First AI steps for lean security teams
How to introduce AI into your security workflows in an intentional, lean and consistent manner
LLM chatbots have democratised automation. With little technical effort, you can quickly set up automations to improve painful workflows. Small security teams can reap large time-saving and waste-cutting benefits by using LLM technologies. In fact, most security teams, large or small, have now integrated some form of LLM-driven automation into their processes.
The challenge is that adoption typically happens spontaneously and haphazardly. In most security teams, LLM automations tend to be developed out of necessity to address immediate, tactical needs. As a result, teams can quickly find themselves surrounded by an amalgamation of LLM automations that solve specific tasks without a coherent vision.
When this happens, the individual operators developing the automations feel the immediate gains in their daily work. It is not the overall team. When automations are developed exclusively bottom-up, the team may struggle to share them consistently. It may feel that automation development lacks the coherent, top-down strategy needed to address their problems in a targeted way.
Fortunately, there are three easy steps to quickly establish a simple, yet coherent, top-down LLM automation strategy: first, define LLM-driven automation goals that drive intentional adoption. Second, adopt lean project management. Third, promote consistent automation sharing.
Defining goals that drive intentional adoption
Security teams should always adopt technologies intentionally and in a measured way. If technologies do not actively serve human goals, boost productivity, and protect the team’s well-being, they will inevitably create costly distractions that waste resources and lead to digital burnout.
This is especially true for lean security teams, given the higher operational demands and pressures they face. Adopting LLM automations in this context is no exception.
Two fantastic LLM automation goals can drive intentional adoption. The first goal is to design, deploy and refine LLM workflow automations across core operational processes to double output capacity. This goal reduces manual effort so lean security teams can deliver twice the value without increasing headcount or work hours. Put simply, this goal embodies “growing without growing”.
Under this goal, success is proven by measuring 2x productivity gains. Here, you can use two measurements to see whether this goal is being achieved: first, at the management level, measure Total Output per FTE. This metric is computed from two additional workflow-level measurements: Cycle Time and Manual Process Touchpoints.
Cycle Time refers to the total time elapsed from workflow request to completion. Manual Process Touchpoints measure the number of human interventions required per task. Together, these metrics let your team measure whether the developed automations reduce cycle time by 50 to 70% and cut repetitive manual steps by at least 80%.
The point of this goal is to amplify human potential, not squeeze out more labour. By removing high-friction, repetitive tasks, these automations aim to create space for your team to think. This way, they can focus on higher-leverage functions such as problem-solving, cross-department collaboration and strategy development.
The second goal complements the first. It aims to reduce or eliminate manual steps in security tasks that typically require high effort, such as pre-deployment threat modelling or security architecture reviews that cause deployment bottlenecks. Similarly, manual security remediations caused by poor deployment practices can take up far too many hours of a team's time.
A good waste reduction goal aims to streamline security operations by cutting non-value-added security overhead by 50% while accelerating delivery and maintaining a secure technology stack.
To work towards this goal, lean teams must first define what non-value-adding security overhead is. On the flip side, three high-value-add activities can be targeted: embedding threat modelling into early feature shape-up, shifting controls left into development and down into infrastructure, and improving automated security checks during pre-deployment phases.
Lean security teams can use two metrics to measure a 50% reduction in non-value-added security work: the ratio of Pre-Deployment to Post-Deployment Vulnerabilities Found. The second is Mean Time to Resolve once vulnerabilities are discovered. Both capture how effectively architectural flaws and code defects are caught upstream.
This waste-reduction goal can eliminate security bottlenecks and turn the team into an architectural enabler. LLM-driven automations can embed threat modelling directly into the feature shape-up phase. Conversely, LLM-driven dynamic security testing can easily complement Static Application Security Testing tooling in the pre-deployment phases.
Adopting lean project management
Great, well-defined goals will help drive intentional adoption, but they won't actually do the work for you. Your team will have to. The automation potential within LLM platforms is great. However, they don't carry the intrinsic knowledge needed to build an automation development process that fits your team.
Moreover, the generalist nature of LLM platforms introduces hidden engineering challenges that can hinder rapid adoption. First, LLM frameworks are not tailored for security work. In most cases, your team will need access to LLM models trained for cybersecurity tasks, especially when security testing is a central goal.
Second, LLM tools alone can't handle complex, multi-step procedures, so teams need to adopt an agentic approach when building automations. This inevitably compounds the learning curve and development complexity.
Third, while multiple agentic harnesses exist today (like Mistral Workflows) to accelerate the development of agentic systems, security teams still need to spend significant engineering effort configuring agents properly. For starters, they need execution guardrails so agents can operate within predefined security and cost parameters. Second, they need engineering skills that align agents with your team's existing specialised workflows.
There is only one way to overcome these engineering challenges: getting your team organised so it can deliver improvement work without drowning in project management overhead.
One approach is to adopt a lean project management framework. As mentioned in our article on project management for lean security teams, most small teams naturally start by adopting frameworks such as Scrum or Agile.

These frameworks can all provide useful ideas. The problem is that lean security teams tend to adopt these frameworks too literally, creating project management frameworks that - despite claiming to be Agile - are too bloated for small teams and tend to get in the way inevitably.
A lean project management framework embraces a minimalist approach that uses only the parts that give the team enough delivery structure without forcing them to abandon their natural way of working. This way, the team can channel consistent improvement work in a manner that can be sustained during a normal operational week.
The ingredients of lean project management are simple: define the team's needs, borrow the useful parts of existing frameworks, and rely on the smallest possible toolset your team favours.
Promoting consistent sharing
Clear goals and an effective project management framework will help kickstart consistent development. However, neither will ensure your team knows what LLM automations are available to them and how to use them. Moreover, they won't ensure your team has a consistent way to share feedback on existing automations or new ideas. Only a closed feedback loop built on consistent information sharing will.
The closer you can place this feedback loop to security operations, the better. In cybersecurity, the best feedback and most impactful automation ideas generally come from colleagues who are involved in daily operations. They're most exposed to workflow bottlenecks and pain points, so they are best positioned to highlight where improvement is needed.
The challenge for lean security teams is picking the right meeting or format to share feedback or improvement ideas. A good approach is to focus on weekly, well-attended meetings. Weekly security handovers, for example, are a good medium to promote information sharing; even more so if their subject is security operations.
SecOps handover meetings are the perfect space to communicate new automations or share feedback on how to improve them. As mentioned in our article on running agile security operations, a good weekly handover meeting (also known as a handshake protocol) should always have a well-defined structure that covers engineering topics such as detection developments and any relevant engineering backlogs.

The engineering agenda section of these weekly meetings can be easily expanded to cover LLM automations and discuss whether they are concretely reducing waste and speeding up operations. Moreover, they provide the right channel for sharing automation news with the widest possible audience, since attendance at these meetings is generally mandatory.
Conclusion
It is easy to feel that AI isn't being adopted fast enough when teams are stuck building LLM automations haphazardly, without the guidance of a coherent development strategy. Getting out of this rut requires stepping back and introducing AI into your security workflows in an intentional, lean and consistent manner.
For small security teams, this sounds more complicated than it is: defining goals focused on intentional AI adoption will help channel development efforts in a clear direction. At the same time, embracing a lightweight delivery framework supported by regular exchanges on the topic will bring the development focus, momentum, and energy your team has been craving.


