Systems don't fail on features. They fail on people.
Most rollouts don't collapse because the software is broken. They collapse because the people expected to use it feel threatened, overloaded or ignored — and quietly carry on the old way.
I hold a degree in communication psychology and twenty years in the room. Below is a real NHS project, start to finish, showing exactly what I do at every stage — and exactly where an unqualified trainer would have failed it.
Skim the bold lines and you'll get the whole story in 30 seconds. The detail is there for the people whose job is to check it.
The premise
What actually kills a rollout.
Threat, not difficulty
People reject a system when it threatens their competence, control or status — not because the buttons are hard.
Power, not the org chart
The person who can quietly kill adoption is rarely the one with the senior title. Miss them and the programme stalls.
Consent, not compliance
You can mandate attendance. You cannot mandate use. Real adoption is negotiated, one honest objection at a time.
One real project, five stages
One real project, five stages
Bradford District Care NHS Foundation Trust · SystmOne outcome recording · Adult Mental Health
The brief: appointment outcome recording — a recognised challenge right across the NHS — wasn't happening consistently among clinical staff, including psychologists. Instruction alone hadn't shifted it, and national reporting requirements were at stake. The Trust needed the behaviour to change, and needed it to stick.
This is the kind of contract where the average unqualified trainer lasts eight weeks. Here's why I didn't.
Stage 1 · Analysis
Diagnose the resistance before building anything
The situation. On paper, a training problem: a recording process not being followed. Every previous instruction-led attempt had failed to move it.
What the standard trainer does
Assumes non-compliance means ignorance. Builds a module that explains the process louder, mandates completion, and calls it done. The audience — clinicians with doctorates — sits through it, nods, and changes nothing. Eight weeks later the trainer is gone.
What I did
I treated the resistance as data. Analysis showed the blocker was behavioural — and intelligent, not ignorant. These were evidence-led professionals declining to act on an instruction that had never been justified in the terms their profession runs on. Telling them to do it was always going to fail, because "because we said so" is precisely the standard of evidence they are trained to reject.
The theory driving it. This is Lukes' third dimension of power in practice — the resistance wasn't in the org chart, it was in the professional culture shaping what staff saw as legitimate. And it's loss aversion (Kahneman & Tversky): the change read as an attack on clinical autonomy, not an admin tweak. You cannot train past a grievance no one has named. I named it.
What the client gotA diagnosis that reframed the entire project: not "staff need training on the process" but "staff need a professionally credible reason to adopt it." Everything downstream was built on that finding.
Stage 2 · Design
Build the learning in the audience's own framework
The situation. An audience that rejects internal mandates — but whose entire working life runs on one thing: externally validated, evidence-based guidance.
What the standard trainer does
Writes content from the internal process documents. To this audience, that's the same discredited voice that already failed — repackaged with a progress bar.
What I did
I designed the module in the audience's own epistemic framework. Every justification was sourced from recognised national authorities — NHS England, NHS Resolution, national patient-safety bodies the audience already trusts and is professionally obliged to follow. Nothing sourced internally. Deliberately. The design reframed outcome recording as what it actually is: a national professional standard, not a local demand. The consequences of incomplete recording were stated plainly, including the human cost, then calibrated with clinical reviewers to land without alienating.
The theory driving it. Knowles' andragogy — adults adopt what is relevant, self-directed and consistent with their own experience. Self-determination theory (Deci & Ryan) — the design handed autonomy back: here is the evidence; you're the professional; draw the conclusion. And a deliberate reframe of the perceived loss: from "you're being policed" to "this is the standard your profession already holds."
What the client gotA design the audience could not dismiss without contradicting their own professional standards. It was signed off and delivered complete.
Stage 3 · Development
Protect the build — from overload, and from scope
The situation. A live build in a regulated environment, with clinical SMEs, a systems developer, and a sign-off chain — and, inevitably, mid-build requests for things outside the brief.
What the standard trainer does
Absorbs every request to seem helpful, burns the hours, misses the deadline, and looks like the failure. This is the single most common way the eight-week trainer dies.
What I did
Built screenshot-accurate simulation in Articulate Storyline against the real workflows, with structured SME input and a formal storyboard sign-off gate before development — so clinical leads corrected a document, not a finished build. When an out-of-scope technical request landed mid-build, it was documented and routed through the agency the same day. Nothing was built until it was authorised — and the legitimate extra work was covered by a properly negotiated extension, not swallowed.
The theory driving it. Cognitive load theory (Sweller) governed the build itself — chunked, sequenced, one workflow shown one way, because overload reads to the user as "this system is too hard" and hardens into resistance. The change-control discipline isn't theory; it's what twenty years of watching contracts die teaches you.
What the client gotAn accurate, clinically signed-off build, delivered inside a controlled scope — and an agency relationship protected rather than burned.
Stage 4 · Implementation
Land it with the hardest audience in the building
The situation. Delivery into a workforce where instruction had already failed once — including the profession best equipped on earth to see through a manipulative pitch: psychologists.
What the standard trainer does
Talks at the room. Reads the slides. Treats silence as agreement. In a sceptical clinical audience, silence is not agreement — it's the disengagement that shows up as "failed adoption" three weeks later.
What I did
The module led with the audience's own objections, gathered in analysis and stated in their own words, before asking for anything — and conceded what was true: the old process genuinely was difficult, and that was not their failure. Only then did it show the fix. Adoption was framed as a clinical responsibility, evidenced by national authorities, with support explicitly offered rather than competence assumed.
The theory driving it. Psychological safety (Edmondson) — making it safe to have been resistant, so no one had to defend their past position to accept the new one. Active listening applied in design: the objections in the module were the audience's real objections, not invented ones.
What the client gotA rollout that met its audience where the resistance actually lived — instead of pretending the resistance wasn't there.
Stage 5 · Evaluation
Define success before you build — and build the measure in
The situation. The Trust didn't buy a training module. It bought a change in recording behaviour, under national reporting requirements.
What the standard trainer does
Reports completion rates and a happy feedback form, and leaves. Evaluation is never designed at all — it's improvised at the end, if anyone asks.
What I did
Success was defined at the analysis stage — recording behaviour in the live system, not module completion — and the evaluation instrument was built into the module itself: completion structured as evidence of process acknowledgement, assessment by demonstration in a simulated live workflow rather than quiz recall. The evaluation wasn't an afterthought; it was the destination the whole build was aimed at, agreed before a single slide existed.
The theory driving it. Kirkpatrick's four levels, designed for the level that matters: behaviour on the job and the organisational result — not whether people enjoyed the course.
What the client gotAn evaluation aimed at the thing the Trust actually bought — behaviour in the live system — designed in from the analysis stage, not bolted on at the end.
The method is the point — because it repeats
The method doesn't change.
For anyone evaluating a bid: nothing above was improvised. It's a repeatable structure — ADDIE as the backbone, SAM where an agile build fits — with formal sign-off gates, documented change control, and evaluation defined before design begins. The sector changes. The resistance changes. The method doesn't.
Full methodology detail, including the wider learning-science grounding, on the methodology page.
For the agencies who place me
I'm the safe bet.
Your risk is a contractor who needs managing, or who wobbles the moment the room turns. I'm the opposite. Put me in front of a difficult client, a hostile audience or a programme that's already drifting, and I come out making you look like you found exactly the right person.
Get in touch
Got a rollout people are resisting?
Tell me where it's stuck. I'll tell you straight whether it's a training problem or a behaviour problem — and exactly how I'd fix it.