party platform across multiple security domains all at once.

Third party integrations allow platforms to meet customers where they are and still produce outcomes without owning the full stack. This also gives customers flexibility to standardize over time (accounting for renewal cycles, budgeting, politics, and myriad other factors required to make changes in large organizations).

Native platform capabilities are ideal, and they’re still my first preference when possible. A first-class integration with a third party is a close second, though. Especially when it behaves like a core part of the system even though another company built it.

Third-party platform participation also has an economic side. That’s part of the MIT definition of a platform I mentioned earlier. Summarizing that part of the research from Gawer and Cusumao: An external partner may build a capability the platform owner would not build itself. Customers gain another useful option, and the partner gains a new way to reach customers.

The first-party technology platform is only part of the story. Companies like Sophos, CrowdStrike, and others have massive ecosystems that materially contribute to the overall value of their platforms.

We are still in the middle of building broader platforms across cybersecurity domains. The work may never end, and that’s okay. Platforms are a hard problem to solve.

At the same time, AI is both pushing and enabling our industry toward systems that can coordinate more of the work people do today. If platforms have features and benefits, systems have objectives and outcomes.

Thinking in terms of systems, objectives, and outcomes feels a bit foreign while we’re still building and adjusting to the idea of cybersecurity platforms. It’s not as far of a leap as you might think, though. Let me explain.

What a system adds

At a conceptual level, most security work is a combination of people, process, and technology used to share information and take action. Security objectives guide the work.⁷

That already is a system.

Donella Meadows wrote the canonical book on this. It’s called Thinking in Systems. You’ve probably heard of it, maybe even read it. Meadows’ definition of a system has three essential parts: elements, interconnections, and a function or purpose.

In cybersecurity terms, people and technologies are the elements. Processes and information flows connect them, and security objectives give them purpose.

Now, back to our discussion about the shift from cybersecurity platforms to systems…

Security teams need shared context, interfaces, and controls to do their work. Products, suites, and platforms can give us part of the foundation, but they’re only part of the overall sequence of real “work.”

The full sequence executed by a human is to observe, build context, decide, act through a control point, check the result, and learn.

Let’s walk through the sequence with some SecOps-related examples.

After observing an anomaly, the first practical step for a security practitioner is to build context across control points. That’s more than just collecting alerts in one place.

Endpoint, network, email, data, identity, and other tools may each see a different part of an incident. An effective cybersecurity system can help connect those observations in a way that drives a final decision or verdict faster or better than humans can do on our own.

The response may need to happen somewhere other than the source of the first signal. For example, an MDR or EDR case could lead a security team to remove a malicious message through an email control or revoke sessions through an identity control.

This type of integration is useful both ways: information comes in, and decisive action goes back out through the right control point.

From there, the system has to know what happened. Did the email get removed? Were the sessions revoked? Did the affected device or account stop generating the signals that prompted the response?

Confirmation of individual response actions (at a technical level) is helpful, but that’s not the same as knowing for certain the security outcome we wanted was achieved.

Learning is easy to neglect, but systems can help codify it. A team can use the result of one incident to make future decisions faster or more effective.

Over a longer period, reviewed cases can improve detections, playbooks, permissions, or models. Some of that work happens within a company’s own environment. Some of it can also come from across a cybersecurity company’s customer base.

Cross-product coordination is the foundation of a cybersecurity system. We needed this long before AI. Agents just made the problem more obvious and urgent because of the speed they work at.

Here’s the other thing: Systems don’t replace platforms. They make platforms more valuable.

AI can help coordinate the work we do faster and more flexibly, but the quality of the underlying cybersecurity platform(s) will still dictate what the system can do reliably.

Systems might just be the missing ingredient that helps platforms deliver on their potential.

Where this leads: coordinated outcomes

I am way, way more excited about systems and outcomes than products or platforms alone.

Our whole world changes when products or modules in a platform start working as a system. I know this may sound utopian, especially right now, but I don’t think it’s as far off as it seems.

There will still be focused products. There will still be suites. Broader platforms may keep expanding without ever covering the entire security market. But the real end game was never just platforms.

Unlike platforms, I don’t think the next phase of cybersecurity will be decided by who uses the word “system” in marketing assets. You have to prove it now.

AI may make the transition between stages happen faster, but it will also expose weak foundations when an agent tries to act across them.

A true cybersecurity system helps a team understand a problem, choose an authorized response, carry it out through the right controls, and learn from the result. A coordinated outcome, not just a piece of software.

Our industry’s systems phase will be decided by whether our human and agent actors can do useful work together, whether the results make the next decision better, and whether we can use them to meet a whole new set of security challenges we’ve never faced before.


Footnotes

¹ Admittedly, I’m slightly less certain about not having one single cybersecurity platform now than when I first wrote this in 2024. AI coding was decent then, but it’s come a long way since. I still think enterprise-grade cybersecurity software is complex enough that we’re many years away from one platform, though.

² Emphasis on “historically.” Many cybersecurity product categories have matured, so recent generations of companies can enter a relatively defined problem space with significant capital and try to speedrun the incumbents. Emerging markets like AI security, cloud security (before that), and so on still tend to follow the traditional pattern, though.

³ Give or take. Total deal count for financing has been either just above or just below 1,000 transactions this decade. This isn’t net-new companies (many raise multiple rounds), but it illustrates the broader point that our industry invests a lot in building new companies.

⁴ I know these stats are basically an industry trope now. I wouldn’t get fixated on the exact number here. You know it’s at least directionally correct if you’ve spent time in an enterprise security environment.

⁵ The total of public company listings goes up to 49 if you include hybrid companies with significant cybersecurity business units.

⁶ People get angry when a bundle gets called a platform, though. That’s collateral damage when suites and platforms co-exist. This happens more than it should.

⁷ I know, I know. So do fire drills. I don’t like those either. Even if it feels like most of your day is driven by panic over objectives, I would still say you’re doing important work towards broad (maybe even idealistic) objectives we all share in the industry.

Share Article

Related Articles

SOS Logo

Cybersecurity, clarified

Strategic intelligence on the cybersecurity ecosystem, straight to your inbox.