Asset intelligence for solar and storage
Knows your assets.
Reads your ledger.
Writes back.
The operating layer for energy assets.
A utility-scale battery has a design case, a stack of contracts, a telemetry feed, a settlement statement and a project-finance model. Five systems, five owners, none of them talking to each other. Brian holds all five against one asset and grades what actually happened against what was contracted.
ONE ASSET · THREE TRUTHS
- Designedlayer 01 · what the asset was built to do
- Contractedlayer 02 · what you are actually owed
- Realisedlayer 03 · what the meter and the settlement say
The problem
One asset. Five systems that never talk.
Working out how an asset's performance maps to its returns is obscured by fragmentation. The systems that track performance are disconnected from what the asset was designed to do, from what was contracted, and from where the money is recorded.
Storage is the hardest case. State of charge is inferred rather than measured, round-trip efficiency only means something over an energy-balanced window, and the revenue arrives as a settlement statement in somebody else's format, weeks after the half-hour it refers to.
Silent triggers
Nobody was watching
A DSCR covenant breaches. A performance guarantee falls short. The trigger fired in the data weeks before anyone read it in a quarterly pack, and by then the conversation with the lender is a different conversation.
Lapsed claims
The window closed
Availability and capacity-retention guarantees carry claim windows measured in weeks. Miss the window and the remedy is gone, whether or not the underperformance was real and provable.
Contract conflicts
Two definitions, one asset
The availability definition in the O&M agreement is not the one in the offtake. The curtailment carve-out sits in a side letter. Nobody notices until both are invoked in the same month.
Diligence as archaeology
Priced as a discount
At sale, somebody spends three months reassembling five years of paperwork from inboxes and shared drives. Every gap in the record is doubt, and doubt is taken off the price.
The real cost
Analysts spend most of their time producing information and very little of it understanding what the information means.
Invoice processing, contract summarisation, covenant calculations, report collation. It looks like work. It is latency between data existing and a human knowing what it says. Brian runs the production half continuously, so by the time a person opens anything the comparison has already been made and their job is the judgement: is this right, what does it mean, what do we do.
The obligations register
Every promise in every contract, with a number attached.
Contracts go in. Brian extracts each obligation — the warranted level, how it is measured, the remedy, the claim window — holds it against the asset, and grades it against what the asset actually did.
Drift to claim
Measured against warranted
A warranted round-trip efficiency of 88% against a metered 84.1% is not a chart, it is a gap, a remedy figure and a deadline. Brian assembles the claim packet — exposure, remedy calculation, notice requirement, evidence and the contractual procedure — while the claim window is still open.
Cross-contract anomalies
Priced in IRR basis points
Where two agreements covering the same asset disagree — an availability definition, an escalation basis, a curtailment carve-out — Brian surfaces the conflict against the clauses that create it and prices what it costs in basis points of equity IRR.
Entitlement countdowns
Time remaining, not a date
Claim windows, warranty terms, option dates and step-in rights run as live countdowns attached to the asset. An entitlement you still hold is worth something. One that quietly expired is worth nothing.
Reporting calendar
A per-asset to-do list
Deliverable obligations to lenders and equity — what is due, to whom, on which date, from which clause — as a working list rather than something reconstructed from a diary the week it is late.
Provenance
Every figure says where it came from.
Nothing is presented as fact that is not one, and where the system is uncertain it says so rather than guessing. Every revenue and performance figure sits on one of six rungs, and the interface tags which one. A projection never sits beside a settled number pretending to be the same kind of thing.
A line starts as an assumption. Bind the contract and it becomes contracted. Connect the market feed and you get a graded dispatch signal. Upload telemetry and the electrical actuals are real. Upload the settlement statement and the revenue is settled. The number improves as the evidence arrives.
| Rung | Name | What it is | Where it comes from |
|---|---|---|---|
| 1 | Settled actuals | Money that was actually paid, by revenue stream | Settlement statements |
| 2 | Metered telemetry | The real electrical behaviour of the asset | SCADA and metering feed |
| 3 | Dispatch engine | What a good operator would have captured | Modelled dispatch over real market prices |
| 4 | Contracted | Revenue fixed by an agreement | Bound PPA, tolling, capacity or O&M contract |
| 5 | Register assumption | The planning baseline the model was built on | Asset register inputs |
| 6 | Labelled placeholder | A stand-in so the screen exists before a feed lands | Retired the moment a real feed connects |
Boundary
No model in the number path.
A language model reads documents at ingest and proposes structured data, which a human confirms. It never computes a figure. The project-finance model, the debt sculpting, the returns, the dispatch benchmark and the drift calculation are deterministic code — the same inputs give the same outputs every time, and every one of them traces back to the register and the contracts.
The weekly report
An asset-intelligence newspaper, every week.
One continuous read per portfolio: what the assets did, what that was worth against plan, and what fired.
Verdicts, grouped by trust
Plan · Live · Obligations
Equity IRR and minimum DSCR sit under Plan. Dispatch capture and today's spread sit under Live. Breached obligations sit under Obligations. Grouped by rung, so nobody reads a projection as an outcome. Every tile links through to the model line that produced it.
Dispatch, benchmarked honestly
Three runs, one chargeable gap
Perfect hindsight, a causal fair operator that only sees what had already happened, and the realised position. Only the gap against the fair operator is chargeable. The structural gap no causal operator could ever have captured is never held against the asset.
Alerts that are already claims
With the remedy calculated
A breach does not arrive as a red dot. It arrives as a claim packet: the measured actual, the warranted level, the remedy under the contract, the notice required and the days left on the window.
Transaction readiness
A data room that accumulates instead of being assembled.
Contracts, performance data, warranty and guarantee claims, settlement statements, and the emails and invoices sitting behind all of them, connected to the asset across its whole life rather than gathered up in the month somebody decides to sell.
The value is transactional. Doubt about the record is what discounts a price and what slows a legal process down. An asset whose history is already evidenced and already traceable sells for more, and sells faster.
It stays on your infrastructure.
Brian writes processed outputs back to your own Google Drive or SharePoint, so the record accumulates where your organisation already keeps things. The substrate is scoped per client. You are never holding your asset history inside somebody else's format.
Always ready, never assembled.
Because the register is built as the asset operates, the diligence pack is a view of something that already exists. There is no reconstruction phase, and no three months of somebody reading old inboxes to find out whether a claim was ever made.
The interface
Ask for the reporting pack in a sentence.
Brian is not a dashboard you go and read. It is the layer between you and the asset data, reachable in the tools you already work in.
Generate the covering reporting pack for Ridgeway Solar, quarter ending March 2026.
Brian pulls the live asset data, compares it against the contracted case and the plan, builds the narrative with the charts, and files it to your data room. When you have read it and you are satisfied, it goes to the lender distribution list.
The analysis does not start when you ask. Background agents have been running against generation, market prices and operations continuously, and the report is assembled from work that was already done. Your time goes on interrogating it, not producing it.
Candid
What is solid, and what is not yet.
The same discipline the product applies to its own numbers, applied to the product.
Solid.
Market prices, demand and generation are real GB half-hourly data. Settlement statements are the ground truth for revenue and the telemetry feed is the ground truth for electrical behaviour; both accumulate, and an unfamiliar file format is mapped once and then flows through on its own. The financial maths is deterministic and traceable to the register and the contracts.
Not yet — and labelled that way.
Until your settlement feed connects, the realised dispatch leg is a modelled operator rather than your actual trades. Round-trip efficiency from a single day of telemetry is indicative, not asserted. Availability guarantees cannot honestly be graded from a power and state-of-charge export — they need a fault signal, and Brian says which feed it is waiting for rather than inventing a measurement it cannot make.
Connect one asset. See what it finds.
We are taking a small number of design partners across the NEM and GB markets.