There are plenty of podcasts about the energy transition and AI. Not many go deep on the engineering layer where the two actually meet; the architectures, the trade-offs, and the lessons that only come from shipping real systems in a domain where data is messy.
That’s the gap we’re here to fill. AI x Climate Tech is a new podcast series from Enjins for CTOs, Heads of Data, and senior engineers in climate and energy. Each episode is a 30–45 minute conversation focused on concrete system design.
Episode 1: Agentic AI in Climate Tech
For the first episode, we started close to home. Joanne, who leads multi-agent system development at Enjins, joins us to talk through what it actually takes to build and deploy agentic AI in climate tech.
We cover:
- Why multi-agent architectures outperform single-model approaches for climate and energy use cases
- How to handle the structured/unstructured data mix that breaks simpler systems
- What we learned building an agentic system for building decarbonization — real project, real constraints (case study)
- How to run a first internal deployment that actually changes minds
What’s next
We have strong conversations lined up — covering everything from productionizing agentic AI at scale, to real-time energy data handling, to how climate tech investors evaluate AI value propositions.
If you work in climate or energy and want to talk architecture, we’d love to hear from you.
TL;DR – recap of Episode 1
- Several converging breakthroughs have made agentic AI possible right now: transformer architecture, the ability to process unstructured data (not just structured databases), and the shift from prompt-and-compute to AI that can actually call tools and act.
- Climate tech companies face a specific economic pressure: grow revenue without growing headcount. Agentic AI is one of the few levers that makes this possible, automating high volume workflows at a fraction of the cost of new hires.
- The sector’s hardware/software complexity (solar, EVs, grid assets, sensors) creates a dense layer of technical documentation and regulation. Agents can handle the first 80% of that knowledge work autonomously, freeing human experts for the judgement calls that actually need them.
- Not all agentic approaches are equal. Off-the-shelf tools and managed platforms have a ceiling. A custom multi-agent architecture built on open-source frameworks gives you full control, unlimited composability, hybrid reasoning, and, critically, the ability to deploy agents as a product for your own customers, not just as an internal tool.
- Successful implementation starts small and builds with domain experts, not around them. You can’t anticipate where the real bottlenecks are without them. Prove value fast, then expand.
Full story on “How Agentic AI is transforming climate tech impacts?”
In the race to net zero, speed is a crucial tool. As data and climate technology go hand in hand, Agentic AI is perfectly positioned to accelerate positive impact and increase efficiency. Forward-looking organizations are realizing that Agentic AI can simplify their complex and time-consuming processes. But the question remains: “What is needed to ensure that Agentic AI has a positive impact, and how do you implement it successfully?”
Why Agentic AI (and why now)?
Highly accessible interfaces based on Large Language Models (LLMs) have become widespread, and many people are starting to see the real-world impact these can provide. On the more complex end of the spectrum, AI agents can make numerous processes more efficient, effective, and accurate.
Several converging breakthroughs have made this moment possible:
1. The Transformer Revolution: The transformer architecture fundamentally changed what AI can do, enabling models to understand context, reason across domains, and process language at scale. This is the foundational shift that made everything below possible.
2. Breaking Free from Structured Data: Traditional analytics required clean, structured data stored in data lakes or warehouses. LLMs have shattered that constraint. Satellite imagery, scientific papers, maintenance logs, field reports, and policy documents can now all be processed and reasoned over. In climate tech, where critical knowledge is often locked in unstructured formats, this is a genuine paradigm shift.
3. From Answering to Acting: Early LLMs were limited to prompt and response. The move to tool calling and agentic frameworks means AI can now execute code, query live databases, trigger APIs, and orchestrate multi-step workflows autonomously. This transforms AI from an advisor into an operator.
4. Model Context Protocol (MCP): This emerging standard gives AI models modular, standardized access to software tools and external data sources, making it far easier to compose capable, domain-specific agents without building everything from scratch.
5. Retrieval Augmented Generation (RAG): RAG grounds LLM outputs in specialized, up to date knowledge bases, ensuring reliability and sector-specific accuracy rather than relying solely on what was learned during training.
The role of Agentic AI in Climate Tech
There are two strong arguments for why agentic AI is particularly well suited to climate tech. One is about economics, the other about the nature of the work itself.
The climate tech sector is at an inflection point. The green premium is gone, investors are demanding profitability, and hiring your way to growth is no longer viable. Agentic AI offers a different path: augmenting your workforce with agents rather than new hires. High volume, repetitive tasks like processing data feeds, system monitoring, and customer workflows can run continuously and at a fraction of the cost, letting you scale revenue without scaling headcount.
The second argument is specific to climate tech’s DNA. Most players operate at the intersection of hardware and software (solar, EV infrastructure, grid assets, sensor networks), creating a dense layer of technical documentation, product specs, and regulation that teams must constantly navigate. Agentic AI can absorb that knowledge base and autonomously handle the first 80% of the work: answering technical queries, cross-referencing regulation, drafting documentation, and flagging edge cases. Human experts then focus on the final 20%, the judgement calls, exceptions, and relationships where they add irreplaceable value.
Why a custom multi-agent architecture is the right long-term bet
Not all agentic AI approaches are created equal. Off-the-shelf tools like Copilot or Claude’s Co-work offer a fast start for internal productivity tasks, and managed platforms reduce the engineering burden with pre-built orchestration. But both come with a ceiling, and ambitious climate tech companies will hit it sooner than they expect.
A custom multi-agent architecture, built on open-source frameworks, is the approach that scales with your ambitions. Here is why:
Full control and observability: In regulated industries, you need to explain and audit every decision. A custom build gives you complete visibility into every step of the reasoning chain, something that black-box platforms simply cannot offer.
Unlimited composability: Each agent handles a specific role: orchestrating, retrieving data, reasoning over regulation, or triggering actions in external systems. New data sources, tools, and use cases snap in without rebuilding from scratch. Your architecture grows with your ambitions, not against them.
A product, not just a tool: Off-the-shelf solutions are built for internal use. A custom multi-agent system can be deployed externally, as a capability you offer your own customers. What starts as an internal efficiency play becomes a genuine product differentiator.
Hybrid reasoning: The most powerful architectures combine deterministic logic for steps that require consistency and compliance, with LLM-based reasoning for the parts that require judgement and flexibility. That balance is only achievable when you control the architecture end to end.
Case study: lessons from creating an agentic system for building decarbonization
Why: The solution needed to support experts in their work, which involves a lot of data, expert knowledge, and manual effort. By removing operational bottlenecks, automated agents support experts to achieve more impact in less time.
How: We worked closely with their experts to map out their processes and identify targeted automations for each business case. Using integrated feedback loops the agents can continuously improve, and new use cases are identified.
What: We designed a production-ready multi-agent architecture, deployed in a self-hosted environment. This guarantees privacy, compliance, and full control over sensitive building data. It consisted of 3 core agents: an Orchestrator Agent, a Raw Data Agent, and a Contextual Data Agent.
The result: A greatly decreased workload for building performance experts, who can do their job faster and with less stress. Instead of being overwhelmed by the complexity of their role, they feel in control and supported to achieve the best results.
You can read the full case study here.
How to run an initial deployment that changes minds
Users frequently underestimate the impact that AI agents can have on their work, or how effective they can be for performing complex tasks. Perhaps they have had disappointing interactions with AI tools in the past, or believe that an AI agent is only as capable as a ‘general’ model like ChatGPT.
Whatever the cause, initial reluctance can be overcome. The key is to build with the experts or users, and not for them. Expert involvement is a crucial ingredient, because you cannot anticipate which domain knowledge is needed, or where they lose time in their current processes.
By starting with a small use-case, you can demonstrate the value. Then, using the addition of more capabilities to a multi-agent system, you can help them uncover new use cases and efficiencies.
In other words: build fast, prove the value, and change minds, one workflow at a time. For climate tech companies navigating the pressure to scale without inflating headcount, that’s not a small thing. It’s how you stay competitive, keep your experts focused on what matters, and turn your AI investment into something that compounds over time.
