That’s most of what’s happening in business right now.

There are many signals waiting to be deciphered.

Anything familiar, or known enables us to respond faster than our awareness of our actions and the situation.

Stability makes prediction simpler. The laws of physics don’t change, but every once in a while new discoveries force us to revisit our understanding of the principles.

In business, the laws of supply and demand are not the only meaningful economic principles. Utility theory, the usefulness of something informs us of its value. Value can be further enhanced by distinction in features or form that makes it more appealing to some but not all.

The capability of an object may not change, and the capacity of a market may also remain stable. Still bolting on design, with new features/flourishes can add value that alters supply and trigger more demand.

The Interface versus the Infrastructure

Humans move physically quicker then their ability to understand and learn. Organizations/institutions move even slower. Let’s consider what this means in terms of raw worker, buyer and business entity capabilities, and the role of design.

It’s why organizations/institutions lay down infrastructure, and establish dedicated through lanes. However the consensus among parties necessary to build and fund them takes considerable time and effort. With consensus, the actual construction often goes fast.

Note, this is why we imagine the private sector to be more efficient. Their plans and execution can reach internal agreement seemingly faster than creating public consensus necessary to move public initiatives forward to fruition.

The design of an interface to look familiar goes a long way in easing access and adoption of new infrastructure and systems. But it also obscures the extent of greater capabilities that reside in the newly accessed infrastructure.

The lament that most people are using AI for simple tasks misses the benefits that other tools and models had already embedded. I was building predictive models in the early 1980s to enhance the accuracy of on-air elections forecasting. I then contributed to the scaling of the credit card businesses by derisking their offers, and optimizing their limited resources by accurately targeting outbound collections calls. The Genetic algorithms and machine learning tools were constrained by computation resources.

It was George Box, the British statistician, whose comment about all models are wrong but some are useful seems appropos. He reminds us that reality is not really simple, but incredibly complex.

In the last few years, the high tech service organizations’ appetite for processing and analysis was absorbing new capacity faster than the suppliers could deliver. The computational giants had to reinvent the structures that enabled AI’s near-frontier capability. Some hyped AI while others steady integration quietly absorbed the benefits.

Organizations already building sophisticated prediction models with multi-variate processing gained even further benefit. As AI firms competed to release improvements faster and offer more efficient development tools, they gained the same capabilities, but delivered faster and cheaper.

The problem, however, is not in the absorption of the new models, but in your infrastructure which you already optimized using simpler automation. Bolting on AI doesn’t deliver greater scalable capabilities or value. Particularly true if it’s availability isn’t nearly as stable or cheap as what you already own and operate.

Stability versus Transformation

The difference between growing capabilities and growing the capacity to absorb/utilize them is what creates bottlenecks. In economics if supply is the capability, then demand is the activated use case. Establishing and maintaining equilibrium proves difficult for several reasons.

1. The developmental path of AI models has taken decades and considerably more investment resources, than the recent period of returns has recouped.
2. The demand fueled greater competition in AI development and that competition significantly constrained the parallel investment needed to deliver, stabilize availability, and let others absorb these capabilities.
3. Finance likes assurance against risk, and expects predictability that stable environments can deliver. The time needed to design, assemble and transform materials into infrastructure that produces returns constrains capital. Without upfront assurance of stable demand, the cost of that capital may exceed the projected returns.

Google’s narrow search interface had been steadily advancing its capabilities without the fanfare. OpenAI saw an opportunity to compete and launched itself using the familiar interface of search, but with no workflow or utility. It’s that slow build out that has proven highly disruptive.

All of the existing AI tools offer real time processing. Other industries have found this transformation daunting , requiring wider adjustments to workflow, data, incentives, and trust. At present, changes in the interfaces offer that promise without touching the underlying processes. It frees the user in real time, but the data awaits later queued-up batch run.

Determining present needs to put present capability to use is a bit of a fools’ errand, as it strips out the slack and redundancy a business actually needs to grow. A model made more widely aware and responsive can flag risk sooner; that part is fast. But the mitigation? The actions to prevent or respond to what the model flags takes more serious thought, and physical coordination that can’t move at the same speed.

Compliant, structured systems offer utility and proven capabilities that conform to needs established in a previous period. Systems of record are the infrastructure, that give institutions stability, reliability, and predictability within known boundaries.

Have you ever experienced an overnight software update to a device and found another function impacted? That’s the nature of complexity; but it’s also the nature of linear thinking that makes it hard to consider and sufficiently evaluate simultaneous activities. Our brain doesn’t do that. It prioritizes sensory signal processes, makes other functionality redundant, and preserves extra capacity.

The business world, by contrast, doesn’t reward redundancy. It favors the elimination of excess materials, processes, and personnel.

The Dashboard Is Interface Too

The summary isn’t the information, and dashboards though dynamic don’t change which or how the decisions in the underlying systems get made.

We don’t have to imagine the experiences. The Agile project management, I’ve taught and managed make clear that organizationally it’s more effective and efficient to let people closest to the action make the decision. Not autonomy for its own sake, but real veto and adjustment authority staying with the person closest to the consequence.

Transferring calls to supervisors has added cost, and delayed resolution for countless customer encounters. Dashboards for managers and supervisory teams track aggregate system activity well. The value tradeoff for gains in efficiency costs frontline engagement among staff and customers.

Call it infrastructure and the debate stops there, as if a datacenter and a re-imagined workflow were the same investment. They’re not. Capital flowing into the physical buildings that hold the models is diverting away from the work that keeps your business meeting changing needs.

Some signals are worth decoding in the moment, by the person who heard them first. A dashboard, like a half-turned head on a bicyle, only ever gets you the outline.