The platform explained

IT MAKES BETTER
DECISIONS THAN YOU CAN.

Not because it's smarter than you. Because it never sleeps, never gets tired, and can simultaneously consider more variables than any human team can hold in their head at once. Here's what it actually does — and what that means for your bottom line.

The core idea

EVERY 5 MINUTES,
A BETTER DECISION.

A battery storage asset makes hundreds of dispatch decisions every day. Each one has a financial consequence. The difference between a good decision and a mediocre one is typically $40–200 per interval. Across a year, that adds up to hundreds of thousands of dollars.

Avrenix makes those decisions — or informs them — using nine purpose-built engines that each handle a different dimension of the problem, with the Chai co-pilot to explain them. They run in sequence, each conditioning the next, producing a dispatch recommendation that is simultaneously optimised for market conditions, weather risk, portfolio balance, and regulatory constraints.

Every decision is explainable. Every outcome is attributed. Nothing is a black box.

Your data
Front-of-meter revenue uplift is measured against your own baseline schedule using a 12-month walk-forward backtest on out-of-sample periods the system never trained on — five baselines, Diebold-Mariano significance testing. Avrenix is pilot-stage: the number you see is yours, not someone else's headline.
Your bill
Behind-the-meter savings — through demand-charge reduction and joint tariff optimisation — are proven directly against your facility's own interval-meter data and utility tariff during the free pilot.
What the system actually does

REAL SITUATIONS.
REAL DECISIONS.

These are illustrative examples of the kinds of decisions Avrenix makes — or protects you from making badly — every day. The figures are worked scenarios, not customer results; your own numbers come from your own data during the pilot.

Tuesday 14:23 · ERCOT West
Wind generation just dropped 18% unexpectedly. Prices are rising fast.
Without Avrenix: The system continues its pre-planned discharge schedule, unaware of what's about to happen to prices. Sells 60 MW at $84/MWh.
With Avrenix: FENRIR detected the wind deviation 40 minutes earlier while stress-testing adversarial scenarios. HELIOS revised its price forecast upward. The system held reserve and discharged 80 MW at $127/MWh when the spike arrived.
Revenue difference this interval: +$3,440
FENRIR adversarial detection · HELIOS forecast revision · Dispatch controller execution
Wednesday 16:45 · PJM West Hub · Day 22 of billing period
Facility load is climbing toward what might become this month's peak demand event.
Without Avrenix: Load hits 1,840 kW at 16:52. This becomes the monthly peak. Demand charge: $33,120.
With Avrenix: PHALANX flagged 87% peak probability at 16:30. The battery discharged 400 kW into the facility load, capping it at 1,440 kW. STRIX had pre-validated this schedule 8 hours earlier, confirming it would keep peak below 1,500 kW in 96% of scenarios.
Demand charge avoided: $7,200 this month alone
PHALANX peak probability engine · STRIX schedule validation · Dispatch execution
Thursday 09:15 · CAISO · Ancillary market window
Grid operator offers $76/MW for regulation service. Spot price is $81/MWh.
Without Avrenix: Operator manually decides — usually defaults to spot, misses the regulation bid, or accepts it without knowing the true opportunity cost.
With Avrenix: CONSUL calculated that spot opportunity cost for the 8 MW offered is $72/MW adjusted for dispatch timing. Regulation at $76/MW is marginally better — but a counter-bid at $82/MW is the right move. CONSUL submitted the counter. It was accepted at $80/MW. Three minutes of autonomous negotiation.
Ancillary uplift: +$640 this interval. Annualised across similar windows: +$38,000
CONSUL autonomous negotiation · SYNDEX microstructure check · Portfolio optimisation
Monday 08:00 · Post-weekend analysis · MISO
Revenue came in below forecast last week. Why?
Without Avrenix: Review settlement statements, compare to forecast. Hard to isolate cause from correlation. Takes hours, often inconclusive.
With Avrenix: KAIROS ran causal attribution overnight. Example root cause: wind capacity factor ran well below forecast for three consecutive afternoon intervals. Counterfactual analysis quantifies how much of the shortfall an accurate forecast would have recovered, and the model recalibrates.
Root cause identified in seconds. Model improvement applied automatically.
KAIROS causal attribution · Counterfactual analysis · HELIOS forecast recalibration
The nine engines

EACH ONE DOES
ONE THING EXCEPTIONALLY WELL.

They run in sequence. Each module's output conditions what comes next. The result is a decision that reflects more dimensions simultaneously than any single model can.

01
HELIOS
Forecasts electricity prices and demand 48 hours ahead. Also tells you how confident it is — wide uncertainty triggers more conservative planning downstream.
Output: price trajectory + confidence intervals
↓
02
SPECTER
Reads the grid's current personality — what regime it's in, how it typically behaves, which patterns are active today. Tells every other module what kind of day to expect.
Output: regime classification + grid profile
↓
03
FENRIR
Stress-tests the plan against 50 adversarial weather and market scenarios. Finds the dispatch strategy that holds up even when things go wrong — not just when everything goes right.
Output: robust dispatch plan + revenue floor
↓
04
SYNDEX
Checks whether other large batteries are likely to dispatch at the same time — which would push prices down. Adjusts execution timing to avoid selling into a crowded market.
Output: optimal execution timing + adjusted MW
↓
05
CONSUL
Autonomously balances spot dispatch against ancillary service opportunities. Submits and negotiates bids across regulation, reserves, and synthetic inertia without manual intervention.
Output: optimised portfolio allocation + bid submissions
↓
06
DISPATCH
The final decision — a specific MW value, direction, and timing — is sent to your battery. Every element of the above is reflected in this single executable instruction.
Output: executable dispatch command

STRIX validates every candidate schedule against 2,000 scenarios before it's recommended. KAIROS attributes revenue after the fact, identifying what drove each outcome and automatically improving future decisions. PHALANX and TARSIS run the same sequence for behind-the-meter facilities, replacing the market layer with demand charge and tariff optimisation.

What you actually see

THE PORTAL IS WHERE
INTELLIGENCE BECOMES ACTION.

You don't need to understand every module to use the platform. The portal surfaces what matters — what happened, why, and what to do about it.

📊
Revenue attribution

Every dollar of revenue — and every dollar you missed — is explained. Not just "market conditions were unfavourable." Specifically: wind deviated by 15% at 14:00 and you could have recovered $4,200 by holding reserve until 15:30.

⚡
Actionable alerts

Alerts that tell you what to do, not just what happened. "Peak risk is 87% for the next 2 hours — PHALANX recommends discharging 200 kW now." You act or you don't. Either way, you're informed.

💬
Chai — ask anything

Ask Chai "why did revenue drop Tuesday afternoon?" and get a causal analysis in plain English. Ask "what's our peak risk for the rest of the month?" and get a probability, a timeline, and a recommendation. Site-specific intelligence, conversational interface.

Common questions

THINGS PEOPLE
USUALLY ASK.

Avrenix is a decision support and intelligence system — it produces recommendations that your existing dispatch infrastructure executes. It does not directly control your battery, and it does not actuate any hardware. This is intentional: operators retain full authority over every dispatch decision, with Avrenix providing the analysis that makes those decisions significantly better.
From account creation to a working intelligence layer: typically 2–5 business days. The platform runs on simulated market data immediately after signup — so you can explore the portal, run STRIX simulations, and understand the interface before connecting your real asset. Connecting your live market data feed takes a few hours. Connecting your BMS telemetry depends on your system but is usually a day's work with vendor support.
Minimum viable: your market node/hub (for real-time price data), your battery specs (MW, MWh, chemistry), and your operating market. That's enough to run HELIOS, STRIX, SPECTER, and FENRIR on real market data. For full intelligence, we also need battery telemetry (SOC, MW readings from your BMS) so KAIROS can attribute your actual revenue and CONSUL can optimise ancillary bids in real-time. Historical price and dispatch data accelerates the learning curve but is not required.
The 90-day pilot is completely free — no credit card, no commitment. After the pilot, you choose a plan based on your needs. A Front-of-Meter plan grants the seven FoM engines: Operator is $12,000/month for a single site, Fleet is $32,000/month for up to 10 sites (Fleet is a dual plan and also includes the two behind-the-meter engines). Behind-the-Meter plans grant PHALANX and TARSIS — BTM Facility is $4,200/month and BTM Portfolio is $14,000/month. The point of the free pilot is that you see the actual revenue uplift or bill savings on your own data before you pay, so you can judge the ROI against your real numbers rather than ours.
Your site data, dispatch history, and revenue figures are yours alone. We do not share, sell, or use your operational data to train models that benefit other customers. Each account's data is isolated. For Enterprise customers, on-premises deployment is available — your data never leaves your infrastructure. Full security documentation is available on the Security page.
An Avrenix account is an organization, not a single login. You invite teammates by email with single-use, expiring invitations, and each gets a role: Owner (full control, including billing and the permission configuration), Admin (manages members and sites and runs the engines), Member (operational — sites and engines, no billing), or Viewer (read-only, built for executives and auditors who need the numbers but shouldn't touch a control). The Owner can tune what admins and members are allowed to do within anchored bounds, and can optionally allow teammates from your company's email domain to join automatically (off by default). Access is enforced on the server on every request — the interface only hides what a role can't do; it's the backend that actually stops it.
Every forecast comes with confidence intervals — when the system is uncertain, it tells you. FENRIR's adversarial planning is specifically designed for the case when the forecast is wrong. STRIX validates every schedule before execution, rejecting any plan with greater than 5% probability of a serious downside outcome. And because you retain final dispatch authority, no recommendation is ever executed without your infrastructure acting on it. The system is designed with the assumption that it will sometimes be wrong — and the protections are built around that assumption.

BETTER DECISIONS.
STARTING THIS WEEK.

90-day free pilot. Nine engines and the Chai co-pilot. Simulated data running immediately — live data when you're ready.