Deterministic AI agents

A new intelligence for energy infrastructure.

Brian’s AI agents connect operational, commercial and financial data to automate complex workflows, deliver traceable results and help energy owners and investors make better decisions.

A data centre site shown as a technical model: a data hall with rooftop solar and cooling plant, gas generators, battery storage and a grid connection
  • Rooftop solar. What has the rooftop solar generated this month? 110 MWh month to date. That is 2% above the performance guarantee.
  • Data hall. What revenue did the data hall earn last week? $665,000, from 24 MW of the 30 MW contracted.
  • Cooling plant. What did cooling cost us yesterday? $9,600, at an average PUE of 1.26. That is inside the 1.35 limit.
  • Gas generators. What has running the generators saved this month? $98,700 month to date. Their power cost $71/MWh against an average $141/MWh from the grid.
  • Battery storage. How many cycles has the battery done this year? 285 year to date, against the 365 a year its warranty allows.
  • Grid connection. How close are we to the grid import limit? Importing 38 MW right now, 95% of the 40 MW limit. That is near the limit.
Illustrative day 19:00

01 / The problem

Energy assets are complex. Understanding them shouldn’t be.

From the earliest investment assumptions through development, construction and operations, energy infrastructure generates vast amounts of technical, commercial and financial information.

Too often, that information is fragmented across spreadsheets, contracts, consultant reports and disconnected systems. As projects evolve, the assumptions, evidence and decisions behind their financial performance become increasingly difficult to trace.

Brian is being developed to connect that information across the entire asset lifecycle, creating a continuous, auditable picture of each investment, from initial feasibility and development through construction, operations, refinancing and eventual divestment.

02 / What Brian helps you understand

Four questions. One connected picture.

What is the investment case?

Understand the assumptions, costs, revenues and projected returns behind an energy investment.

How is it evolving?

Track how new information, project milestones and commercial decisions change the investment outlook.

Can we trust the numbers?

Connect financial models and reporting to the documents, assumptions and calculations behind them.

Is it delivering?

Understand operational and financial performance against the investment case and identify where outcomes differ from expectations.

03 / The asset lifecycle

One asset. One continuous investment history.

From the first assumptions to the final exit, Brian’s vision is to preserve the models, evidence and decisions behind an investment as it progresses through each stage.

  1. 01

    Feasibility

    Initial assumptions and investment returns

  2. 02

    Development

    Studies, contracts and evolving forecasts

  3. 03

    Investment decision

    Approved financial model and funding

  4. 04

    Construction

    Capital expenditure and delivery performance

  5. 05

    Operations

    Actual performance against investment expectations

  6. 06

    Refinance or divestment

    Valuation, refinancing and exit outcomes

One continuous investment record

  • Assumptions
  • Financial models
  • Contracts
  • Decisions
  • Performance
  • Returns

04 / Deterministic AI agents

AI where it adds value. Deterministic execution where accuracy matters.

AI is powerful at understanding complex information. But when it comes to financial calculations, commercial obligations and investment decisions, accuracy and accountability matter.

Brian’s AI agents combine the intelligence of AI models with deterministic execution. They follow defined, repeatable workflows, while critical calculations are performed by software, not generated by language models.

Reliable by design

Defined workflows and reproducible calculations deliver consistent results from the same validated inputs.

Traceable from source to result

Follow calculations and decisions back to the data, documents and assumptions behind them.

Intelligence without unnecessary AI costs

AI models are used selectively. Routine calculations and scheduled workflows can execute without repeated model calls.

Governed and auditable

Defined execution rules, recorded actions and human approval where required provide oversight of business-critical work.

AI-powered intelligence.Deterministic execution.

05 / Asset Intelligence

From complex information to decisions that matter.

Brian’s first application, Asset Intelligence, brings operational data, commercial agreements and financial analysis together to give energy asset owners a clearer picture of performance.

Starting with battery energy storage, Brian is already being developed and tested against real-world asset data.

Ask Brian

Is the rooftop solar doing what was promised?
The rooftop array is generating 0.74 MW, about 2.6% of what the site is using. Yield is running 2% above the level set in the performance guarantee. It is a small part of the site, and has saved $5,500 of grid energy so far this month. Checked against metered data, the performance guarantee and the financial model.
How is the data hall doing?
The data hall is drawing 23.2 MW for its customers. That is 77% of the 30 MW contracted under the customer agreement, so there is room to grow. It has earned $1.71 million in revenue so far this month. Checked against metered data, the customer agreement and the financial model.
What is cooling costing us?
The cooling plant is drawing 5.9 MW to remove 30.5 MW of heat from the site. That puts efficiency (PUE) at 1.24, inside the 1.35 the service agreement allows. Cooling has cost $173,400 in energy so far this month. Checked against metered data, the service agreement and the financial model.
What are the gas generators doing?
They are generating 16.0 MW because grid power costs $105/MWh and their fuel costs $71/MWh. They have taken 84% of the minimum daily quantity in the gas supply agreement. Running them at peak prices has saved $98,700 against grid power so far this month. Checked against metered data, the gas supply agreement and the financial model.
What is the battery doing right now?
The battery is discharging at 10.0 MW into the evening price peak. It has used 0.6 of the one cycle a day its warranty allows. Charging at low prices and supplying at the peak has been worth $77,400 so far this month. Checked against metered data, the battery warranty and the financial model.
Are we close to our grid limit?
The site is importing 38.0 MW from the grid at $30/MWh. That is 95% of the 40 MW the connection agreement allows, so it is near the limit while the battery charges. Grid energy has cost $753,000 so far this month, the largest cost on site. Checked against metered data, the connection agreement and the financial model.
Battery charge, %So far todayExpected
Contracts and obligations6 of 6 within
Customer agreement / Data hall83% of 30 MWWithin
Performance guarantee / Rooftop solar2% above guaranteeMet
Connection agreement / Grid connection14% of 40 MWWithin
Battery warranty / Battery storage0.6 of 1.0 cyclesWithin
Gas supply agreement / Gas generators41% of minimumOn track
Service agreement / Cooling plantPUE 1.26 of 1.35Within

No obligations are at risk right now.

Battery value this month$77,400The asset in focus
EBITDA this month$598,300Case $583,200
Free cash flow this month$317,500Case $306,000
Forecast IRR11.4%Case 11.2%
Forecast payback7.8 yearsCase 7.9 years

See the numbers. Understand the story.

Explore revenue analysis, contractual performance, asset cycling and financial scenarios, with results linked to their supporting information. Figures shown are calculated by software, not written by a language model.

06 / When something goes wrong

An alarm is only the start of the question.

When an asset trips, most systems tell you that an alarm went off. The questions that matter come next: why it happened, whether it has happened before, what it will really cost and what to do about it.

Brian is being developed to read the communications around an asset as well as its data: emails from the O&M provider, site reports, work orders and invoices. When something goes wrong, it connects the event to its history, sets out the full financial impact and drafts the next step for you to approve.

Alarm at 14:02: battery storage, inverter 2

Brian, unprompted
  1. Inverter 2 on the battery block tripped on over-temperature at 14:02. A quarter of the battery is offline.
  2. With 2.5 MW of its 10 MW unavailable, the site is losing about $1,100 a day in peak supply.
  3. This is the second trip in three months. On 12 August the O&M provider attended, replaced a cooling fan and reported there was not enough evidence to replace the inverter.
  4. Two trips on the same unit point to the inverter itself. It is 26 months into a five-year warranty, so I recommend replacing it and claiming the cost.
  5. I have drafted an email asking the O&M provider to reassess it and send a report, a work order and a quote. The full cost forecast is alongside. You approve, and the email is sent.
  6. Sent. I will track the reply and update the forecast when the quote arrives.
What Brian read, from August
Email, 12 AugustFrom the O&M providerAttending site after inverter 2 tripped on over-temperature.
Site report, 14 AugustInverter 2 inspectionCooling fan replaced. Not enough evidence to replace the inverter.
Work order WO-0412Fan replacementSigned off on 14 August.
Invoice INV-2291$1,850Booked to maintenance costs.
Forecast cost of this faultCalculated by software
Lost peak supply until replaced, 21 days$23,100
O&M reassessment and report$2,400
Replacement inverter, paid up front$86,000
Installation and commissioning$9,500
Battery offline for the changeover, 1 day$4,300
Cash out before the warranty claim$97,900
Expected back under warranty−$86,000
Net cost to the investment$39,300

About 7% of one month’s EBITDA at this site.

07 / Ready for diligence

The data room assembles itself.

Refinancing or selling an asset usually starts with weeks of gathering documents. With Brian, the record already exists. Every reading, model, contract, invoice, report and email is filed as it arrives, linked to the asset and to the decisions it informed.

Brian is being developed so that when a transaction comes, the data room is assembled from that record: organised for due diligence, with every number traceable to its source.

Data room / Data centre site
  1. 01Operating dataMetered generation, load, storage and grid import.4.2 million readings
  2. 02Financial modelsFeasibility, investment decision and operating models, with approvals and comments.7 versions
  3. 03ContractsCustomer, connection, gas supply, service and warranty agreements.14 documents
  4. 04Invoices and paymentsO&M, fuel, grid energy and equipment.312 invoices
  5. 05O&M reports and work ordersInspections, faults and remedial works.46 reports
  6. 06CorrespondenceEmails with the O&M provider, suppliers and counterparties.1,180 emails
Latest addition: site report and invoice INV-2291, inverter 2.Assembled for a refinancing or a sale.

08 / Built on industry experience

Built by people who understand what’s at stake.

Energy infrastructure investments involve complex commercial arrangements, significant capital commitments and decisions with consequences that can last decades.

Brian brings together experience in infrastructure investment, project finance, asset management and software development to address the challenges faced by the people responsible for these investments.

Infrastructure finance

Experience across infrastructure investment, financial management and asset economics.

Purpose-built technology

AI-native software designed around the financial and commercial realities of infrastructure assets.

09 / Let’s talk

The future of energy infrastructure deserves better intelligence.

Whether you’re developing new energy projects, managing operating assets or investing in infrastructure, we’d like to understand the challenges you’re solving.

Discover how Brian’s approach to intelligent automation and asset intelligence could support your business.