Kinetic Information: properties needed in an AI-mediated world
Showing how the PODIUMS framework for kinetic information looks like in an industrial setting.
In this article, I’d like to elaborate on the PODIUMS framework for kinetic information and examine what kinetic information looks like in an industrial setting.
The definition of kinesis is movement in response to a stimulus. Potential energy is stored energy; kinetic energy is energy in motion. Applied to information, kinesis is content and data that is retrievable in response to a call by users via some sort of system.
Properties of kinetic information
The concept behind kinetic information allows for the following properties.
Profiled: personalised and automatically directed to the right audience according to context.
Offered: rather than simply stored, content objects are assembled and delivered when called by downstream systems, not just when a user happens to search.
Dynamic: updated automatically in response to new data, usage metrics, and content needs.
Independent: usable across contexts without editorial rework or reformatting.
Ubiquitous: always ready for use, online, searchable, and findable wherever a user might look.
Molecular: stored as independently-usable modular units of microcontent ready to be delivered as Content as a Service (CaaS).
Spontaneous: information assembly triggered by context in real time, delivered without human intervention.
Kinetic information in action
Let’s look at a theoretical company. This fictitious company is fairly representative of many industrial manufacturers. Let’s call them ACME.
ACME makes IoT-enabled industrial pumps that get used across manufacturing, water treatment, and oil & gas sites worldwide. In this company, a range of people, from field technicians, in-house support engineers, to distributor partners, and an AI troubleshooting assistant all need the same technical content, but none of them need it in the same form, at the same time, or delivered through the same channel.
Traditionally, this content would be published in a static manual, one PDF per pump model which would get updated on a release cycle. Users search through the documentation manually. A kinetic information architecture replaces that with content that behaves more like infrastructure than like a document.
Let’s look at how each of the properties of kinetic information are used to improve the customer experience and optimise the operating model for content production.
Profiled
Information is profiled for each type of product and personalised for each type of user. So when a technician uses a handheld scanner to scan a pump’s QR code, the system can identify information about the pump model, firmware version, and the technician’s location, and even the technician’s certification level. Content is filtered and directed automatically: a Level 1 technician in Germany can get the German-language basic diagnostic module; a Level 3 technician in Texas can get the advanced electrical isolation procedure, because their credentials clear them for delivery.
Offered
None of the content needs to be published. When a pump’s sensors report an anomaly, the diagnostic system pulls the content, aggregates it into a troubleshooting module, and the relevant module is pushed to the technician’s device. The content can be assembled and delivered in real time, triggered by the downstream need rather than search behaviour.
Dynamic
When ACME’s engineering team issues a firmware update that changes a calibration threshold in a sensor, the module storing that specification updates once, centrally, in the database. Every downstream surface: the technician app, distributor portal, AI assistant, printed job sheet generator, and so on, automatically accesses the current version. In a different dynamic process, usage data can be fed back to authors. If technicians repeatedly abandon a module part way through, that module can be flagged for editorial review.
Independent
Each module is self-contained. That means that the a “hydraulic seal replacement” module can be dropped into a technician’s repair workflow, a distributor’s training course, and a warranty claim form without the need to rewrite or reformat to adapt to a new context. The module carries its own meaning and can be used, regardless of where it gets delivered.
Ubiquitous
The same modules are findable and usable everywhere a need might arise: in the field technician’s app, in the distributor’s parts-ordering portal, in the customer-facing knowledge base, and inside the AI assistant’s retrieval layer. A technician doesn’t need to know which “manual” to open because the relevant module surfaces according to the context.
Molecular
Instead of a long-form manual, ACME’s content is broken into atomic information objects: one object per error code, one object per torque spec, one object per safety lockout procedure, one object per part number cross-reference, and so on. Each object is tagged with metadata, such as pump model, firmware version, region, hazard class, and certification level required. Objects are stored in a content repository as independent objects which can be exposed via an API as Content as a Service. No unit depends on the others to be understood or delivered.
Spontaneous
This is where the system becomes agentic rather than simply automated. When the AI troubleshooting assistant is asked a question such as “why is sensor 5A showing a pressure fault,” there’s no need to run a keyword search against a number of documents in a repository. The agent queries a knowledge graph that assembles the relevant objects into a module. A single contextual answer, with the right error code definition, the current firmware’s calibration spec, the site’s maintenance history, and so on gets assembled in real time and offered up without human intervention. The graph acts as the deterministic retrieval layer underneath the assembly. Content debt in any one information object, such as an outdated spec or even a broken cross-reference, shows up immediately as an error.
This example is but one way that the principles could play out within a company. Ever week, I see a post or two where someone discusses how they are optimising their operating model for production of their product information.
Note: In the interest of transparency, I used AI to generate a partial draft, which I then shaped into a proper article.