predictes is an AI transformation partner: strategy, engineering and managed services that take AI from a slide to something running in your business.https://predictes.comenhttps://predictes.com/blog/the-device-estate-is-missing-from-the-business-casehttps://predictes.com/blog/the-device-estate-is-missing-from-the-business-caseField AI business cases count the model and the software and forget the devices. Who enrols, updates and replaces them decides the real cost of owning the system.Wed, 07 Oct 2026 00:00:00 GMTfield deploymentdevice managementbusiness caseedge AIhttps://predictes.com/blog/support-contracts-grow-by-default-write-one-that-shrinkshttps://predictes.com/blog/support-contracts-grow-by-default-write-one-that-shrinksKnowledge drifts to the supplier and the support contract grows. For an AI system the opposite works better: a handover schedule and a declining support tier.Sun, 04 Oct 2026 00:00:00 GMToperationshandovercontractsknowledge transferhttps://predictes.com/blog/how-to-judge-a-supplier-who-promises-a-lights-out-planthttps://predictes.com/blog/how-to-judge-a-supplier-who-promises-a-lights-out-plantLights-out proposals tend to lead with machines and models. The ones worth trusting lead with exceptions, read-only access and what should stay staffed.Fri, 02 Oct 2026 00:00:00 GMTdark factorysupplier selectionunattended operationdecision recordshttps://predictes.com/blog/ai-training-converts-weeks-after-the-workshophttps://predictes.com/blog/ai-training-converts-weeks-after-the-workshopThe workshop shows people what is possible. The weeks after show them what is in the way, and that is where training is won or lost.Tue, 29 Sep 2026 00:00:00 GMTtrainingcapabilityadoptionmeasurementhttps://predictes.com/blog/govern-the-ai-portfolio-not-each-project-separatelyhttps://predictes.com/blog/govern-the-ai-portfolio-not-each-project-separatelyPer-project approval gates slow delivery and still miss the real risks. Governance that keeps pace uses proportionate lanes and evidence teams already produce.Fri, 18 Sep 2026 00:00:00 GMTAI governanceportfolioriskISO/IEC 42001https://predictes.com/blog/before-anyone-builds-your-ai-application-agree-whose-it-will-behttps://predictes.com/blog/before-anyone-builds-your-ai-application-agree-whose-it-will-beOwnership of an AI application is decided in the first week, not at handover. What to agree before the work starts, and the questions that reveal a hidden dependency.Thu, 17 Sep 2026 00:00:00 GMTownershiphandovervendor lock-incontractshttps://predictes.com/blog/what-a-good-ontology-project-leaves-behindhttps://predictes.com/blog/what-a-good-ontology-project-leaves-behindAn ontology project is easy to sign off as a careful diagram. What should remain for the model to outlive a quarter, and how to spot it in the proposal.Mon, 14 Sep 2026 00:00:00 GMTontologyacceptancevendor selectionhttps://predictes.com/blog/how-much-really-goes-through-untouchedhttps://predictes.com/blog/how-much-really-goes-through-untouchedStraight-through processing is the headline number of every automation programme and the easiest one to flatter. Four definitions decide whether it means anything.Sun, 13 Sep 2026 00:00:00 GMTstraight-through processingmetricshyperautomationoperationshttps://predictes.com/blog/write-down-what-would-make-you-stop-before-the-pilot-startshttps://predictes.com/blog/write-down-what-would-make-you-stop-before-the-pilot-startsA spec-driven pilot without a stopping rule never fails, it just gets extended. Agree what evidence means widen and what means stop before the first spec is written.Sat, 12 Sep 2026 00:00:00 GMTspec-drivenpilotsdecision makingAI codinghttps://predictes.com/blog/a-good-transformation-partner-gets-smallerhttps://predictes.com/blog/a-good-transformation-partner-gets-smallerIf a transformation partner is as involved in month nine as in month two, something has gone wrong. How to write the decline into the contract, and check it.Thu, 10 Sep 2026 00:00:00 GMTknowledge transfersupplier managementcontractstransformationhttps://predictes.com/blog/the-first-hour-of-an-agent-incidenthttps://predictes.com/blog/the-first-hour-of-an-agent-incidentWhen an agent does something nobody intended, the first hour depends on decisions made months earlier: its identity, its logs and who can switch it off.Wed, 09 Sep 2026 00:00:00 GMTAI agentsincident responseaudit traillogginghttps://predictes.com/blog/a-process-that-works-for-one-team-erodes-at-twentyhttps://predictes.com/blog/a-process-that-works-for-one-team-erodes-at-twentySpec-driven delivery works in the pilot. Rolled out to twenty teams, gates get bypassed, reviews thin out and the dashboards stay green. How to keep it honest.Sat, 05 Sep 2026 00:00:00 GMTengineering leadershipspec-drivencode reviewprocesshttps://predictes.com/blog/ai-use-case-list-with-no-ownerhttps://predictes.com/blog/ai-use-case-list-with-no-ownerLong use-case lists feel like progress and function as a way of avoiding a decision. What replaces them is shorter, harder to write and actually gets built.Fri, 04 Sep 2026 00:00:00 GMTAI strategyprioritisationuse casesdecision makinghttps://predictes.com/blog/in-an-ai-design-proposal-read-the-prototype-plan-firsthttps://predictes.com/blog/in-an-ai-design-proposal-read-the-prototype-plan-firstDesign proposals for AI products usually look like any other. A few details show whether the author has ever designed for a system that is sometimes wrong.Sat, 29 Aug 2026 00:00:00 GMTvendor selectionproduct designprototypingAI interfaceshttps://predictes.com/blog/adoption-is-a-budget-linehttps://predictes.com/blog/adoption-is-a-budget-lineEvery transformation plan has a technology budget and an adoption paragraph. That imbalance predicts the outcome more reliably than anything about the technology.Fri, 28 Aug 2026 00:00:00 GMTchange managementadoptionprogrammesbudgetinghttps://predictes.com/blog/data-decisions-an-engineer-should-not-make-for-youhttps://predictes.com/blog/data-decisions-an-engineer-should-not-make-for-youAI data work often stalls on business decisions rather than on technology. Which decisions those are, how much time they take, and what happens when nobody makes them.Tue, 25 Aug 2026 00:00:00 GMTdata qualitydata ownershippersonal datadocumentshttps://predictes.com/blog/ask-an-agent-supplier-for-the-decision-log-not-the-demohttps://predictes.com/blog/ask-an-agent-supplier-for-the-decision-log-not-the-demoA demonstration shows an agent on its best day. The questions that separate a production supplier from a demo supplier are about bad days, logs and limits.Sat, 22 Aug 2026 00:00:00 GMTAI agentssupplier selectionprocurementdecision logshttps://predictes.com/blog/an-agent-is-as-safe-as-its-toolshttps://predictes.com/blog/an-agent-is-as-safe-as-its-toolsMost attempts to make agents safe happen in the prompt. Almost all of the real safety comes from the tool layer, which is an engineering decision.Fri, 21 Aug 2026 00:00:00 GMTAI agentssecuritytool designprompt injectionhttps://predictes.com/blog/an-ai-business-case-should-be-argued-withhttps://predictes.com/blog/an-ai-business-case-should-be-argued-withAI business cases fail less often because the number is wrong than because nobody can see what the number rests on. Write the assumptions down first.Thu, 20 Aug 2026 00:00:00 GMTbusiness caseAI strategycostsmeasurementhttps://predictes.com/blog/design-for-the-review-not-the-answerhttps://predictes.com/blog/design-for-the-review-not-the-answerThe metric that decides whether an AI feature saves anybody time is not accuracy. It is how long it takes a person to confirm or correct what it produced.Fri, 14 Aug 2026 00:00:00 GMTUXAI interfacesreviewtrusthttps://predictes.com/blog/on-an-ai-product-team-the-domain-expert-is-not-a-stakeholderhttps://predictes.com/blog/on-an-ai-product-team-the-domain-expert-is-not-a-stakeholderSoftware teams consult the people who know the work. AI product teams need them inside, with time set aside, because the product is measured against their judgement.Wed, 12 Aug 2026 00:00:00 GMTteam designdomain expertsoperating modelproduct managementhttps://predictes.com/blog/a-pilot-with-one-crew-is-a-specificationhttps://predictes.com/blog/a-pilot-with-one-crew-is-a-specificationField pilots are often run to prove that a system works. The useful ones are run to find out how it fails, and they need a different crew, scope and exit criteria.Sun, 09 Aug 2026 00:00:00 GMTfield deploymentpilotsrolloutadoptionhttps://predictes.com/blog/retry-the-model-never-the-paymenthttps://predictes.com/blog/retry-the-model-never-the-paymentPipelines that mix models with ordinary systems retry constantly. Unless every step that changes state is idempotent and recorded, a retry becomes a duplicate.Sat, 08 Aug 2026 00:00:00 GMTidempotencyretriesaudit trailorchestrationhttps://predictes.com/blog/build-the-evaluation-harness-firsthttps://predictes.com/blog/build-the-evaluation-harness-firstWithout a way to score changes, an AI codebase gets slower every month. Teams feel it around month three and usually blame the wrong thing.Fri, 07 Aug 2026 00:00:00 GMTevaluationtestingLLMregressionhttps://predictes.com/blog/an-sla-for-an-ai-system-what-can-be-promisedhttps://predictes.com/blog/an-sla-for-an-ai-system-what-can-be-promisedA conventional SLA measures uptime, and an AI system rarely goes down. What belongs in a contract for running one, and which promises should worry you.Wed, 05 Aug 2026 00:00:00 GMTSLAcontractsoperationsprocurementhttps://predictes.com/blog/an-agent-sandbox-is-mostly-about-what-can-leave-ithttps://predictes.com/blog/an-agent-sandbox-is-mostly-about-what-can-leave-itOnce an agent writes and runs code, typed tools no longer bound it; the environment does. Isolating the process is the easy half. The rest is what can leave.Mon, 03 Aug 2026 00:00:00 GMTAI agentssandboxingegress controlisolationhttps://predictes.com/blog/night-shift-knowledge-belongs-on-the-risk-registerhttps://predictes.com/blog/night-shift-knowledge-belongs-on-the-risk-registerIn many plants the judgement that keeps production running sits with a few experienced people. Capturing it pays off whether or not the lights ever go out.Sun, 02 Aug 2026 00:00:00 GMTdark factorytacit knowledgerisksuccessionhttps://predictes.com/blog/your-board-does-not-need-a-prompting-coursehttps://predictes.com/blog/your-board-does-not-need-a-prompting-courseExecutives rarely need to use AI tools better. They need to judge proposals, costs and risks well enough to decide, and that calls for a different half-day.Sat, 01 Aug 2026 00:00:00 GMTexecutivestrainingproposal reviewriskhttps://predictes.com/blog/most-retrieval-failures-are-definition-failureshttps://predictes.com/blog/most-retrieval-failures-are-definition-failuresWhen search returns something technically related and practically wrong, the usual response is to tune the retriever. The usual cause is elsewhere.Fri, 31 Jul 2026 00:00:00 GMTontologyretrievalmaster datadefinitionshttps://predictes.com/blog/a-release-is-the-code-the-prompt-the-model-and-the-indexhttps://predictes.com/blog/a-release-is-the-code-the-prompt-the-model-and-the-indexIn an AI system, behaviour changes when the prompt, the model version or the retrieval index changes. Release and rollback have to treat them as one versioned unit.Thu, 30 Jul 2026 00:00:00 GMTrelease managementrollbackLLMdeploymenthttps://predictes.com/blog/the-answer-was-wrong-because-the-document-was-oldhttps://predictes.com/blog/the-answer-was-wrong-because-the-document-was-oldFreshness is the quietest failure in an AI system. Nothing errors, nothing alerts, and the answers are perfectly consistent with a version of reality that has moved on.Fri, 24 Jul 2026 00:00:00 GMTdata qualityfreshnesspipelinesmonitoringhttps://predictes.com/blog/the-ai-feature-nobody-dares-to-changehttps://predictes.com/blog/the-ai-feature-nobody-dares-to-changeA live AI feature with unmeasured quality becomes a liability the day its model is retired or its bill grows. What taking it over involves, and what to expect from it.Sat, 18 Jul 2026 00:00:00 GMTlegacy systemsevaluationmodel migrationriskhttps://predictes.com/blog/the-ai-policy-people-actually-followhttps://predictes.com/blog/the-ai-policy-people-actually-followMost AI policies are written to satisfy a risk register and are ignored within a month. The version that works is shorter, more permissive and considerably more useful.Fri, 17 Jul 2026 00:00:00 GMTAI policygovernancetrainingshadow AIhttps://predictes.com/blog/an-ontology-needs-a-decision-maker-more-than-a-modellerhttps://predictes.com/blog/an-ontology-needs-a-decision-maker-more-than-a-modellerInterviews, modelling and implementation can be bought. Deciding which definition of customer each system means cannot, and that is the real cost of the project.Thu, 16 Jul 2026 00:00:00 GMTontologydefinitionsdata governancedecision makinghttps://predictes.com/blog/spec-driven-development-does-not-buy-a-smaller-teamhttps://predictes.com/blog/spec-driven-development-does-not-buy-a-smaller-teamThe budget case for agent-written code is often a headcount case. What actually changes is where engineering time goes, and which skills become scarce.Mon, 13 Jul 2026 00:00:00 GMTspec-drivenengineering teamsbusiness caseaccountabilityhttps://predictes.com/blog/the-system-went-live-nobody-changed-who-does-whathttps://predictes.com/blog/the-system-went-live-nobody-changed-who-does-whatAn AI system changes the work around it: who checks, who takes the exceptions, who decides. If nobody redesigns that, the old process carries on beside it.Sat, 11 Jul 2026 00:00:00 GMToperating modelrolesgovernancechange managementhttps://predictes.com/blog/ai-systems-do-not-crash-they-drifthttps://predictes.com/blog/ai-systems-do-not-crash-they-driftConventional monitoring watches for things that stop. The failure mode that matters here produces no error, no alert and perfectly formatted answers that are quietly worse.Fri, 10 Jul 2026 00:00:00 GMTmonitoringdriftSLAoperationshttps://predictes.com/blog/cost-per-task-and-response-time-are-product-decisionshttps://predictes.com/blog/cost-per-task-and-response-time-are-product-decisionsModel choice, retrieval depth and extra checks each buy quality with money and seconds. Left to engineering, those trades still get made, just never on purpose.Tue, 07 Jul 2026 00:00:00 GMTunit economicslatencyproduct managementLLMhttps://predictes.com/blog/where-the-automation-stops-somebody-has-to-starthttps://predictes.com/blog/where-the-automation-stops-somebody-has-to-startIn an automated process the exception path decides the result: what reaches a person, who that is when nobody is on shift, and how the decision gets back in.Sat, 04 Jul 2026 00:00:00 GMTexceptionshand-offhyperautomationunattended operationhttps://predictes.com/blog/go-and-watch-a-shift-firsthttps://predictes.com/blog/go-and-watch-a-shift-firstField systems are rejected for reasons that never appear in a requirements document, and every one of them is visible within two hours of standing where the work happens.Fri, 03 Jul 2026 00:00:00 GMTfield deploymentrequirementsadoptionoperationshttps://predictes.com/blog/your-design-team-and-its-first-ai-producthttps://predictes.com/blog/your-design-team-and-its-first-ai-productAn in-house design team knows your users and product. What it usually lacks is a language for software that is sometimes wrong, and that is quicker to add than to hire.Mon, 29 Jun 2026 00:00:00 GMTdesign teamsAI interfacesknowledge transfertrusthttps://predictes.com/blog/third-party-connectors-are-your-agents-supply-chainhttps://predictes.com/blog/third-party-connectors-are-your-agents-supply-chainA ready-made connector brings code that runs with your credentials and tool descriptions your model reads as instructions. It deserves dependency discipline.Sun, 28 Jun 2026 00:00:00 GMTAI agentsconnectorssupply chainMCPhttps://predictes.com/blog/what-to-ask-a-data-engineering-providerhttps://predictes.com/blog/what-to-ask-a-data-engineering-providerProposals for AI data work look alike until the first month. The questions that tell you, before signing, whether a provider builds trust in your data or a platform.Sat, 27 Jun 2026 00:00:00 GMTvendor selectiondata qualityprocurementhttps://predictes.com/blog/when-agents-write-the-code-review-is-the-bottleneckhttps://predictes.com/blog/when-agents-write-the-code-review-is-the-bottleneckGeneration got ten times faster and review did not. Teams that measure only the first half accumulate under-reviewed code, and the debt arrives all at once.Fri, 26 Jun 2026 00:00:00 GMTAI codingcode reviewspec-drivenengineering practicehttps://predictes.com/blog/delivery-metrics-that-lie-once-agents-write-the-codehttps://predictes.com/blog/delivery-metrics-that-lie-once-agents-write-the-codePull requests, velocity and lines of code all rise when agents write the code. None of them says whether delivery improved. A few measures still do.Wed, 24 Jun 2026 00:00:00 GMTmetricsDORAAI codingengineering leadershiphttps://predictes.com/blog/an-agents-business-case-is-a-cost-per-task-not-an-autonomy-ratehttps://predictes.com/blog/an-agents-business-case-is-a-cost-per-task-not-an-autonomy-rateAgent proposals tend to promise a share of work done without people. The number that decides the investment is the full cost of a finished task, people included.Tue, 23 Jun 2026 00:00:00 GMTAI agentsbusiness casecostsapprovalshttps://predictes.com/blog/ai-strategy-proposal-or-sales-documenthttps://predictes.com/blog/ai-strategy-proposal-or-sales-documentStrategy proposals look alike on paper. A handful of questions about pricing, interviews and what the provider sells next separate the useful ones.Sun, 21 Jun 2026 00:00:00 GMTAI strategysupplier selectionproposalsprocurementhttps://predictes.com/blog/the-night-shift-knows-things-nobody-wrote-downhttps://predictes.com/blog/the-night-shift-knows-things-nobody-wrote-downUnattended production is rarely blocked by robotics. It is blocked by the judgement three experienced people carry, which exists nowhere a system can read.Fri, 19 Jun 2026 00:00:00 GMTdark factorytacit knowledgecontextunattended operationhttps://predictes.com/blog/has-your-ai-provider-ever-stood-on-a-shop-floorhttps://predictes.com/blog/has-your-ai-provider-ever-stood-on-a-shop-floorField AI projects fail on connectivity, gloves and devices nobody manages. A handful of questions in the first meeting shows whether a provider knows that.Mon, 08 Jun 2026 00:00:00 GMTfield deploymentvendor selectionoffline-firstoperationshttps://predictes.com/blog/who-runs-your-ai-system-the-day-after-go-livehttps://predictes.com/blog/who-runs-your-ai-system-the-day-after-go-liveThe project budget ends at go-live, which is exactly when the cost and the risk begin. How to decide who should run an AI system, and what each option asks of you.Sat, 06 Jun 2026 00:00:00 GMToperationshandoverownershiphttps://predictes.com/blog/the-business-case-for-unattended-hours-is-written-in-stoppageshttps://predictes.com/blog/the-business-case-for-unattended-hours-is-written-in-stoppagesLights-out cases are usually built on machine capacity times hours. What decides the return is how often the operation halts with nobody there to restart it.Wed, 03 Jun 2026 00:00:00 GMTdark factorybusiness caseunattended operationexceptionshttps://predictes.com/blog/how-to-buy-ai-training-that-changes-somethinghttps://predictes.com/blog/how-to-buy-ai-training-that-changes-somethingMost AI training is judged by how much people enjoyed it. The questions worth asking a provider are about what exists the week after the session.Tue, 02 Jun 2026 00:00:00 GMTtrainingworkshopsprocurementcapabilityhttps://predictes.com/blog/an-ai-roadmap-should-promise-quality-bars-not-featureshttps://predictes.com/blog/an-ai-roadmap-should-promise-quality-bars-not-featuresFeature roadmaps mislead in AI products, because every feature exists by week two. What is uncertain, and worth planning around, is when it becomes good enough.Mon, 01 Jun 2026 00:00:00 GMTroadmapsproduct managementevaluationdefinition of donehttps://predictes.com/blog/a-model-that-clicks-is-still-a-screen-scraperhttps://predictes.com/blog/a-model-that-clicks-is-still-a-screen-scraperRPA bots break because they are bound to screens and templates. A model fixes the reading; driving the same screens with it keeps the brittle half and makes it quiet.Thu, 28 May 2026 00:00:00 GMTRPAintegrationhyperautomationdocumentshttps://predictes.com/blog/whose-credentials-is-your-agent-usinghttps://predictes.com/blog/whose-credentials-is-your-agent-usingMost agents run on a person's token or a shared service account. Both leave the agent's reach undesigned and make the access log say something untrue.Sat, 23 May 2026 00:00:00 GMTAI agentsidentityleast privilegeaccess managementhttps://predictes.com/blog/the-demo-is-the-cheap-parthttps://predictes.com/blog/the-demo-is-the-cheap-partAn AI prototype is quick to build and slow to finish. What the remaining work consists of, how to fund it in stages, and what to ask before the budget is approved.Tue, 19 May 2026 00:00:00 GMTprototypesproductionbudgetingAI applicationshttps://predictes.com/blog/to-a-coding-agent-ci-is-a-target-not-a-checkhttps://predictes.com/blog/to-a-coding-agent-ci-is-a-target-not-a-checkAn agent iterates until the pipeline turns green and finds the shortest way there. CI gates for generated changes have to be designed with that in mind.Mon, 18 May 2026 00:00:00 GMTCIAI codingtestingspec-drivenhttps://predictes.com/blog/when-you-need-an-ontology-and-when-it-is-overheadhttps://predictes.com/blog/when-you-need-an-ontology-and-when-it-is-overheadAn ontology can be the most load-bearing part of an AI programme, or pure overhead. The signals that tell you which, before anybody draws a diagram.Sat, 16 May 2026 00:00:00 GMTontologymaster datasystem consolidationhttps://predictes.com/blog/before-funding-spec-driven-development-sort-your-backloghttps://predictes.com/blog/before-funding-spec-driven-development-sort-your-backlogSpec-driven development makes some work much cheaper and some work slower. Which of the two dominates your backlog decides the business case.Thu, 14 May 2026 00:00:00 GMTspec-drivenAI codingbusiness casebackloghttps://predictes.com/blog/the-most-valuable-decision-is-what-to-stophttps://predictes.com/blog/the-most-valuable-decision-is-what-to-stopAI programmes rarely fail on the build. They fail because they share six people with four other programmes, and nobody decided which ones would wait.Mon, 11 May 2026 00:00:00 GMTtransformationportfoliochange capacityprioritisationhttps://predictes.com/blog/the-ai-feature-works-nobody-uses-ithttps://predictes.com/blog/the-ai-feature-works-nobody-uses-itWhen an AI feature is technically sound and usage is flat, the instinct is to buy a better model. The problem is usually the interaction, and it is cheaper to fix.Wed, 29 Apr 2026 00:00:00 GMTadoptionAI interfacestrustproduct designhttps://predictes.com/blog/data-work-gets-funded-for-answers-not-platformshttps://predictes.com/blog/data-work-gets-funded-for-answers-not-platformsA request for a data platform loses to every other line in the budget. A request to stop one AI system from being wrong usually wins, if it is costed honestly.Mon, 27 Apr 2026 00:00:00 GMTbusiness casedata qualitybudgetinghttps://predictes.com/blog/in-an-ai-product-feasibility-belongs-in-discoveryhttps://predictes.com/blog/in-an-ai-product-feasibility-belongs-in-discoveryClassic product discovery asks whether people want something and assumes it can be built. With AI, whether it works well enough on your data is the open question.Sun, 26 Apr 2026 00:00:00 GMTproduct discoveryproduct managementprototypingfeasibilityhttps://predictes.com/blog/your-first-agent-should-do-the-boring-jobhttps://predictes.com/blog/your-first-agent-should-do-the-boring-jobThe first agent most companies want is the impressive one. The one that reaches production is narrow, frequent, checkable and dull, and it earns the right to grow.Fri, 24 Apr 2026 00:00:00 GMTAI agentsuse casesprioritisationbusiness casehttps://predictes.com/blog/before-you-buy-an-ai-strategyhttps://predictes.com/blog/before-you-buy-an-ai-strategyAn AI strategy solves one kind of problem well and three others badly. Which one you have decides whether to buy it, and what it should cost you.Wed, 22 Apr 2026 00:00:00 GMTAI strategyprocurementchange capacitydecision makinghttps://predictes.com/blog/the-model-proposes-the-code-decideshttps://predictes.com/blog/the-model-proposes-the-code-decidesHyperautomation works when each step goes to the right worker: the model reads and drafts, deterministic code checks and commits, and people take what is left.Tue, 21 Apr 2026 00:00:00 GMThyperautomationvalidationprocess designAI agentshttps://predictes.com/blog/everything-your-agent-reads-can-give-it-ordershttps://predictes.com/blog/everything-your-agent-reads-can-give-it-ordersPrompt injection rarely arrives through the chat box. It arrives in the email, ticket or web page the agent was asked to read, and the defence is structural.Fri, 17 Apr 2026 00:00:00 GMTAI agentsprompt injectiondata exfiltrationarchitecturehttps://predictes.com/blog/if-the-spec-is-the-source-of-truth-it-needs-a-lifecyclehttps://predictes.com/blog/if-the-spec-is-the-source-of-truth-it-needs-a-lifecycleWithin a quarter the code and the specs disagree. A specification stays the source of truth only if every change, hotfixes included, goes through it first.Sun, 12 Apr 2026 00:00:00 GMTspec-drivenspecificationsAI codingengineering practice