<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
<channel>
  <title>Beyond the Pilot</title>
  <link>https://inteevo.com/writing</link>
  <description>Short, useful writing on AI strategy, governance and delivery, from Inteevo.</description>
  <language>en-gb</language>
  <item>
    <title>Everybody is an AI expert now, and nobody carries the can</title>
    <link>https://inteevo.com/writing/everybody-ai-expert-nobody-carries-can</link>
    <guid>https://inteevo.com/writing/everybody-ai-expert-nobody-carries-can</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>The interface changed, so fluency started passing for understanding. Meanwhile the money in any real AI project goes on data quality, integration and accountability, none of which demos well. Demonstration debt is what you owe when the proof of concept cannot reach production, and someone else always repays it.</description>
  </item>
  <item>
    <title>The models are eating what the models wrote</title>
    <link>https://inteevo.com/writing/models-eating-what-models-wrote</link>
    <guid>https://inteevo.com/writing/models-eating-what-models-wrote</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>The photocopy analogy for model collapse is too gentle. The real version has someone filling in the unreadable gaps with a pen, in good faith, until the page reads perfectly and means nothing. New research suggests the fix is smaller than expected, and it turns provenance into a buying question.</description>
  </item>
  <item>
    <title>Brain in a jar, please standby</title>
    <link>https://inteevo.com/writing/brain-in-a-jar-consumer-ai-privacy-gap</link>
    <guid>https://inteevo.com/writing/brain-in-a-jar-consumer-ai-privacy-gap</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Same model, same Tuesday, two people in one building. One is protected by encryption, retention policy and a governance document. The other is telling a chatbot about their mother. The questions every business asks its suppliers as a matter of routine are the ones no consumer thinks to ask.</description>
  </item>
  <item>
    <title>Thirty spelling mistakes and you bin it, thirty wrong facts and you sign it</title>
    <link>https://inteevo.com/writing/overconfident-ai-automation-calibration</link>
    <guid>https://inteevo.com/writing/overconfident-ai-automation-calibration</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>A proposal with thirty spelling mistakes gets binned on sight. The same proposal with thirty wrong facts gets actioned. Every surface signal we learned to read has been decoupled from the thing it used to indicate, and in chained systems the errors get polished rather than degraded.</description>
  </item>
  <item>
    <title>Fix the brakes before you fit the new engine</title>
    <link>https://inteevo.com/writing/transformation-small-steps-fix-brakes-first</link>
    <guid>https://inteevo.com/writing/transformation-small-steps-fix-brakes-first</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Every struggling organisation has two corners: one where people copy data between spreadsheets by hand, and one where an executive presents a two million pound revolution. Incremental work has a property big-bang work cannot: the benefit arrives before the money runs out.</description>
  </item>
  <item>
    <title>The best automation idea in your business is already in the building</title>
    <link>https://inteevo.com/writing/democracy-of-discovery-automation-ideas</link>
    <guid>https://inteevo.com/writing/democracy-of-discovery-automation-ideas</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Broken processes stop being visible once the workaround becomes muscle memory. Ask people where AI could help and you get silence; ask what part of their week makes them want to lie down and you get detail, with feeling. The value usually hides between departments.</description>
  </item>
  <item>
    <title>If you think AI is coming for your job, you are the person best placed to take it</title>
    <link>https://inteevo.com/writing/microsoft-stack-developers-ai-advantage</link>
    <guid>https://inteevo.com/writing/microsoft-stack-developers-ai-advantage</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>You did not spend twenty years learning syntax. You accumulated an instinct for how technology meets a business problem, built by being present when things went wrong. A model has read about that. It has not stood in it at half past six on a Friday.</description>
  </item>
  <item>
    <title>One pair of brackets, four hundred thousand pounds</title>
    <link>https://inteevo.com/writing/document-processing-garbage-in-garbage-out</link>
    <guid>https://inteevo.com/writing/document-processing-garbage-in-garbage-out</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Brackets mean negative in accounting. When extraction loses them, a loss becomes a profit and everything downstream inherits it as fact. The arithmetic is impeccable, the premise never true, and three steps at 95% accuracy leaves you at 86% overall.</description>
  </item>
  <item>
    <title>Stop asking it to write the thing, start asking it to take the thing apart</title>
    <link>https://inteevo.com/writing/ai-persona-critic-not-author</link>
    <guid>https://inteevo.com/writing/ai-persona-critic-not-author</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>You should not ask these systems to write your work, because you cannot tell a board to hold on while you ask why it suggested something. What you can build is the critic you never get access to: a commercial sceptic who asks what happens at thirty per cent of projection.</description>
  </item>
  <item>
    <title>The analyst who will not show their working</title>
    <link>https://inteevo.com/writing/blind-trust-intelligent-verification-enterprise-ai</link>
    <guid>https://inteevo.com/writing/blind-trust-intelligent-verification-enterprise-ai</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>You would not tolerate an analyst who refused to show their working. The enterprise platforms generating your forecasts do exactly that, and never get awkward about it. Three questions for the supplier meeting, and why scaling scrutiny beats spreading it thinly.</description>
  </item>
  <item>
    <title>Hallucination is the kindest word we could have chosen</title>
    <link>https://inteevo.com/writing/when-ai-lies-hallucination-wrong-word</link>
    <guid>https://inteevo.com/writing/when-ai-lies-hallucination-wrong-word</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Hallucination is a medical word and a merciful one: it grants the status of a patient rather than an author. What actually happens is the manufacture of supporting evidence, complete with plausible citations that lead nowhere, and we have different words for that when people do it.</description>
  </item>
  <item>
    <title>The risk was on page seventeen and the summary did not think it mattered</title>
    <link>https://inteevo.com/writing/great-dumbing-down-productive-friction</link>
    <guid>https://inteevo.com/writing/great-dumbing-down-productive-friction</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Someone senior explains cheerfully that they have not read a full document in months. Then you ask about the risk on page seventeen, and the pleasure goes out of the conversation. Summarising is a judgement about what matters, made by something that does not know what matters to you.</description>
  </item>
  <item>
    <title>Parameter count is the new horsepower</title>
    <link>https://inteevo.com/writing/ai-arms-race-parameter-count-horsepower</link>
    <guid>https://inteevo.com/writing/ai-arms-race-parameter-count-horsepower</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Nobody selects accounting software on calculations per second, yet we compare AI by parameter count. The specification argument is harmless; the release cadence is what costs money, freezing decisions that would otherwise be straightforward.</description>
  </item>
  <item>
    <title>The blindfolded painter</title>
    <link>https://inteevo.com/writing/two-ai-agents-beat-one-second-pass</link>
    <guid>https://inteevo.com/writing/two-ai-agents-beat-one-second-pass</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Text generation is a painter working blindfolded, placing each stroke by where they think the last one landed. Asking for quality mid-generation asks for assessment of something that does not exist yet. A genuinely fresh second instance sees the whole, and costs almost nothing.</description>
  </item>
  <item>
    <title>The more we automate, the busier everyone gets</title>
    <link>https://inteevo.com/writing/workslop-epidemic-ai-productivity</link>
    <guid>https://inteevo.com/writing/workslop-epidemic-ai-productivity</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Workslop is work that looks finished and is not, and it does not eliminate labour, it relocates it onto whoever opens the attachment. Almost nobody sending it is being cynical: the polish now exceeds the recipient's ability to judge it, which is a genuinely new problem.</description>
  </item>
  <item>
    <title>Substantially less prone</title>
    <link>https://inteevo.com/writing/substantially-less-prone-ai-sycophancy</link>
    <guid>https://inteevo.com/writing/substantially-less-prone-ai-sycophancy</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Not free from. Not eliminated. Substantially less prone. A reduced tendency towards deception and sycophancy is still a tendency, and of the two, the eager agreement is the one that will cost you money, because it does not feel like flattery.</description>
  </item>
  <item>
    <title>Can we just use this?</title>
    <link>https://inteevo.com/writing/illusion-simplicity-prototype-production</link>
    <guid>https://inteevo.com/writing/illusion-simplicity-prototype-production</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>The demonstration convinces precisely because everything absent from it is invisible. Error handling, audit trails, accessibility, integration and maintainability are perhaps sixty per cent of the work and none of it photographs. Meanwhile the tool has already chosen your architecture.</description>
  </item>
  <item>
    <title>Nobody comes home and announces they only used 0.45 gallons</title>
    <link>https://inteevo.com/writing/stop-counting-tokens-measure-outcomes</link>
    <guid>https://inteevo.com/writing/stop-counting-tokens-measure-outcomes</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Each attempt looks cheap, which is what makes the total invisible: rephrase, clarify, regenerate, and twenty five thousand tokens later you have something usable. Token consumption became the headline metric while output quality got filed under secondary considerations.</description>
  </item>
  <item>
    <title>The review you need is never available when you need it</title>
    <link>https://inteevo.com/writing/micro-workforce-ai-personas-review</link>
    <guid>https://inteevo.com/writing/micro-workforce-ai-personas-review</guid>
    <pubDate>Tue, 28 Jul 2026 09:00:00 GMT</pubDate>
    <description>Scrutiny is not unavailable in principle, it is unavailable now, and now is when the document gets finished. Building a useful critic is about temperament rather than job title: not "a CFO" but someone impatient with the phrase "once we scale".</description>
  </item>
</channel>
</rss>
