Building a Champions Network That Works, part 1 of 7
Security champions are not new. Organisations have been recruiting volunteers to carry security into their teams for at least fifteen years, with mixed results. So when we tell clients that this is the moment to start, or restart, a champions programme, the fair question is what has changed. Quite a lot, as it happens. Five separate shifts have converged in the last two or three years, and together they make the case stronger than ever.
The attacks have moved to where champions sit
For most of the last decade, the security team could reasonably expect to catch most social engineering at the perimeter. Filters improved, phishing got easier to spot, and awareness training could focus on the obvious tells. That period is over. Generative AI has removed the spelling errors, clumsy grammar, and generic greetings that made phishing recognisable, and it has made voice and video impersonation cheap enough to use against middle managers rather than only against chief executives. The attack that matters now is the plausible one: the supplier email that reads like a real supplier, the call from a colleague whose voice is genuinely their voice.
Nothing at the perimeter catches that. It is caught, or missed, in the moment a person decides whether to check. That decision is shaped far more by what their immediate team does than by anything the security function says. A champion in the room, visibly pausing and verifying, is the control that operates at exactly the point where the modern attack lands.
AI is quietly eroding the habit of checking
The second shift is inside the organisation, not outside it. As people offload more thinking to AI assistants, research is starting to show a consistent pattern: productivity goes up, and independent verification goes down. A recent three-wave longitudinal study found that as daily AI use rose past 90 per cent, participants' confidence in their own ability to check AI output fell, most sharply on complex tasks. The older automation literature predicted this. Reliable automation reduces monitoring, and it reduces it most when people are busy or tired.
Verification underpins nearly every security behaviour, so a decline in it is a security problem even though it looks like a productivity story. The corrective isn't a ban on AI tools, which would fail, but a social one: keeping checking behaviour visible and normal within teams. That is what champions do. They model the pause. They talk about the time the assistant was confidently wrong. They keep "I checked" in the team's vocabulary at the moment it is most at risk of dropping out.
Regulators now expect culture, not just controls
The third shift is regulatory, and it has moved faster than many security leaders have noticed. NIS2 requires management bodies to approve and oversee cyber risk measures and to ensure staff receive appropriate training. DORA asks financial entities to build ICT risk awareness across the whole workforce, not only in the technology function. The NCSC's Cyber Assessment Framework has a dedicated objective on culture and awareness, and ISO/IEC 27001:2022 tightened the awareness and competence requirements in its updated Annex A. Boards are increasingly being asked, in plain terms, what they are doing about security culture.
A champions programme is one of the few interventions that answers that question with evidence. It produces coverage figures, engagement data, reporting rates and network measures that an auditor can inspect. Training completion rates were once enough. They are not anymore, and a champions network is the most credible thing a security leader can put in front of a regulator asking what happens in the business after training is complete.
Security teams cannot scale, but networks can
The fourth shift is economic. Security budgets are flat or falling in real terms in most sectors, headcount is hard to secure, and the organisations we work with are asking their security functions to cover more ground with the same people. Hiring is not the answer, and in most cases it is not available.
A champions network is the only part of a security programme that gets cheaper and more effective as it grows. Each new champion extends coverage into a team the central function would never reach, and each one strengthens the signal for the champions around them. At a time when every function is being asked to justify its cost, a control that scales without headcount is rare.
The tools to run it well finally exist
The fifth shift is the one we find most interesting. Earlier champions programmes failed for a reason that had little to do with the idea and everything to do with execution. They could not see themselves. A programme manager had a list of names and no way to know whether those names connected to anyone, whether the network reached the teams that mattered, or whether it was changing behaviour at all. Programmes drifted into newsletters and lunch-and-learns because those were the things you could count.
That constraint has lifted. Organisational network analysis, once a research technique, is now practical at enterprise scale, and it shows whether champions actually sit in the paths of influence within a team or are isolated at the edge. Behavioural measurement has moved beyond click rates to reporting behaviour, escalation patterns and norm indicators. The research base has matured alongside the tooling: we now understand from the work on complex contagion why a champion's value depends on their ties rather than their title, and from behavioural economics why a visible cooperating peer sustains cooperation in others. A programme started today can be designed around that evidence from the first week, rather than discovered by trial and error over three years.
What waiting costs
One version of this argument says the pressures above will only intensify, so the sooner you start, the better. That is true, but it undersells the point. The real cost of waiting is that norms are being set right now, in every team, around how people use AI tools and how carefully they check what those tools produce. Once a team has converged on "the tool checked it", moving it back is far harder than shaping the norm while it is still forming. Established behaviour is expensive to change. Behaviour that is still settling can be steered.
A champions programme launched in the next twelve months arrives while those norms are still open. One launched in three years arrives to find them closed.
Starting well
None of this means rushing. The programmes that fail are the ones launched as a call for volunteers, with no sponsor, no measurement, and no idea what champions are for. The ones that last begin with a clear mechanism, a named executive owner, a benefits model and a way of seeing the network as it grows. We have written separately about how to build that business case. The argument is simpler: the conditions for a champions programme to succeed have never been better aligned, and the window to shape rather than repair behaviour is open now.
About this series. Building a Champions Network That Works is a seven-part guide from CyBehave to starting, funding, recruiting, scaling and measuring a security champions programme, and to the shift from awareness to behaviour change that sits underneath all of it. Next: Part 2, Making the business case for security champions.