Data Science and Engineering Analytics Technical Section

 View Only

  • 1.  You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 20 days ago

    Noam Chomsky once offered a deceptively simple image. Imagine yourself in a luxurious car, moving along the most fascinating road you have ever seen. The engineering is flawless. The ride is effortless. The scenery commands your full attention. And then — the realization: you are not the one driving.

    Someone else is.

    I have been thinking about this image in relation to something far larger than any single vehicle. When I visualize the extraordinary architectures of modern AI, quantum computation, and networked infrastructure, Chomsky's image becomes almost unbearably precise. We are seated in a cabin of astonishing sophistication — and we have systematically confused the luxury of the seat with the power of the wheel.

    The seduction of the asset

    The dominant framework for evaluating technological power still centers on ownership. Organizations acquire a model license, deploy a system, and declare themselves technologically sovereign. Strategy consultants produce capability assessments that catalogue what has been purchased. Investors evaluate portfolios by what has been acquired.

    This framework is not wrong exactly. It is dangerously incomplete.

    What ownership frameworks cannot see is the distinction between a static asset and a living system. A sophisticated AI deployment is not a building you own or a patent you hold. It is a dynamic, dependency-laden organism. It requires computational infrastructure that belongs to someone else. It draws on data pipelines whose provenance you may not fully trace. It operates within parameter spaces that can be altered — often contractually, sometimes unilaterally — without your consent.

    You hold the deed. Someone else holds the keys.

    The silo problem no one admits

    Here is where the danger compounds.

    Most organizations have encountered the four universal questions of resilience — the same four questions that engineers call fault tolerance, that physicians call homeostasis, that military strategists call command continuity, that ecologists call adaptive capacity. How do we maintain continuity? How do we create redundancy? How do we avoid single points of failure? How do we remain functional under stress?

    And most organizations have built genuine answers. Within their own domain.

    The IT security team answered all four. The supply chain function answered all four. Finance answered all four. Each framework would satisfy a rigorous audit.

    Then a real disruption arrives — and the cascade begins crossing layers.

    A SCADA breach becomes a production halt. A production halt becomes a payroll crisis. A payroll crisis becomes an institutional authority vacuum. The failure doesn't respect the boundaries between functional silos. It travels through them as if they weren't there.

    This is what Aquarian Systematic Resilience identifies as the defining blind spot of conventional strategic analysis: a framework that only asks the four questions within its own domain answers them well — right until the moment the failure arrives from outside that domain. Which, in a real disruption, it almost always does.

    The concept of Flow Sovereignty

    This is precisely why ASR introduces the concept of Flow Sovereignty as its corrective.

    Flow Sovereignty does not ask who owns the asset. It asks: who commands the flows — of power, data, and institutional legitimacy — passing through the system? And it demands that this question be answered simultaneously across what the framework calls the Six Sovereign Layers: Physical Infrastructure (the Muscle), Digital Architecture (the Nerves), Human Capital (the Will), Institutional Authority (the Permission), Strategic Geography (the Horizon), and Financial Liquidity (the Velocity).

    The organization that can trace, redirect, and stabilize those flows under conditions of shock is genuinely sovereign. The organization that cannot is a well-appointed passenger.

    The distinction carries a second, equally critical dimension: the difference between reactive and architectural resilience.

    Conventional resilience is assembled during the crisis. Crisis detected. Response constructed. Central command required to authorize. Improvisation under maximum pressure. The 72-hour window — the critical period immediately following any major disruption, during which cascades can still be compressed rather than allowed to become irreversible — consumed by coordination overhead.

    ASR doctrine inverts this entirely. The answer is built before the event. Protocols pre-positioned. Command distributed and pre-authorized, so degraded-condition procedures activate autonomously — without waiting for a central authority that may itself be compromised or unreachable. The 72-hour window is covered by design, not consumed by it.

    The deeper anxiety

    What gives Chomsky's image its particular power is not the vulnerability it describes. It is the comfort it depicts.

    The car is luxurious. The road is genuinely fascinating. The experience of being in the vehicle is, by every immediate measure, excellent. This is why the discovery of non-control is so destabilizing — not because the ride was unpleasant, but because the ride was so thoroughly convincing.

    Modern technology is extraordinarily good at this. The interfaces are extraordinary. The outputs are remarkable. The sense of capability is real, even when the control underlying that capability is not.

    And this is where the ASR framework's third contribution lands with particular force. Resilience has historically been treated as a cost center — necessary, like insurance, but not investable. ASR proposes three instruments to change this: the Net Fragility Matrix, which scores every node across six dimensions and identifies every Pivot Domino whose failure cascades into total system collapse; the 72-Hour Survival Window, which establishes the minimum engineering benchmark for autonomous operation without external command, supply, or connectivity; and the Sentinel Risk Value, a capital protection ROI formula. The documented case: a $1.4 million investment in pre-positioned resilience architecture against $111 million in avoided loss — a return of 69 to 79 times.

    Continuity, in this framing, is not a defensive posture. It is an investment thesis.

    The work — the genuinely difficult strategic work — is to separate the experience of power from its substance. To ask, with uncomfortable rigor: at what points in this system am I actually sovereign, and at what points am I simply well-seated?

    Chomsky gave us the image. The question is what we are prepared to do with it.

    Aquarian Systematic Resilience — connect@syedabidshah.com | syedabidshah.com

    image


    ------------------------------
    Syed Abid Shah
    Founder Aquarian Systematic Resilience
    connect@syedabidshah.com
    +92 331 3330 188
    Karachi Pakistan / Dubai, UAE
    ------------------------------


  • 2.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 19 days ago
    This was a rather unique post, so worth trying to get a different perspective on the themes presented.

    My sense was it tries to frame a business problem which can only be addressed by a specific product that the author was associated with. Caveat emptor.

    Given the big crew change, most folks seeing the post might not be aware that there is some historic track record in applying tools in E&P organizations to understand somewhat related issues. Top of my mind would be :

    The Zachman Framework: Enterprise architects have tried to use the Zachman framework to understand an enterprise and its information systems interactions with various customers as an integrated whole.

    Bowtie Analysis of Risks: Bowtie analysis allows an organization to assess a risk and look at both the causes and protections already in place against the consequences and mitigations.

    Probably multiple ways to combine both types of approaches.

    Does AI deployment pose unique and different risks than the wave of self-service analytics that preceded it? Yes- in terms of the pace of updates to the technology and its capabilities itself. But in terms of the quality assurance and controls on the inputs and outputs that should have been there before, perhaps not.

    The big risk that we had on analytics that was difficult to mitigate was that service companies acquire data for multiple operators and were using the data collection to improve their own analytics to create new value-added services over the top which they could then sell to the market- and potentially create opportunities for themselves through disintermediation by becoming a form of operator themselves. Seems entirely likely that the same risk is even higher in the AI domain – and even harder to constrain irrespective of your risk mitigation approach.




  • 3.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 19 days ago
    Edited by Syed Shah 19 days ago

    David, thank you for this. It is genuinely the kind of engagement that makes a discussion worthwhile.

    On the caveat emptor point, fair, and worth addressing directly. Aquarian Systematic Resilience (ASR) is a framework I have developed, and I have an obvious interest in it. The real test, however, is whether the underlying argument holds regardless of who is making it.

    I agree that Zachman and Bowtie have established track records and answer important questions. My view is that they address different analytical problems. Zachman helps map and classify systems. Bowtie traces causes, controls, and consequences for defined risks. Both are valuable.

    Where ASR is attempting to contribute is in understanding what happens when disruptions cross domains that may each be individually well designed. A digital disruption becomes an operational interruption, which becomes a financial issue, which begins affecting institutional decision making. The cascade is not a property of any single node; it emerges from the interactions between them.

    Your observation regarding service companies accumulating data and progressively becoming operators is particularly interesting. To me, that illustrates a shift in control, influence, and dependency that extends beyond technology alone. It raises questions about who ultimately controls critical flows of data, authority, and value, the issue ASR attempts to examine through the concept of Flow Sovereignty.

    I would be interested in your perspective on whether the frameworks you mentioned have been successfully applied to modelling such cross domain cascades, or whether in practice they tend to remain largely domain specific.

    Thank you again for the thoughtful challenge.

    ------------------------------
    Syed Abid Shah
    Founder Aquarian Systematic Resilience
    connect@syedabidshah.com
    +92 331 3330 188
    Karachi Pakistan / Dubai, UAE
    ------------------------------

    Original Message


  • 4.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 18 days ago

    Syed

    I was holding off on responding to you to give time for others to weigh in on an important topic.  But I will get back on the soap box to make a few more comments.

    This morning I typed into Google a query "risks to business continuity from ai" twice and got two different answers.  The second one was

    "AI introduces highly dynamic, silent, and systemic risks to business continuity that traditional Disaster Recovery (DR) frameworks-built for simple server outages-are entirely unprepared to handle. When traditional IT fails, it leaves clear error logs; when AI fails, it often decays silently, hallucinates, or breaks critical automated workflows without warning."

    The first answer was completely different and emphasized how Business Impact Analysis needs to be updated for a long list of risks like agentic failures, model drift, user developed AI solutions, dependency of workflows on AI, data integrity issues, and over automating work to remove human judgement in sensitive areas.

    My point – besides highlighting how the non-stationarity of AI answers could lead one to very different approaches to recognizing and dealing with its own risks- is that some of the risks are unique to workflows enabled by AI, and some are issues that were with us before when applying business analytics in E&P.  And my own experience from information related bow tie analysis is that recognizing and naming a risk is often much easier than implementing barriers and controls to minimize their impact & recover when they are triggered in an event. Disruption caused by an Ai supplier taking the knowledge gained across multiple companies and using it to launch its own competing business ventures could be one such risk.

    Is it entirely possible to do cross domain business process modeling that includes the linkage context to outside the enterprise organizations like suppliers, regulators, and customers- yes, but it takes time, tools and expertise to do so.  I only mentioned Zachman in terms of there being lots of easily accessible information on a framework that considers multiple dimensions of information moving through an enterprise.

    The ability to analyze cross domain linkages of issues and risks to business delivery is theoretically possible- the key question you need to ask is who is concerned enough with the potential risks and impacts to dedicate the resources to analyze the problem and implement changes.



    ------------------------------
    Dave Feineman
    ------------------------------

    Original Message


  • 5.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 17 days ago

    Very thoughtful follow-ups here. I think the observations about non-stationarity is important, but I would draw a distinction between variation in outputs and variation in underlying information.

    On your first point, yes, Google, SEO-driven search results, and even human experts can provide different answers depending on geography, personalization, context, and timing. The difference with generative AI is that the variability can occur even when the same user submits the same prompt minutes apart. That doesn't necessarily make AI unreliable, but it does mean organizations need governance frameworks that focus on repeatability, traceability, and decision accountability when AI is incorporated into business-critical workflows.

    To me... this is less an AI-specific problem than an extension of a longstanding challenge in information management: how do we distinguish between exploratory tools designed to generate possibilities and operational systems designed to support consistent decisions?

    Regarding weights, biases, and model sovereignty, my view is that organizations should think about these as strategic choices rather than purely technical ones.

    • Closed commercial models offer convenience, scale, and rapid innovation but require trust in external providers and their governance practices.

    • Open-source models and weights provide greater transparency, auditability, and control, particularly for sensitive industries and sovereign data environments.

    • Federated architectures or model meshes may ultimately become the preferred middle ground. Rather than relying on a single model, organizations can route tasks across specialized models, retain local control of sensitive data, and reduce concentration risk associated with any one provider.


    That said, open weights do not eliminate bias. They simply make the sources of bias more inspectable and, in some cases, more correctable. Every model reflects choices made in training data, architecture, tuning, and governance. The question is not whether bias exists, but again... whether it can be understood, monitored, and managed.  And who controls the weights? 

    Your point about service companies accumulating operational knowledge is particularly relevant here. The strategic issue may not be AI itself but the concentration of data, expertise, and decision-making influence. Whether that resides in a service company, cloud provider, model developer, or platform operator, the underlying business continuity question becomes one of dependency management and resilience.

    In that respect, I think your earlier observation about cross-domain cascades is exactly where the conversation should be focused. The most significant risks may emerge not from a model failure, but from the interactions among technology providers, operators, regulators, suppliers, and business processes that become increasingly interconnected over time.

    Best,

    Nick Robbins, P. Eng. 

    Research Director, Oil and Gas (Production & Facilities)

    Darcy Partners


    Original Message


  • 6.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 17 days ago

    Nick,

    the output-variation versus information-variation distinction is the most useful clarification this thread has produced, and it reframes my original point in a way I think is more accurate: the governance need isn't "AI is unpredictable" but "we lack the repeatability and decision-accountability architecture to know when variability matters and when it doesn't." That's a real correction, and I'll take it.

    Your weights spectrum (closed / open / federated) is the right way to think about this as a strategic rather than technical choice, and I'd push on one implication of it: a federated model-mesh architecture doesn't just reduce concentration risk at the provider level but it also distributes the governance burden, because you can no longer rely on a single point of inspection or accountability. You gain resilience against any one provider's failure or defection, but you trade it for a harder cross-model traceability problem. Worth naming as a cost, not just a benefit, of the federated path.

    On "who controls the weights", that's the question underneath everything we've been discussing, and I don't think it has a clean answer yet, which is itself informative. The honest position is that nobody currently has a doctrine for sovereign command over a model whose weights, training data provenance, and update cadence are controlled by a third party operating under a different jurisdiction's incentives. That's not a gap any single organization closes with better contracts. It's closer to the dependency-concentration problem you named at the end; except the "supplier" in this case is the model itself, and the thing being concentrated is interpretive authority over what counts as a correct decision.

    Which is where your closing point lands hardest for me: the risk isn't the model failing. It's the slow transfer of decision-making authority to whoever sits closest to the weights, accumulated quietly across thousands of small interactions, with no single failure event to mark when it happened. That's a harder thing to put in a Business Impact Analysis than a server outage, because there's no outage. There's just, one day, the realization that a meaningful share of operational judgment now lives somewhere you don't have visibility into.

    Genuinely valuable thread - appreciate both of you pushing this further than I took it initially.



    ------------------------------
    Syed Abid Shah
    Founder Aquarian Systematic Resilience
    connect@syedabidshah.com
    +92 331 3330 188
    Karachi Pakistan / Dubai, UAE
    ------------------------------

    Original Message


  • 7.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 17 days ago

    David,

    I think your final observation is the one that matters most, and I want to answer it directly rather than around it: in my experience, the people who actually fund resilience work are not the ones who see the risk most clearly; they're the ones who have already paid the cost of not having it once, and have a number attached to that memory.

    Pre-event, resilience competes for budget against everything with a visible ROI. Post-event, it's the only thing anyone wants to talk about. The investment-framing matters because it's the only lens that makes the case before the event has happened, by translating "this could fail" into "this is what failing costs, and here is the multiple you get by paying now instead." That reframing doesn't eliminate your question about who is concerned enough,  it just narrows the audience to the CFOs and boards who already think in those terms, which is a smaller but more persuadable group than risk committees.

    On your bowtie point , agreed, and I'd push it further. Naming a risk is a one-time analytical act. Building the barrier is an ongoing operational commitment, and recovering from a triggered event is a third, distinct capability that often gets the least investment of the three because it's the hardest to rehearse without disrupting the business you're trying to protect. The AI supplier-disintermediation risk you raised earlier is a good test case: naming it is easy (I just did it in a LinkedIn post). Building a contractual or architectural barrier against it is much harder and rarely gets done, because it requires sacrificing some short-term integration convenience for long-term independence. Recovering from it once a supplier has already absorbed your data into a competing offering may not be possible at all, which is the uncomfortable case for treating some risks as something you architect against from day one, not something you plan to recover from later.

    Appreciate you sticking with this thread



    ------------------------------
    Syed Abid Shah
    Founder Aquarian Systematic Resilience
    connect@syedabidshah.com
    +92 331 3330 188
    Karachi Pakistan / Dubai, UAE
    ------------------------------

    Original Message


  • 8.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 12 days ago

    Syed, your statement about resilience "they're the ones who have already paid the cost of not having it" brought to mind an article about safety practices on seismic crews in Western Canada. Which leader has already paid the cost of not having?

    "Leaders who visit a workplace the day after an incident simply demonstrate their interest in incidents. Leaders who visit the workplace the day before an incident demonstrate their interest in safety. - Dr. Bruce Staley, "The Source", 2009-01, p30.



    ------------------------------
    William Dickson

    Dickson Intl. Geosciences(DIGs)

    Houston

    billd@digsgeo.com
    ------------------------------

    Original Message


  • 9.  RE: You're in the Passenger Seat: Sovereignty, Technology, and the Illusion of Control

    Posted 12 days ago

    William,

    That is an excellent observation, and a quotation I had not come across before. Thank you for sharing it.

    The distinction between leading before disruption rather than responding after it captures an essential principle of resilience. Once an incident occurs, the cost has already been paid. The real question is whether leaders invested in understanding and reducing the conditions that made it possible in the first place.

    That is very much how I think about resilience. It is less about responding well to failure than about continuously strengthening the architecture that makes failure less likely to cascade when it does occur.

    Thank you for adding this perspective to the discussion.



    ------------------------------
    Brig Syed Abid Shah (Ret.)
    Founder Aquarian Systematic Resilience
    connect@syedabidshah.com
    +92 331 3330 188
    Karachi Pakistan / Dubai, UAE
    ------------------------------

    Original Message