Author: Wayne Loke

  • What AI Exposes – Part 6 of 6

    AI is an extension of me in a machine.

    I was three iterations deep into a comic strip, watching Gemini put a collar on a character who’d never had a collar in any version, when I realised I was looking at the same problem I’d spent five articles describing in organisations with thousands of people. Style drift. Definition of done gone soft. The same dysfunction, just compressed down to one person, one laptop, one Friday afternoon.

    The difference was what happened next. I didn’t need a steering committee. I didn’t need to wait for a financial threshold to clear or a board paper to go in three weeks ahead of a meeting. I just changed the prompt and tried again.

    That’s the thing about working at the scale of one. There’s nobody between me and the outcome. No middle layer, no Chinese whispers carrying my intent through three other people before it reaches the work. If the comic strip doesn’t look right, that’s mine to fix, immediately, and there’s a kind of clarity in that, a clean line from decision to consequence that almost never existed in any of the organisations I spent thirty years inside. The same problem that took Stats NZ years to half-resolve, Scrum or Kanban, which method actually fits the work, I can resolve for myself in an afternoon, because I’m the only one who has to agree with the answer.

    It can be a lonely way to work, though. So I’ve built something to sit in that gap. I’ve put an About Me file into Claude, my taste, my voice, the way I actually think, so that what comes back has a chance of sounding like me rather than like the four thousandth LinkedIn post on the same subject. I use other platforms differently, for research, for testing an idea from another angle, for seeing whether a thought holds up when something with a different bias pushes back on it. It’s a handful of different counsel rather than one tool, each one good at a different part of the job, closer to how I used to work inside an organisation than people might expect. I just don’t need an org chart to do it anymore.

    The voice-to-text side of this has surprised me more than anything else.

    For most of my life, I wrote the way I’d been taught to, type it, read it back, correct it, type some more. Mind maps. Handwritten notes. Bullet points I’d later try to expand into sentences. It worked, in the sense that things got written, but it never quite captured how I actually think, because how I actually think doesn’t arrive in tidy bullet points. It arrives the way I talk, in long connected runs, one idea pulling the next one behind it.

    I’d tried voice-to-text before, years ago, and it never worked for someone like me. English is my third language, after Cantonese and Bahasa Malaysia, and I don’t sound British or American, and the older speech recognition tools simply couldn’t hear me properly. Wrong words, wrong spelling, wrong context, dictation that fought my accent instead of capturing it. I gave up on it more than once.

    Wispr Flow is the first one that actually heard me. It captures the nuance, the pronunciation, the pace, and somewhere in using it I came across Zinsser’s Writing to Learn, which gave me the missing piece, the idea that thinking happens in the act of saying something out loud, not after it, and that stopping to fix a sentence halfway through is stopping the thought itself. Once I let that go, once I let myself talk fast when the idea was exciting and trust that it would be captured properly, the writing changed. All five articles before this one came out of that process, voice first, structure after. I don’t think I’d have written any of them the way I did, ten years ago, with the tools that existed then.

    So when people ask me what I actually think AI is, after trying most of the major platforms over the past couple of years, ChatGPT, Perplexity, Gemini, Claude, Deepseek, Manus, Grok, after the courses, the prompt engineering, the experiments that went nowhere, here’s the true answer.

    It’s still a bicycle. That part hasn’t changed since I first wrote it. You learn the rules of the road, the etiquette, what it’s good for and what it isn’t, and once you know how to ride it, it gets you from one place to another faster than walking, without doing the walking for you. I’m not dependent on it. I’m faster because of it.

    What’s changed is what I think the bicycle is actually carrying.

    It feels less like a personal assistant now and more like an extension of my own thinking, put into a machine that can hold a version of my taste, my voice, my judgement, and reflect it back to me in different forms, depending on what I need that day. Gemini draws in something close to my comic style. Claude drafts in something close to my voice. The research tools dig into a topic the way I would, if I had the hours to do it myself. Each one is carrying a piece of how I think, and I’m the one deciding what to do with what comes back. None of them is me.

    There’s a newer part to this that wasn’t here when I started the series. Some of the counsel runs on its own now, on a schedule I set once and mostly leave alone. Every week a report on my finances arrives without me asking for it, the numbers pulled together, the movements flagged, the few things I’d actually want to look at sitting near the top. Every week the same happens for the Pipitea Mews Body Corporate Committee I sit on, the updates gathered and drafted so I meet the BC Manager already knowing what has changed, instead of finding out halfway through the meeting like everyone else. I read them, I decide what matters, I act or I don’t. But the fetching and the first draft were done before I sat down, and that used to be somebody’s whole job.

    That’s the part I didn’t quite see coming. At the scale of one, I’d assumed the recurring work, the weekly rhythm a team used to carry, would either land back on me or simply not get done. It turns out a good deal of it can be handed over and checked rather than done from scratch, which is closer to working with a capable team than anything I expected to find on my own. The machine doesn’t decide what the numbers mean. It doesn’t sit in the committee. It does the gathering and the first pass, the way a junior analyst would, and it leaves the judgement where it has to stay, with me.

    That’s the part the bicycle metaphor still gets right, even now. The bicycle doesn’t choose the destination. I do. It just makes the distance shorter, and it means the work doesn’t have to be lonely the way I once worried it might be in retirement, no office, no team, no water cooler. I find I’m more connected now than I was for a lot of my working life, because I can take an idea, set it down, walk away from it, come back later with something new to add, and the thinking is still there waiting, sharper than it would have been if I’d had to hold the whole thing in my head the entire time.

    Looking back across everything in this series, the pattern holds at every scale I looked at it. AI won’t remove ANZ’s production bottleneck. It won’t dissolve Stats NZ’s immunity system or smooth over the gap between a steering committee report and what was actually happening on the ground. It won’t move Te Pūkenga’s funding any faster, and it can’t read a room the way thirty years of watching people abandon ship taught me to. None of that changed because the scale dropped from an organisation of thousands down to one retired man and his laptop. What changed is that at this scale, I’m the only constraint left to find, and most days, I’m also the one who gets to fix it.

    Age and experience don’t buy certainty. I don’t have much more of that than I did at thirty. What they buy is recognition: I’ve seen enough versions of the same mistake, in enough organisations, that I spot it faster now, in myself, in a prompt that isn’t working, in a strip that needs a sixth pass before the collar disappears. AI hasn’t replaced that. If anything, it’s given me more places to practise it.

    My daughter grew up mostly hearing titles. Project Manager. CIO. Senior Manager. Platform Engineering Director, toward the end, a title I’m not sure she ever fully understood, and neither did most people I worked with. None of those titles ever told her what the work actually was.

    What I’d want her to know, more than any of it, is that the work was always about the people in the room, not the tool on the desk. That was true with Scrum. It was true with cloud migrations and steering committees and budgets stuck above someone’s financial delegation. It’s still true now, with AI sitting on my desk instead of a project plan. The tool changes what’s possible. It has never once changed what actually matters, which is whether you can read the people around you really, and whether you treat them well while you do it.

    This website was never meant to be a memoir. I’d like her to read it one day, not for the organisations or the job titles, but to see how her father learned to work with what he was given, with enough discipline to keep going and enough humility to know when to change course. If she takes one thing from it, I hope it’s this: learn from the people who went before you, take their mistakes as your own lessons, and you’ll fumble less than I did.

    AI makes that easier now than it’s ever been. It gives you a faster bicycle to go find out what you actually believe. It doesn’t do the believing for you.

    I’m still riding mine.

  • Knowledge Is Cheap Now

    For 4 years those boxes sat in the closet, the filing cabinets, the shelf in the corner, and I never once opened them. I moved to Hamilton in 2022 and the paper came with me, taped up, labelled, promising to be useful one day. Now I’m moving house again in 3 weeks, so I finally cut the tape. Most of what I found was printouts. Articles from Harvard Business Review, research notes from Gartner and Forrester, journal pieces, newsletters, things I printed off before AI, before the internet was much use, and a fair bit of it from before the internet at all. I printed them because that’s what you did. You couldn’t read a Gartner report on your phone on the train back then, so you printed it, folded it into your bag, and read it on the commute, or in a meeting where you looked busy and were, occasionally, actually learning something.

    People talk about hoarding as retail therapy, and I understand the pull of it: buying things with your own hard-earned money is a small hit of happiness, and there’s nothing wrong with that. My hoarding was a different animal. Mine was knowledge. Over 30-odd years across Southern Bank, ABN AMRO, the National Bank/ANZ, Inland Revenue, Stats NZ, I gathered paper the way other people gather shoes, because in my line of work the paper was the point. It was the raw material of doing the job well. So I kept it, and kept it, and moved it from house to house, and told myself I’d use it again one day.

    The maybe bucket is a trap

    So I did what you’re supposed to do. Three buckets. Keep, throw, and maybe. The keep pile is easy and the throw pile is easy. The maybe pile is where good decluttering quietly goes to die.

    I read something years ago about the maybe bucket that stuck with me, and it turns out to be true every single time. The maybe bucket is not a decision. It’s a holding pen for your own uncertainty. You put a thing in there because you can’t decide, and when you come back to look at it again, you are exactly as undecided as you were the first time, so it lives forever. Never filter the maybe bucket. The moment you start sorting it, picking through it a second time, you’ve handed your uncertainty a way to keep everything.

    My rule is blunter than the books make it sound. Hell yes, I keep it. Hell no, it’s gone. And maybe is hell no, because the hesitation is the answer. If I have to think about it, I already know. Then I run one test on the whole lot: would I miss this a month from now, 2 months from now? These boxes have been sealed in a closet for 4 years and I haven’t missed a single page. So the answer writes itself. Would I use it again? Probably not. Would I ever find it again if I wanted it, buried in a box, filed under a system I set up a decade ago and have long forgotten? No. You never find the thing you archived by hand. And if I genuinely want the idea back, well, you’ve got AI now. You’ll remember you once read it, you’ll ask, and it will hand you a better version than the yellowing page you were clinging to.

    Knowledge got cheap

    Here’s the thing that landed on me somewhere in the middle of the second filing cabinet, kneeling on the floor with a bin bag. Knowledge is cheap now.

    When I started out at 19, people paid for what you knew. You acquired it, you studied it, you carried it around, and that accumulation was your edge. It was why you got hired, why you got promoted, why someone senior asked for your read on a problem. I made CIO at ABN AMRO in my early thirties, and a good chunk of what got me there was simply that I’d done the reading nobody else had bothered to do. The paper in my bag was professional capital. It had a price, and I’d paid it, in commutes and late nights and subscription fees.

    Now anyone can Google it. Better than that, they can ask an AI to do the deep dive and explain it back at whatever level suits them, like they’re 5 years old, or like they’re a corporate executive, or like they’re a CIO who needs it in the language of a board paper. The thing that took me years to gather, a curious teenager can now summon in a sentence, in better shape than I ever kept it. That paper mountain in my closet, the one that was capital once, is recycling now. I’m not being dramatic about it. I’m literally putting it in the recycling.

    A comic where two people talks about dumping collection of books and research where the internet is now the source. That means your lived experience is potentially a database too.

    What I couldn’t throw out

    And yet I kept some of it. Not much, but some.

    I kept my own notes. The newsletters I wrote, the journals of my own thinking, the observations I jotted down over the years that were never online and never will be. That’s not knowledge I can Google, because nobody has published it, including me. It’s mine, and I keep it because it feeds what I write now on Strategic Kata. The rest, the borrowed knowledge, the reprints of other people’s cleverness, I let go, because the internet holds a better copy than my cabinet ever did.

    My wife ran the same test faster than I did. She had the course materials from her Cafe Service Level 2 Certification and her Certificate in Dental Assisting from the New Zealand Dental Association. I asked if she wanted to keep them. She said, give me some time. Less than an hour later she came back and said, chuck it out, I’m not working anymore, and the information’s probably dated by now anyway. Then she said the thing that is the whole point of this piece, without meaning to say anything profound at all: the rest is in my head. The course materials went in the no bucket. The certificates she kept. The knowledge stayed exactly where it had always lived.

    The information was the cheap part all along. What held its value was what we did with it, and what stayed in our heads after the paper was filed and forgotten. AI didn’t take that from anyone. It just made it obvious, by making the information itself free. The stuff that’s still worth something is the stuff that was never really on the page: the judgment, the pattern recognition, the scar tissue from having been in the room when it went wrong.

    Books are where my rule breaks, and I’ll admit it plainly. I can’t part with physical books. I have a hardcover of Management Control Systems, 9th edition, Anthony and Govindarajan, from my MBA days. I read it front to back and went back to it for years. The information in it is dated now too, of course it is. That’s not why I keep it. I keep it for the earmarks, the notes crammed in the margins, the sticky tabs, the yellowing pages, the whole record of me arguing with the author across the page. Master that book and you know what to look at and what to consider, and the rest is delegation. What I value there is the layer of my own thinking sitting on top of it, and you can’t print that off Gartner.

    One room at a time

    Then there’s the rest of it. The extra furniture nobody sits on, the kitchen appliances we bought and used twice and left to gather dust. There’s a quiet lesson in an appliance you haven’t touched in a year. I’m a minimalist by nature, and my wife is learning to be one, though it’s taken her time and a few house moves to get there.

    We’ve got 3 weeks, so we’re doing one room at a time. AI handled the checklist and the paperwork, which used to eat a weekend on its own. Thinking about it, this move is the least stressful we’ve had, and the reason is simple: we’re both home full time now. The hard moves were the ones where we both worked, packing at midnight, arguing about who booked the truck. This time we can sit on the floor with a box of old crockery and utensils and stuff we’d forgotten we owned, and talk, and laugh about where it came from. Sentimental value? Plenty. Worth keeping? Not really. So it’s sold, or given to someone who’ll use it, or dropped at the Salvation Army or the op shop.

    The lawn mower and the trimmer are the last things to go, and they can’t go yet. I’ve still got to mow and trim the lawn one last time before I hand the keys back to the property manager, so those two stay until the tenancy exit is done. Small mercy on the other end: the place we’re moving to has no lawn to mow at all.

    We upsized once. We downsized once. Now we downsize again. Life is good.

  • What AI Exposes – Part 5 of 6

    AI can read what’s written down. It can’t read the room.

    I was naive the first time I got close to real power.

    I was CIO for the Malaysian operation at ABN AMRO, and it was the first time I’d been anywhere near the corridors where decisions actually got made. Up to that point, I’d assumed that what happened in meetings was mostly what was happening. I learned quickly that wasn’t true. There were deals behind the scenes, agreements reached in rooms I wasn’t in, what I’d call horse trading, the kind of give and take between senior people that never appears in any paper, any minutes, any business case.

    I didn’t have anyone showing me the ropes before that. No mentor pulling me aside to explain how things actually worked, the way some people get. Most of what I learned, I learned by getting it wrong first, reading senior leadership moves badly, more than once, because I’d assumed, the way management textbooks teach you to assume, that everyone in the room was working toward the organisation’s stated goals and that personal interest sat somewhere well below that. I went looking for better teachers wherever I could find them, suppliers, older employees who’d seen more cycles than I had, and through my MBA, where I sought out senior executives at other listed companies who could tell me what the textbooks left out. Some of them taught strategy and financial management at business school themselves. What they gave me wasn’t in the assigned reading. Looking back, those were my most formative years, the MBA and ABN AMRO running at the same time, each one teaching me something the other couldn’t.

    What you see in the room may not be what it actually means. And if you raise something, even something real, something you’ve genuinely identified as a problem, it can go nowhere. Not because anyone disagrees with you. Often nobody disagrees with you at all. You get lip service, you’re told it’s a good point, worth considering, and then you’re kept busy, genuinely busy, with work that feels important, until at some point you’re told to stop. No budget. Leadership doesn’t approve it. Other priorities. Whatever the reason given, it’s rarely the real one.

    I used to think that was wasted effort. I don’t think that anymore. I think it’s closer to camouflage. Keeping you occupied with something real enough to feel meaningful means you’re the one focused elsewhere, not the one asking questions, not the one creating friction around whatever the actual direction of travel is. You become part of the cover, without ever being told that’s the role you’re playing.

    Here’s what that looked like in practice. The Managing Director issued a directive, get rid of personal colour printers across the operation, replace them with three leased colour copiers instead. Sound economics, better cost control, easier to monitor consumption. I followed it. Then the Asian Financial Crisis hit, and in the middle of it, Treasury delivered the bank’s entire year’s profit target in a single week. The Head of Treasury came to me asking for a colour printer of his own. I declined. We had a directive from the MD, and as far as I was concerned, that was the end of the conversation.

    I hadn’t read the room at all. I was reprimanded for it, called penny wise and pound foolish, for refusing to sign off something so small against the scale of what Treasury had just delivered for the bank. The policy was real. The directive was real. None of that mattered once a senior leader had just made the year for the entire Malaysian operation. That’s when I understood what equity, fairness, and meritocracy actually meant inside an organisation, and what they didn’t. The dots I needed to connect weren’t in the policy document. They were in who had just earned the standing to ignore it.

    That was ABN AMRO. The lessons came one humiliation at a time, and I was new enough to all of it that I didn’t yet see the whole pattern underneath.

    Years later, at Te Pūkenga, I understood it completely.

    We had real work underway, a digital transformation to bring 25 separate polytechnic and vocational education platforms down to one, a programme that would have cut close to a third of the operating cost across the network. The case for it was strong. And it kept hitting roadblocks, one after another, each with its own reason, no funding yet, no approval yet, other priorities ahead of it, despite earlier agreement that it would proceed.

    By that point I’d learned to read the room differently. Human nature doesn’t change much, whatever era you’re in. When the people closest to the centre, the ones who hear things in rooms you’re not in, start quietly updating their CVs, start having conversations that sound like they’re already planning their next move, that tells you something the org chart never will. It’s a filter of a filter. They’re reacting to information you don’t have access to, and their behaviour is the signal.

    At Te Pūkenga, a change in direction was coming for the organisation, well before it was formally announced. I’m not going to comment on whether that decision was right or wrong, that’s well outside anything I have a view worth sharing on, and frankly it was beyond anyone at my level to influence either way. What I will say is that the roadblocks we kept hitting, the funding that never quite arrived, the approvals that kept slipping, made a different kind of sense once you understood that the organisation’s future was already an open question at a level none of us could see directly. The work was being held in place because committing real money to a transformation that would take years to deliver didn’t fit with a future that might not include the organisation in its current form.

    Nobody told us that. Nobody could have, probably, even if they’d wanted to. We worked it out by watching what the people around us did, not what they said.

    Sense-making is connecting what you observe, what’s said directly, what’s implied, how people’s behaviour shifts, against everything you already know about how organisations and the people running them tend to behave, until a shape emerges that nobody has confirmed but that turns out, later, to be correct.

    Not every signal points toward self-interest, though. I learned that at ANZ too, watching a different kind of decision get made.

    Graham Hodges joined ANZ New Zealand as CEO in November 2005. The decision to merge the ANZ and National Bank mainframes and systems had already been sitting in analysis since 2003, two rounds of review, a mountain of documentation, no call made. During his tenure, he made it, the kind of decision a more cautious leader might have let sit for another review cycle. Then he did something nobody had budgeted for. He proposed ANZ lead the development of a Snapper card for public transport payment in New Zealand, something like London’s Oyster card, or the Octopus card he’d have seen used in Hong Kong. Nothing in the existing programme of work accounted for it. He asked for budget to be pulled from other initiatives to fund it. I was in the room at a leaders’ update when he told the whole leadership team why, that New Zealand needed it, and that ANZ’s brand belonged at the front of building it, not because any individual business unit had asked for it. Snapper went live in July 2008.

    Nobody saw that coming. Every CEO ANZ HQ had sent out before him had played it safe. This read differently to me, closer to conviction than caution, the kind of bet a leader makes when belief in an idea outweighs the safety of staying inside existing budget lines. The bet didn’t cost him either. He left ANZ New Zealand in May 2009, promoted to Deputy CEO of the ANZ Group. Reading the difference between someone protecting themselves and someone genuinely backing an idea they believe in is its own kind of sense-making, rarer than the other kind, and worth getting right, because mistaking one for the other costs you something different each time.

    And here’s where I think AI sits in all of this, and where it doesn’t.

    AI is extraordinarily good at the documented layer of an organisation. The papers, the reports, the meeting minutes, the business cases, the dashboards. All of that, AI can read, summarise, generate, analyse, faster than any of us could. Parts 3 and 4 in this series were both, in a sense, about that documented layer, the steering committee pack, the business case waiting for approval above a financial threshold. AI makes all of that faster and more polished.

    The decision lived in belief and fear, sometimes tangled together the way they were at ABN AMRO, sometimes belief running clean on its own, the way it did with Hodges and his Snapper card. Either way, it sat with a small number of people at the apex of an organisation, employees too, whatever the org chart said about their seniority, their real stake nothing more than their position, their income, their title. And almost none of that gets written down. It can’t, often, because writing it down would make it real in a way that creates its own problems. So it travels through behaviour instead. Through who goes quiet. Through which topics never quite get scheduled. Through the slight change in tone in a room that, on paper, hasn’t changed at all.

    AI has no access to that layer. It was never going to. The information doesn’t exist anywhere AI can read it, because it was never put anywhere in the first place, deliberately.

    So here’s what I think happens as AI takes over more and more of the documented layer. The documented layer gets faster, cleaner, more abundant, and it becomes less and less where anything is actually decided, because it never really was. What’s left, the undocumented layer, the belief and fear sitting with a handful of people, doesn’t shrink. It becomes more clearly the only layer that ever mattered, just harder to see precisely because everything around it now looks so comprehensive and so well-documented.

    AI is a product, like every other product that’s come through an organisation before it. New tools change what gets produced and how fast. They don’t change who holds the decision, or what that decision actually runs on. That’s been true of every wave of technology I’ve worked through, and I don’t see why this one would be different.

    What’s changed for me, I think, is that I stopped expecting the documented layer to tell me anything useful about where things were heading, a long time ago. I learned to watch the room instead.

    It’s served me well. Whether it’s something AI can ever do, I genuinely don’t know. I suspect not.

  • Arcane by Necessity

    At 2:00 am, just after the bots from the tax refund agents finished scraping the portal, we rebooted the servers.

    The servers were healthy. That was the point. By the time an out-of-support WebSphere Application Server tells you it’s unhealthy, you’re already inside a priority one incident, and a system serving every taxpayer in New Zealand is falling over with the whole country watching. So we rebooted while the country slept, on a schedule built from our own data, and nobody outside the team ever knew it happened.

    This was Inland Revenue between 2013 and 2015, and this piece is about how a team ran the machinery of the New Zealand tax system on technology officially past its life, with methods you could politely call arcane, because we had no other choice.

    The Mandate

    My boss that ran Business Platform Services was a tough fellow who made sure you knew what you signed up for. Hands off otherwise, which suited me fine. He had one famous line, delivered whenever shit hit the fan: “We are here to protect the integrity of the tax system. This is what you signed up for and what the Act says.”

    The Act is the Tax Administration Act 1994, and he wasn’t quoting it for effect. The integrity of the tax system is a statutory obligation, and whenever a P1 landed, or the tax agents flooded us with complaints about the portal, he’d deliver the reminder again: you are responsible and accountable for the integrity of the tax system.

    So the obligation was fixed. Absolute, written into law, indifferent to our circumstances.

    The resources were fixed too, in the other direction. Inland Revenue was waiting for business transformation, and everything in the estate was being sweated to survive until the new world arrived. We were two versions behind. End of WebSphere Application Server 6.1 support from IBM due on September 2013. Aging hardware, aging tools, and no appetite to invest in a platform that transformation was going to retire anyway. My team, Business Platform Services Channels, owned the digital services end to end: online, B2B, voice, and output services. That was about 40 percent of Inland Revenue’s ICT applications, but 68 to 70 percent of all application interfaces flowed through Channel, because we were the layer between the customer and everything behind us in the service line: database, then middleware and application, then the mainframe at the back.

    And the load had teeth. Tax peak ran from the end of March to August. Tax agents ran refund services that scraped our portal relentlessly, their bots hammering the front end through the night on behalf of clients chasing refunds. Every one of those requests travelled through infrastructure that its own vendor no longer supported.

    When the obligation is absolute, the resources are frozen, and the load keeps rising, the only variable left is method.

    Year One: Reactive

    Our first year was fraught. When a WebSphere server hit its limit, the fastest recovery was a reboot, so we rebooted, sometimes in the middle of the business day. And sometimes the second server failed while the first was coming back, because it took the full load alone and buckled under it, which turned a recovery into a bigger incident than the one we started with. We were learning the hard way that on old infrastructure, the cure can cascade.

    Late night calls were routine. Teru, my Senior IT Consultant, would get the pager, an impromptu meeting would form on the phone with the Incident Resolution Lead and whichever technical specialists the failing system had dragged out of bed, and the discussion usually ended the same way: a call to reboot. I made that call, every time. The responsibility and onus sat with me as the end-to-end service owner for digital services, and that’s what owning a service means at two in the morning.

    We logged 33 priority one incidents that first year. I told the story of what the whiteboards and the visible metrics did about that number in Two Kinds of Visibility. This piece is the other half of how the numbers came down.

    The Turn

    We learned the behaviour of our servers the way you learn the behaviour of an old car: by paying attention to it for long enough. We collected data. We correlated system variables, CPU, cache, database log files, with the tax calendar and with application releases, and slowly the patterns showed themselves. It wasn’t rocket science. A server that buckles every time a certain filing date meets a certain cache level isn’t mysterious; it’s predictable, and predictable is workable.

    We built benchmark numbers for when each server would need a reboot to clear its cache. And once we could forecast the failure, we stopped waiting for it. We scheduled the restarts for the hours when New Zealand was inactive and asleep. Dawn, just before people came to work. Or 2:00 am, just after the refund agents’ bots finished their scraping runs. The same reboot that had been a midday emergency in Year One became an invisible maintenance routine in Year Two, and that shift is a large part of why Year Two improved so dramatically on Year One.

    The War Room

    Through tax peak we ran a war room. Every morning, representatives from Channel, Database, Middleware and Application, Mainframe, and Service Management sat in one room, the entire service line from customer to mainframe, and I chaired the briefing.

    The rhythm was fixed. Yesterday’s summary first: what we saw, what happened, what action was taken. Then each area updated the health of its systems and applications in turn. Then the tax calendar, so our activities synchronised with what the country was about to do to us. Then the threshold forecasts, which servers were approaching a reboot and when we’d take them. Then pending releases from Development and any fixes we had scheduled ourselves. Service Management kept us compliant with the ITIL process flows and coordinated the preparation whenever a service outage had to be planned for. And when something critical needed a decision on the spot, the BPS managers joined the room and the decision got made there and then, with everyone who’d have to live with it present.

    Written down like that, it sounds almost formal. In practice it was a group of people running a national system on obsolete kit with a whiteboard, a pager, and a forecast, and everyone in the room knew exactly why. The Act didn’t care what version of WebSphere IBM still supported.

    What Constraint Actually Does

    Constraint tends to get either romanticised or pitied, and I think both readings miss what happens inside it. Nobody in that war room would have designed the operation this way. Given the choice, we’d have taken supported servers, modern tooling, proper capacity headroom, and full nights of sleep. We were never given the choice, and the obligation didn’t shrink to fit the resources, so ingenuity had to close the gap. The data collection, the correlation with the tax calendar, the benchmark thresholds, the dawn reboots: every one of those practices exists because a gap existed that money wasn’t going to close.

    That’s usually what arcane methods are, when you stumble across them in an organisation and wonder why on earth things are done that way. They’re the archaeology of some past gap between what had to be done and what was available to do it with. Before you rip one out, it’s worth finding the gap it was built to span, because the gap may still be there.

    Business transformation did eventually arrive, years after I left in October 2015, and it retired the estate we’d been sweating. I doubt anyone still remembers the 2:00 am reboots, and that was the whole idea; the best nights were the ones nobody noticed.

    It was what we signed up for.