The Systems Lens

Nature Already Has the Math

Utilities do not manage isolated assets. They manage relationships inside the water cycle.

HA Hardeep Anand · 2026-09-21 · 9 min read

Start the morning after a hard rain. The storm is gone. The water is not.

It is in the streets. It is in the soil. It is in the collection system. It is in pump run time, plant loading, overtime, permit risk, flooded basements, customer calls, and tomorrow's capital argument.

A storm is not just rain. By morning, it has become a utility system telling the truth about itself.

Richard Feynman stood at a blackboard in his 1964 Messenger Lectures and made a point that still lands here. He was talking about nature, gravity, and mathematics. I see a lesson here for utilities and AI.

He showed how mathematical reasoning connects physical statements. A force directed toward the sun leads to equal areas swept out in equal times. You can describe each statement in words. The reasoning shows how they belong to each other.

Mathematics gives us that organized reasoning. It is how we connect timing, storage, pressure, flow, and consequence.

Utilities live in that world every day.

Water, wastewater, and stormwater systems follow those relationships before anyone opens a spreadsheet or runs a model.

Operators may not call it math at 2 a.m. when a pump is short cycling. They call it experience. They call it knowing the system. They call it hearing something wrong before the alarm proves it.

That judgment connects what the instrument reads with what the operator knows: elevation, demand, groundwater, pipe condition, maintenance history, and the people depending on the system.

Utilities do not manage isolated assets. They manage relationships inside the water cycle.

The Systems Lens

The Blackboard Lesson

Feynman gives a warning that every utility leader should recognize. A satisfying explanation still has to survive its consequences.

That is where weak explanations fail. In the lecture, he walks through a simple story for gravity. Imagine particles hitting objects from all directions. The sun blocks some of those particles, so the earth gets pushed toward the sun. It gives the mind a picture.

Then the picture breaks. If the earth is moving through those particles, more should hit it from the front than the back. That would create drag. The orbit would slow. The solar system would not behave the way it does.

That is the first utility lesson: test an explanation against the connected system before relying on it.

AI has the same problem. A model can explain a pump failure, a billing anomaly, or a compliance trend in fluent language. That does not mean it understood the relationships. It may have written a clean paragraph while missing the thing that matters.

The test is not, "Did it sound smart?" The test is, "Did it preserve the relationships that decide what happens next?"

Nature Already Runs as a System

The hydrologic cycle is the original operating system for water.

Rain falls. Some runs over land. Some enters soil. Some recharges groundwater. Some feeds streams. Some evaporates. Some is taken up by plants. Some returns to the atmosphere. Then it comes back again.

Nothing in that cycle is separate. Every movement has volume, timing, storage, loss, quality, temperature, direction, and consequence. The cycle is not a school diagram. It is the system utilities are built inside.

Figure 01

The water cycle is already a connected system

Hydrologic cycle and utility intervention A conceptual cycle: rain reaches land, infiltration recharges groundwater, runoff and baseflow feed streams. Utilities withdraw water, supply communities, collect wastewater and return treated flow. Evaporation returns water to the atmosphere. Atmosphere rain, snow, evaporation Land surface runoff and infiltration Groundwater storage and baseflow Streams source water and receiving water Utility systems treat, move, collect, return Community use service, trust, demand precipitation collection supply treated return flow evaporation runoff Water movement and utility intervention Rain and snow reach land. Two parallel routes feed streams: surface runoff, and infiltration followed by groundwater baseflow. Streams evaporate to the atmosphere. Utilities withdraw and treat source water, supply communities, collect wastewater, and return treated flow to receiving water. Side arrows close the evaporation, collection, and treated-return loops. Actual sources and return paths vary. Precipitation: rain and snow reach land Infiltration recharges groundwater Surface runoff reaches streams without passing through groundwater Groundwater baseflow feeds streams Evaporation returns water from streams to the atmosphere Withdrawal: streams provide source water for utility treatment Treated return flow goes from utility systems to receiving water Supply: treated water reaches community users Collection: community wastewater returns to utility systems AtmosphereRain and snow fall;evaporation returns water. Land surfaceRunoff goes to streams;infiltration enters soil. GroundwaterRecharge and storage;baseflow feeds streams. StreamsSource water for utilities;receiving water forrunoff and treated flow. Utility systemsWithdraw, treat, move,store and supply water.Collect wastewater;treat and return flow. Community useService, trust, demand;wastewater collection. Utilities intervene insidethe cycle. Actual sourcesand return pathways vary.

The utility is not outside nature. It is a public intervention inside nature's water movement. A simplified cycle, not a complete flow model; actual sources and return pathways vary. USGS: the water cycle.

We capture water, treat it, move it, store it, use it, collect it, clean it, discharge it, and then nature takes it back into motion.

This is why utility work is so hard to explain from the outside. Its responsibilities cross asset classes and departments, connecting natural assets, physical infrastructure, and people.

A Storm Is Never Just Weather

Take one ordinary event. A heavy storm crosses a service area overnight.

To a resident, it is rain. To a collection system operator, it may be inflow and infiltration. To a treatment plant, it may be hydraulic load. To a pump station, it may be run time and wear. To a compliance officer, it may be permit risk. To a finance director, it may be overtime, emergency repairs, and capital evidence. To a neighborhood, it may be flooded streets and trust lost.

Same storm. Different consequences. One connected system.

Figure 02

One storm becomes many utility consequences

Storm consequence chain A diagram showing a heavy storm moving through groundwater, runoff, sewer pipes, pump stations, treatment load, permit risk, public trust, operating cost, and capital planning. Heavy storm the visible event Wet ground high groundwater Street runoff flooding and inflow Sewer pipes inflow and infiltration Pump stations longer run time, wear Treatment plant hydraulic loading Permit risk overflow, notice, limits Cost evidence overtime and repairs renewal evidence Public trust flooding and calls confidence in service where combined sewers or inflow defects admit runoff One storm, branching utility consequences A heavy storm produces wet ground and street runoff in parallel. Groundwater can infiltrate defective sewer pipes. Runoff enters combined sewers or sanitary sewers only through inflow defects; separate storm drains do not normally flow to the wastewater plant. Sewer flow reaches pumps and treatment. Pump operation creates cost evidence, treatment loading creates possible permit risk, and street flooding affects public trust independently of the wastewater path. Rain can raise groundwater Rain generates street runoff Groundwater enters sewer pipes through infiltration defects Runoff enters only where combined sewers or inflow defects admit it Street flooding can affect public trust, independently of plant loading Added sewer flow reaches pump stations Pumped flow adds hydraulic loading at the treatment plant Longer pump operation contributes operating and renewal cost evidence Treatment loading can increase permit risk Heavy stormThe visible event WetgroundGroundwaterrises.Infiltrationthrough pipedefects addssewer flow. StreetrunoffFlooding;sewer entryonly throughcombinedsewers orinflow defects. Sewer pipesInflow and infiltrationadd to sewer flow. Pump stationsLonger run time;more wear. Treatment plantHydraulic loading;separate storm drainsnormally do not feed it. Permit riskPossible overflow,notice and limit issues. Cost evidenceOvertime and repairs;renewals and capitalplanning. Public trustFlooding, customer calls,confidence in service.

The rain is gone, but its consequences are still moving through the utility. Conceptual pathways: runoff can enter combined sewers or reach sanitary sewers through inflow defects. Separate storm drains do not normally flow to the wastewater plant. EPA: sewer systems.

This is the kind of relationship AI must learn to respect. If AI reads the work orders and says, "There were more pump station callouts after the storm," that is a summary. Useful, maybe. But not enough.

The better question is whether it can help people connect the callouts to rainfall, groundwater, upstream defects, pump age, past maintenance, permit exposure, neighborhood impact, and capital planning.

An Algae Bloom Is Not Just a Lake Problem

The same pattern shows up when the water is quiet.

An algae bloom can look like a surface problem. Green water. Odor. Taste complaints. A warning at the reservoir. A headline after a fish kill.

But the bloom is not a single event either. It may carry a history of nutrients, temperature, sunlight, residence time, runoff, upstream land use, low flow, treatment limits, and public communication. The fish kill may be the visible ending of a chain that started far away from the dead fish.

Respiration and decomposition can draw down dissolved oxygen enough to stress or kill fish. Not every bloom produces toxins, and not every fish kill comes from a bloom. Sampling and local conditions matter. EPA explains these pathways.

Figure 03

A fish kill can be the visible end of a longer chain

Algae bloom and fish kill chain A diagram linking heat, sunlight, nutrients, runoff, low flow, algae bloom, low dissolved oxygen, fish stress, treatment challenges, customer communication, trust, and response choices. Heat sunlight, temperature Nutrients runoff and upstream land use Low flow residence time Algae bloom source-water change Low oxygen respiration, decomposition Fish stress or kill a possible outcome Treatment strain taste, odor, sampling Public response sampling, communication source protection capital choices Bloom conditions and two consequence pathways Heat and sunlight, nutrients from runoff and upstream land use, and low flow with longer residence time are parallel influences on a bloom. A bloom can lower dissolved oxygen through respiration and decomposition, potentially stressing or killing fish. Separately, it can strain treatment through taste, odor and sampling needs. Both pathways inform public response: sampling, customer communication, source protection and capital choices. These are possible pathways, not a diagnosis; not every bloom produces toxins or explains a fish kill. Heat and sunlight can favor a bloom Nutrient inputs can favor a bloom Low flow and residence time can favor a bloom Respiration and decomposition can lower dissolved oxygen A bloom can separately strain treatment Low dissolved oxygen can stress or kill fish Fish stress or kill informs public response Treatment strain informs public response HeatSunlight and temperature NutrientsRunoff and upstreamland use Low flowLonger residence time Algae bloomVisible changein source water Low oxygenRespiration anddecomposition can useup dissolved oxygen. Fish stress or killA possible outcome;check local conditions. Treatment strainTaste and odor;sampling needs andtreatment limits. Public responseSampling; communicationwith customers for trust;source protection;capital choices. Possible paths, not a diagnosis.Not every bloom makes toxins.Not every fish kill is a bloom.

Nature does not hand us isolated incidents. It hands us relationships with consequences. These are possible pathways, not a diagnosis of a particular bloom or fish kill.

A weak answer says, "An algae bloom occurred and water quality declined." A useful answer asks what fed the bloom, what changed in the source water, which records show the sequence, which treatment decisions followed, which customers were affected, which agencies needed notice, and what the next storm or heat wave might do.

The Utility Test for AI

Before we use an AI answer, we need to know what holds it together.

Can it show the source record? Can it connect the asset to the event? Can it show who owns the next action? Can it separate evidence from assumption? Can it preserve the operator's judgment instead of flattening it into generic text? Can it explain what would be wrong if the explanation were wrong?

Figure 04

The answer is not trusted until the relationships survive

Fluent answer versus connected answer A split diagram comparing a fluent AI summary with a connected answer that shows sources, relationships, tested consequences, and human decision authority. Utility question What is happening? What should we check next? Fluent answer clean summary of visible records risk: sounds complete too early Connected answer source, context, consequence human decision authority Sources records used Relationships what connects Consequences what else follows Authority people decide A paragraph can be useful. It is not the same as reasoning. Two ways to answer a utility question The utility question branches into two alternatives, not successive stages: a fluent summary and a connected answer. A summary may help but can sound complete too early. A connected answer must independently expose sources, relationships, testable consequences and human authority. The four checks share a branching connector; they are not a causal chain. Validate records, separate evidence from assumption, preserve operator judgment and identify who owns the next action. This is an evaluation framework, not a measured product comparison. One possible response is a fluent summary An alternative is a connected answer Connected answer must expose its source records Connected answer must preserve relationships Connected answer must test consequences Connected answer must retain human decision authority Utility questionWhat is happening?What should wecheck next? Fluent answerClean summary ofvisible records.Risk: sounds completetoo early. A paragraphcan help; it does notestablish the reasoning. Connected answerSource, context,consequence and humandecision authority. SourcesShow the records used.Validate those records. RelationshipsConnect assets and events;separate evidencefrom assumption. ConsequencesWhat else should follow?What evidence supportsor weakens the answer? AuthorityPreserve operator judgment.Name the next action owner.People retain the decision. Check the answer before acting.An evaluation framework,not a product comparison.

A connected answer gives a utility more to check before acting. The sources still need validation, and people retain the decision. This is an evaluation framework, not a measured comparison of AI products.

If the answer says the likely cause is infiltration, what evidence would support that? What would weaken it? What inspection, sensor, map, complaint, or field note should we check next? Those checks make the reasoning accountable.

Operators Already Do This

Operators already know the system has memory. They know yesterday's rain can show up in today's flows. They know last month's valve work can explain this week's pressure complaint. They know a pump that technically works can still be telling you it is tired. They know an asset record can be officially complete and operationally useless.

The problem is that this intelligence often lives in scattered places. Some of it lives in supervisory control and data acquisition (SCADA). Some in geographic information systems (GIS). Some in computerized maintenance management systems (CMMS). Some in lab results. Some in billing. Some in hydraulic models. Some in permit reports. Some in the head of the person who has been answering calls for decades.

Then we ask AI to help.

If the data is fragmented, the AI can inherit the fragmentation. If the source record is weak, the answer can become weak. If the relationships are missing, the model may fill the silence with confidence.

That is why data readiness comes before AI readiness. The utility's reasoning depends on it.

Utilities Are Where Nature Becomes Public Responsibility

The hydrologic cycle is natural. Utilities make part of that cycle a public promise.

Safe water at the tap. Wastewater collected and treated. Stormwater moved away from homes and streets. Permits respected. Rates explained. Capital spent wisely. Service restored when something fails.

That is the civic layer.

Figure 05

The Systems Lens for utility AI

Systems Lens framework for utility AI A layered loop connecting nature, physics, infrastructure, people, records, AI, accountable decisions, and public trust. Nature moves rain, groundwater, rivers Physics translates flow, pressure, decay Assets intervene pipes, pumps, plants People interpret operators, crews, engineers Records connect SCADA, GIS, CMMS, lab AI must preserve source, context, consequence Accountable decisions repair, renew, respond, explain Public trust service, permits, rates, confidence community feedback human interventions The standard is not fluent answers. The standard is connected reasoning with evidence people can check. Connected reasoning with two feedback loops Nature moves, physics translates, assets intervene, people interpret, records connect and AI must preserve sources, context and consequences. People make accountable decisions that bear on public trust. Two separate feedback loops remain visible: human interventions return from decisions to nature on the left, and community feedback returns from public trust to people on the right. This conceptual framework shows responsibilities, not automatic causation or guaranteed trust. Natural movement has physical relationships Physics informs infrastructure intervention People interpret asset operation People connect observations through records Records provide sources and context for AI Inspectable evidence supports human decisions Accountable public service decisions bear on trust Human interventions return from accountable decisions to nature Community feedback returns from public trust to the people interpreting the system Nature movesRain, groundwater, rivers Physics translatesFlow, pressure, decay Assets intervenePipes, pumps, plants People interpretOperators, crews, engineers;community feedback. Records connectSCADA, GIS, CMMS,lab records AI must preserveSources, context,consequences and evidencepeople can inspect. AccountabledecisionsPeople repair, renew,respond and explain.Human interventionsreturn to nature. Public trustService, permits, rates,confidence; communityfeedback returns to people. Connected reasoning needsevidence people can check.Conceptual responsibilities;trust is not guaranteed.

AI is useful when it helps the utility keep nature, physics, infrastructure, records, people, and trust in the same frame. This conceptual loop shows responsibilities and feedback, not automatic causation or guaranteed trust.

This is why utility AI cannot be treated like a chatbot bolted onto records. It is entering a public system. The answer affects money, risk, compliance, and trust.

It may shape which pipe gets inspected, which pump gets replaced, which neighborhood receives attention, or which capital project gets defended. That means the answer has to show its work.

The Framework

Nature moves. Physics translates the movement. Infrastructure intervenes. People interpret. Records make it computable. AI must preserve the relationships. Decisions become accountable only when the chain can be inspected.

A pump record becomes useful when a crew can connect it to the work performed, the conditions observed, and the decision now in front of them.

Three Utility Examples

Consider three illustrative situations.

The pressure complaint

A resident reports low pressure. A weak AI answer summarizes the complaint and pulls nearby work orders. A useful answer asks what the complaint is connected to: recent demand, elevation, valve status, hydrant flushing, pump operation, tank level, main breaks, customer history, nearby construction, fire flow needs, meter data, and prior complaints.

The plant loading spike

The plant sees a hydraulic loading spike after a storm. A weak AI answer says flow increased after rainfall. Everyone already knew that. A useful answer helps trace the path: which basins responded first, which pump stations ran longer, which older pipe areas correlate with the spike, which manholes have past defects, which neighborhoods saw backups, and which capital projects already touch the same area.

The capital plan

A capital plan should not be a spreadsheet of assets waiting their turn. It should be a chain of evidence: service risk, failure history, hydraulic consequence, permit exposure, equity impact, repair feasibility, funding timing, and public explanation. AI can help only if it keeps that chain intact.

Connected Reasoning

Feynman's lesson comes back to the morning after the storm. The rain gauge, pump log, field note, and customer call each hold part of the explanation. We need to be able to trace the connections and test what they suggest.

For decisions like these, AI is useful when it helps people see the relationships, check the evidence, and make better decisions together.

Not fluent answers. Connected reasoning.

The Systems Lens

Practical takeaways

Email takeawaysShare on LinkedIn
  1. Ask where the AI answer came from, not only whether it sounds right.
  2. Ask what relationships it preserved across nature, assets, records, operations, finance, and trust.
  3. Ask what operator knowledge is missing before the answer can be used.
  4. Ask whether the output can be traced from event, to source record, to decision.
  5. Ask the Feynman question: if this explanation is true, what else should also be true?

Nature Already Has the Math
Hardeep Anand · The Systems Lens
https://hardeepanand.com/writing/nature-already-has-the-math/

APAS Consulting

For a utility exploring AI, the first practical step is not a demo. It is a readiness map: the decision that matters, the source records behind it, who owns them, what relationships are missing, and what evidence a trusted answer must show.

Hardeep Anand | APAS Consulting | apas.ai

A THOUGHTFUL RESPONSE LAYER

What does this raise for you?

This site does not need reaction counters or an open comment feed. If an essay connects to your work, send a considered response, ask a private question, or bring the idea into a real diagnostic conversation.

Ask Hardeep privately →Request a diagnostic →Follow updates →
HA
Hardeep Anand

Thirty years inside water infrastructure, now building the intelligence layer it was missing. The Systems Lens is where the thinking lives, in plain language, with the work shown.

Follow The Systems Lens