Building the AI layer for enterprise energy management

I've spent my entire career at the intersection of energy and software: At McKinsey helping utilities build digital strategies, at SunEdison building out solar and smart grid products, at C3.ai finding new ways to optimize the smart grid. And of course, here at Arcadia, where the core problem has always been the same: data is required to unlock energy optimization, but energy data is messy.

Bad data means inaccurate bills. Inaccurate bills mean bad cost insights. Bad cost insights mean imperfect procurement decisions. Imperfect procurement decisions mean companies overspend on one of their largest controllable cost lines.

Our acquisition of ENGIE Impact changes what Arcadia can do about this problem, because we now have operational depth and data infrastructure in the same organization, at a whole new scale.

Data quality is the foundation

Here's what I've learned about building energy software: data is key. Without good data, you’re flying blind — across decisions, insights, and AI tools. 

A simple example: if you don't know how your building uses energy, how do you know what savings you'll get by putting solar on your roof? When you're feeding sophisticated models with utility data that's inconsistent or inaccurate, it's garbage in, garbage out.

Arcadia's platform covers more than 10,000 utility providers globally. We've spent years building the connectors, the normalization logic, and the validation frameworks to turn raw utility data into something you can trust and act on. We run hundreds of validation checks on every bill before it ever reaches a customer. It's not glamorous work, but it's the work that makes everything else possible.

ENGIE Impact has been doing the hard work of enterprise energy management for three decades — processing billions in annual payments across complex portfolios. That's the kind of scale you can't simulate. You earn it.

Together, what we have is rare: the data infrastructure, analytics, and operational scale to build a true AI-powered energy management platform, tuned by experts, that delivers measurable outcomes for our clients.

Where AI adds value

The energy industry is still in early days with AI, but we've spent two years evaluating use cases and the results are real. Here's where we're placing our bets:

• More (and better-quality) energy data. We're deploying a combination of proprietary AI tools, agentic workflows, and rule-based algorithms to extract, validate, and standardize utility data into a unified model. The result is broader data coverage, more accurate extraction, and faster turnarounds.
• Cost validation. Arcadia recreates and verifies utility bills to ensure accurate cost calculations, applying LLMs to fill data gaps, detect anomalies, and identify billing discrepancies. This enables us to deliver and pay audited bills — at scale — across enterprise portfolios. And if the bills are wrong, our team fights to get the money back for our customers.
• Bill forecasting. Avoiding late fees and shut-offs for our customers is a core mission. Using traditional machine learning and cutting-edge foundational time series models, we forecast bills and make confident estimated payments — so that if a bill is delayed for any reason, we avoid the penalties from utilities.
• Energy optimization. Once we have validated data, we leverage AI to evaluate every available lever — solar, storage, procurement, demand management, efficiency — and identify the optimal energy management strategy for the customer, balancing cost, risk, and sustainability.

Where we're headed

With energy markets becoming more volatile, costs going up, and mitigation strategies requiring complex load management, I believe a fundamental shift is required in how enterprise energy is managed. Spreadsheets and people checking in periodically will not cut it anymore. A cohesive energy platform — taking near real-time granular data, combined with expert intelligence and AI — will be required to manage energy in this new world.

Our product roadmap is focused on delivering that platform: enterprise energy management that is increasingly automated, increasingly proactive, and increasingly integrated.

We're deploying AI agents that extend the capacity of Arcadia's energy experts and accelerate delivery. Right now, even with great tools, customers still have to ask questions and follow up. The platform of the future tells you what you need to know before you ask, and executes proactively.

That's what we're building: agentic workflows that close data gaps automatically, flag risks, surface insights at the portfolio level without manual assembly, and take action on your behalf.

Security is foundational

I want to be direct about one thing: AI in energy means handling sensitive customer data. We take that seriously.

Every AI/ML workload runs inside Arcadia's secured cloud infrastructure. We only license enterprise LLM instances that prohibit using customer data to train third-party models. Customer data is encrypted in transit and at rest, and access is limited to authorized Arcadia personnel. We maintain SOC II certification — and we're continuing to invest in security as we scale.

That's not a footnote. It's foundational.

The human element of energy intelligence

We believe that AI tools in the energy sector are only as trustworthy as the expertise behind them. That's why we employ a deep bench of procurement advisors, tariff analysts, data scientists, utility billing experts, and environmental engineers who have spent their careers understanding how the global energy system actually works.

This operational depth allows us to maintain accuracy standards of 99% — significantly higher than the 90–95% typically achieved by generic AI models in energy data processing. You need experts who can identify when a model is wrong, further tune it, and use AI to extend their capacity to deliver complex solutions to customers.

The combination of Arcadia's technology infrastructure and ENGIE Impact's operational and domain expertise is unlike anything else in this market. And we're just getting started.

If you want to see what that looks like in practice, reach out — I'm happy to walk you through it.