Work examples
The symptoms are always different. The underlying problem rarely is. A selection of times it got sorted.
As Very Good Technology
2026 onwards.01 A Website They Could Run Without Us
01The problem
A business that had changed shape was still being described by a website that hadn’t. Worse, nobody inside the company could change it. Every correction meant a phone call to somebody else, and a monthly fee for the privilege. That is a small cost repeated forever, and it quietly decides what a business is willing to say about itself, because anything worth saying has to be worth the hassle of asking.
02What I did
Fixed scope, fixed fee, and a writer brought in so the words weren’t an afterthought. Six pages, built from a prototype they walked through before anything was committed, so the arguments about design happened on the prototype rather than on the live site. Then a handover guide, a session running them through it, and the whole thing moved into their ownership.
03The result
Built, launched and handed over inside a month. They run it themselves. The monthly hosting and support cost went away with the platform it belonged to. Nobody has been called back to change a page, which was the point.
Handover is a deliverable, not a courtesy. We built for the day we left.
02 Turning AI Noise Into a Costed Decision
01The problem
A consultancy with real technical depth knew AI mattered and couldn’t get past that sentence. The pressure arrived from every direction at once: clients, staff, competitors, the news. Underneath it sat a specific commercial question nobody had asked plainly: how much of an engineer’s week goes on work that cannot be billed, and could any of it realistically be given to a machine. The risk was never doing nothing. The risk was doing the loudest thing first.
02What I did
Four workshops on site, across strategy and direction, people and capability, how work actually gets done, and technology and systems. AI opportunities sit in the gaps between those four, not inside any one of them. What came out became ten requirements, a solution overview, and a roadmap that put order-of-magnitude costs and a sequence against every opportunity. Discovery and strategy only, with nothing sold on the back of it.
03The result
Delivered and paid in full, with the contingency days unused. They went from a list of possibilities to a sequenced, costed roadmap they could act on or decline item by item, with the reasoning attached to each one, including a clear account of where AI was not the answer.
You can’t prioritise a list of things nobody has costed. We put a number and an order on all of it.
Before Very Good Technology
Twenty years of the same problem wearing different clothes.03 Nothing Was Broken, Which Was the Problem
01The problem
Thirty years of research had produced a technology estate nobody had designed. Systems arrived when someone needed one, vendors were chosen by whoever was asking, and sensitive research data sat across infrastructure whose backup arrangements the board had identified as a risk to the organisation continuing at all. Nothing was visibly failing. There was no incident to point at, only spending nobody could see the whole of and no way to tell whether the next decision would be better than the last thirty years of them.
02What I did
Designed and documented a project management methodology with the CEO, executive and department heads, and built the organisation’s first formal tender process alongside it. Then proved it by running the hardest project through it: replacing every server with virtualised infrastructure and off-site redundant backup, with validation built in for research data that could not be lost.
03The result
The infrastructure and disaster recovery work landed ahead of schedule and twenty per cent under budget, with no data loss and no disruption to studies in progress. The methodology went from technology projects to all of corporate services. Spending became visible, and the security position on research data could be demonstrated rather than assumed.
Nothing was broken. That is exactly what made it hard to fund, and why it had to be proven on a real project rather than argued for.
04 The Savings Had Already Been Sold
01The problem
Five thousand people across film, audio, visual effects, digital and print, and clients threatening to leave over competitiveness. The savings automation was meant to produce had already been forecast, sold to clients, and built into the company’s numbers. So the deadlines were not internal targets. They were commitments someone else had already made, in a business that had never once coordinated anything across all five production disciplines.
02What I did
Built the programme structure that did not exist: a steering committee with the head of every production area, scope and work plan agreed with the Chief Production Officer and the PMO, weekly delivery meetings by discipline and monthly steering above them. Each area found and made the case for its own automation. My job was governance and clearing obstacles, not choosing what four very different crafts should have in common.
03The result
Automation delivered across multiple disciplines, including in-house digital localisation and personalisation, automatic subtitle transcription for film and television, and print production localisation. The programme ran eight months as the company’s highest-profile global initiative. The full savings target was not reached inside that window: it depended on change capacity the organisation did not have while it was also delivering to clients.
The savings had already been sold. Finding them and being given time to realise them are two different jobs, and only the first one was mine.
05 Winning More Work and Getting Worse at It
01The problem
The agency was winning more work, and client and team satisfaction were both falling as it did. Sixty people, and no shared way of running a project, so every new win made delivery slightly worse than the last one. Estimation was judgement rather than evidence, which meant profit was discovered after the fact. Nobody was doing anything wrong, which is why it had gone on so long.
02What I did
Established a standardised project management practice across the whole agency, working with the head of every department and reporting to the CEO and owners. Built a six-step design, development and delivery process, and rebuilt estimating on the agency’s own historical financial data instead of on how long people felt things took.
03The result
Client retention improved and the digital team grew by twenty per cent. The agency won a £10m DHL content marketing contract, took Allianz, and pitched Aldi, with credentials it could now evidence rather than assert.
Estimates got better the moment they came from the company’s own numbers rather than from anyone’s experience.
06 Twenty-Seven People and Four Years of Intending To
01The problem
Twenty-seven people, a required ten per cent cost reduction, a brand rollout, and a backlog of new functionality, all at once. The department had needed restructuring for four years and it had never happened, because restructuring is the thing everyone agrees to and nobody schedules. University web teams sit in a particular squeeze: academics expect consultation, administration expects speed, students expect commercial-grade digital, and none of it is funded as though those three things cost anything.
02What I did
Took the department as interim director and rebuilt it: team structure, operating model, delivery approach and development approach together, rather than one at a time. Recruited new roles into the shape rather than fitting the shape to the people already there. Then recruited the permanent Director to run it, and handed over.
03The result
Twenty per cent off overhead against a ten per cent target, with the brand rolled out, the backlog cleared, and search, recruitment campaigns and new sections delivered. Site stability, traffic, engagement and usability all improved. So did the team’s Net Promoter Score, during a restructure.
It had needed doing for four years. Restructuring is always the thing everyone agrees to and nobody schedules, and the only way out is to be the person who schedules it.
07 Keeping the Funding Without Killing the Idea
01The problem
GSK NEXT existed to find market gaps and build products quickly, and it was funded by a pharmaceutical company whose R&D, Finance, Legal and Compliance functions have governance requirements for good reasons. Too little governance and the funding stops. Too much and you have removed the thing you were funding. Four theme founders were running the work, and there was no operating model that satisfied both ends of that at once, nor any incubator for the external start-ups the strategy depended on.
02What I did
Rebuilt the operating model with the CEO and the four founders so it met GSK’s governance requirements while leaving the founders their autonomy, and ran the monthly reporting that kept the parent organisation able to see what it was funding. Built the start-up incubator curriculum with an external provider, and a selection process using weighted criteria against consultation across the whole organisation, so the choices were defensible to Legal and to everyone who had been asked.
03The result
Two cohorts of six start-ups each through the full curriculum, several of which secured funding and continued GSK support. An operating model and selection framework that outlasted the engagement, and continued investment from the parent organisation, which was the thing actually at risk.
The governance had to protect the funding without killing the thing being funded. Those pull in opposite directions and somebody has to hold both.
08 Eighty Per Cent Less Team, Better Website
01The problem
A government directive to centralise web services arrived at the same time as unforeseen funding cuts, so a five-person team had to become one person. The website itself carried live problems: misinformation risk, compliance gaps and functional faults, on a site healthcare providers and patients used to find information that mattered. Cutting the team by eighty per cent and leaving the site as it was would have been the easy version and the wrong one.
02What I did
Documented what was actually wrong, then built a prioritised roadmap agreed with leadership at both organisations and secured funding for it despite the cuts. Ran two tenders for development and content services inside government procurement. Recruited the Senior Content Designer who would carry the service afterwards. Set up a Strategic Working Group with the right seniority, terms of reference and ways of working, so one person had governance behind them rather than only a workload.
03The result
One person, running a better website than five had. User experience improved, target audience usage grew measurably, and organisational risk went down. The operating model held after handover: the team still supports and maintains the site.
Five people to one is only a cut if you leave the service as it was. Rebuild it first and it is a different thing entirely.
09 Half the Budget Gone, Two Users
01The problem
Half a million spent of a million-dollar budget, and two users. The project had run past its approved timeframe and the funding was about to be discontinued with nothing to show. The organisation employed Australia’s most published smoking cessation researcher, whose field is precisely what makes someone stop smoking, and the requirements document had overruled them.
02What I did
Stopped development and went to the funders first, because a project with no time left cannot be fixed quietly. Agreed a new baseline with the Cancer Council and the federal funders that allowed the work to continue at all. Then rebuilt the thing around the Transtheoretical Model of Behavioural Change, running research and workshops with the cessation experts, policy advisers and researchers who had been consulted and ignored, and ran the delivery as a combined waterfall and agile approach with the external agency.
03The result
Traffic up three hundred per cent, and quit.org.au launched as an evidence-based cessation tool that is still one of Australia’s most visited health websites. It won a Leadership and Innovation Award, and it opened ongoing research into technology-assisted cessation.
The expert was already in the building. Nobody had let them decide anything, and the fix was mostly asking them and then defending the answer.
10 Four Countries, Several Acquisitions, One Brand
01The problem
Europe’s largest IT infrastructure provider had grown by acquisition and looked like it. Several companies, four countries, and a digital presence that told prospective clients they were dealing with a collection rather than a leader. The requirement was one brand, centrally manageable, live across the UK, Belgium, France and Germany. The failure mode was never a bad launch. It was four launches, at four times, in four ways, each one locally defensible.
02What I did
Ran it as one delivery rather than four national projects reporting into a centre. Ran workshops to identify the audiences the company was actually targeting and fed those into the design rather than asserting a brand at them, coordinated the bespoke server infrastructure underneath, and kept every stakeholder in every territory informed whether they wanted to be or not.
03The result
Launched on time and on budget across all four territories. It became the largest and most profitable project in the agency’s history, was nominated for UK marketing awards, and the client became the agency’s first retained account, asking for more work immediately.
Several acquisitions do not become one company because the logo changed. They become one company when the four countries agree what they are launching, and that is a conversation, not a design.
11 Three Hundred Courses Sold, None Scheduled
01The problem
The company had promised custom course development to every new client it won, and kept winning clients. Three hundred developments were live with no delivery dates and no view of what resourcing them required. Teams were working hard. Nobody could answer when will it be done, because the answer depended on several teams whose ways of working were each defensible alone and did not fit together, so work stalled wherever two of them met.
02What I did
Went to every team and mapped the process that actually existed rather than the one described, then built a single methodology that kept each team’s approach and fixed the handovers between them. Trained and mentored people into it rather than announcing it, because a methodology nobody was taught is a document.
03The result
All three hundred previously unmanaged projects came under formal management with committed dates. Client relationships recovered, courses went out for AFL clubs, Adidas and others, and one of them won the LearnX Best Budget Course award in 2015.
The backlog was a sales problem wearing delivery clothes. You can improve delivery all you like, but somebody has to say out loud what was promised.
12 Everyone Could Describe Their Value. No Two the Same Way.
01The problem
Two utilities had merged, and a government agreement required the new organisation to demonstrate the value it delivered. The consultation ran into something more awkward than silence: every business unit could describe its value, and no two described it in the same terms. With no common baseline, the executive could not compare anything to anything, which meant it could not implement the operating model, and functions that could not be compared would eventually be measured the only way that is always available, which is cost.
02What I did
Fixed the strategic objectives as the reference point first, then secured executive agreement that value would be measured against them in cost, return on investment or time, and nothing else. Ran the workshops with business units across both former organisations to document and quantify what they actually did against that baseline, working under the Enterprise Value Office with a change manager and a business analyst.
03The result
Adopted in full by the Executive Leadership Team, and it became the method for evaluating the operating model change across every functional unit. The government-required self-audit was completed on the same framework.
Everyone could describe their value. No two teams described it the same way, and until that was fixed nothing could be compared, so nothing could be decided.