Clean Water Current

Operational Intelligence at Paseo Real Water Reclamation Facility: Lessons from Putting Plant Data and AI to Work

Sep 25, 2026

By: Prachi Karve, Onboarding Engineer, Aquasight, Troy, MI; Michael Dozier, Wastewater Director, Paseo Real WRF, City of Santa Fe, NM; and Steve Walker, Program Manager, Paseo Real WRF, City of Santa Fe, NM

At a Glace

Utility: City of Santa Fe, NM — Paseo Real Water Reclamation Facility (13 MGD design capacity; approximately 5 MGD average flow).

Challenge: Operating data scattered across SCADA, laboratory, and instrumentation systems, compounded by the loss of experienced staff to retirement and turnover.
Approach: A phased effort that first unified and validated plant data, then added an AI assistant that interprets that data for operators in plain language.
Operating model: Run by exception — the system monitors continuously and surfaces only the conditions that need investigation or action.
Key takeaway: Data quality and operator involvement, more than the technology itself, determined whether the tools were trusted and used.

The Broader Challenge

Clean water utilities have never had more data, yet many still struggle to turn it into timely decisions. A typical treatment plant generates thousands of data points every day across SCADA historians, laboratory information systems, field instruments, and maintenance records. Each system was usually acquired at a different time, for a different purpose, and by a different group, so few share a common structure or communicate with one another.

At the same time, the pressures on plant staff keep growing. Permit limits continue to tighten, particularly for nutrients. Energy and chemicals are among the largest controllable costs in treatment, with aeration alone often accounting for half or more of a plant’s electricity use. And the workforce is changing: many experienced operators are nearing retirement, and utilities across the country report difficulty recruiting and retaining licensed replacements. Much of what veteran operators know — which instruments tend to drift, how the plant responds to a storm, what an unusual trend usually means — has never been written down.

The result is a widening gap between the data utilities collect and the insight their operators need in the moment. Closing that gap was the goal at Paseo Real.

The Situation at Paseo Real

The Paseo Real Water Reclamation Facility serves the City of Santa Fe, with a design capacity of 13 MGD and an average flow of about 5 MGD. Like many plants its size, it relied on separate systems for process control, laboratory results, and instrumentation. Before this project, operators had to pull information from each source and piece the picture together by hand before making process decisions. That work was time-consuming, depended heavily on the experience of whoever was on shift, and left room for error — a risk that grew as seasoned staff retired or moved on and newer operators had less context to draw on.

The City set out to answer a practical question: could operators spend less time assembling information and more time acting on it, while capturing the institutional knowledge the plant was at risk of losing?

The Approach

The City worked with Aquasight to address the problem in two deliberately sequenced layers.

Layer 1: One trusted source of data. The first step was integration and validation. Using the AQSYNC™ platform, data from SCADA, PLCs, RTUs, field instruments, and the laboratory were brought into a single environment, where each incoming value is checked for gaps, flatlines, out-of-range readings, and other signs of unreliable data before it reaches operators. Process data is also linked to equipment history and lab results, so no single reading is interpreted in isolation.

Layer 2: Interpretation in plain language. Only once that foundation was in place did the team add an AI assistant, AVA™. It draws on laboratory and live process data, with operational logs planned for a later phase, to explain what is happening, why, and what operators may want to consider next, grounded in industry best practices. In day-to-day use this takes the form of color-coded daily status summaries, trend analyses that point to likely causes, and plain-language shift handoff notes. The assistant can also interpret SCADA screenshots, compliance logs, and photos of problem areas.

The guiding design principle was to run by exception. Rather than asking operators to watch screens all day, the system monitors continuously and brings forward only the conditions that warrant investigation or action.

Where Integrated Data Makes a Difference

Connecting and validating plant data creates value well beyond any single tool. The areas where the Paseo Real team sees the greatest potential are common to most treatment plants:

Energy: Aeration and pumping are typically a plant’s largest energy loads. Seeing dissolved oxygen, blower, and loading data together makes it easier to identify setpoint and scheduling changes that reduce power use without compromising treatment.
Chemical use: Connecting dosing data with lab results helps staff match chemical feed to actual demand rather than conservative assumptions.
Maintenance: Changes in equipment behavior can reveal degradation before it becomes a failure, supporting a shift from reactive to condition-based maintenance.
Compliance: Tracking trends against permit limits gives staff time to intervene before an exceedance, rather than explaining one afterward.
Troubleshooting: When process, lab, and equipment data are already correlated, finding a root cause can take minutes instead of hours.
Training and knowledge transfer: Newer operators gain access to the same context and reasoning that previously lived mainly with veteran staff.

Lessons Learned

The most transferable insights from Paseo Real have less to do with any specific technology than with how the project was sequenced and adopted. Several stand out for utilities considering a similar effort.

Data validation comes first. An AI tool is only as reliable as the data beneath it. If operators see it draw conclusions from a drifting probe or a stale lab value, trust erodes quickly and is hard to rebuild. Investing in data cleanup and automated validation up front is the foundation for everything else, not an optional step.
Bring operators in early. Adoption depends on the people who will use the tools every shift. Involving operators in deciding what they need to see, which alerts matter, and how information is presented builds ownership and surfaces practical knowledge that would otherwise be missed.
Context beats volume. Operators do not need more data; they need a smaller set of trusted, connected information that explains what is happening. Starting from the daily decisions operators make, and the information each one requires, is a more useful design approach than starting from the list of available tags.
Treat knowledge capture as a goal in its own right. Workforce turnover was a core driver of this project, not a side benefit. Planning deliberately for how operator expertise will be recorded and shared, through shift notes, documented responses to recurring issues, and consistent explanations, produces more lasting value.
Keep operators in the decision loop. AI assistance works best as decision support: the assistant explains and recommends, while licensed operators decide and act. Framing the tools this way helps address understandable skepticism and keeps accountability where it belongs.
Roll out in phases. A focused initial scope delivered early, visible wins that built confidence among staff and leadership and made the case for expanding. A phased approach also gives the utility room to learn and adjust before committing to the next stage.

Questions to Aske Before You Start

For utilities weighing a similar initiative, the Paseo Real experience suggests a few questions worth answering early:

Which recurring decisions consume the most operator time, and what information does each one require?
How confident are we in our instruments and lab data today, and who is responsible for keeping them reliable?
Where does critical operating knowledge live now, and what happens to it when a key staff member leaves?
How will operations, IT, and cybersecurity staff work together on data access and integration?
What early result would demonstrate value to both operators and leadership?

What's Next

Santa Fe is continuing to build on this foundation in stages. Voice-based input is now in design, so operators will be able to query the assistant hands-free without stopping mid-task. Operational logs will be added as a third data source alongside lab and live process data. Longer term, the validated data layer is intended to support a digital twin of the plant (Aquasight’s APOLLO™), allowing staff to simulate process changes and evaluate optimization strategies before applying them in the field.

Bottom Line

Paseo Real’s experience shows that the path to operational intelligence starts well before artificial intelligence. By first establishing one validated view of plant data and then layering on tools that interpret it in plain language, Santa Fe is moving its operations from reactive to proactive in a way designed to preserve institutional knowledge and scale without adding staff. For other clean water utilities facing the same pressures, the lesson is that the order of operations matters: trust the data, involve the operators, start small, and build from there.