This entry of Machine Readable is by Sam Steyer, Head of Data Science & Head of Customer Success at Halcyon

Cold Outreach Can Freeze Your Pipeline

What makes a great ‘cold’ email? 

If you’ve read previous Machine Readable entries, this may not have been the start you were expecting, but bear with me: effective cold emails are hard to do.

And, they are challenging to explain in the positive. It’s easiest to explain what a great cold email is by thinking through its inverse: what does a nightmare cold email look like? It can be:

• Targeted to someone who does not want the product
• Unclear about who the sender is and what they’re asking for
• Impersonal, or even worse, erroneous in its attempt to be personal

In my past life as the CEO of a software startup, a data vendor incorrectly ingested my name as “Matthias” and, as a result, I received many exemplary bad cold emails. Here’s the opening of one: “Hello Matthias, I'm reaching out one last time to offer my expertise in guiding startups like Greenwork through the intricacies of international tax.” (We were operating only in the US.)

As large language models (LLMs) improve, one potential downside could well be business development agents that clog our inboxes with these impersonal emails. We do not need more emails; we definitely do not need many more emails at the scale that generative AI could enable.  

However, at Halcyon, we believe that we need better emails. And we believe that if BD teams leverage LLM’s strength in reading, rather than writing, then we can collectively enjoy and mutually benefit from fewer, more targeted emails that better understand customer needs. 

Finding Clean Energy Technology Customers

Customer intelligence is a prerequisite to better emails. AI-assisted search is well suited to parsing necessarily huge numbers of utility documents to identify potential partner companies in the electricity industry and the specific decision-makers within.

An example: suppose I am a developer of grid-scale energy storage. I offer a very valuable product, but only some electric utilities are adopting it today (they need both existing renewables already connected to their grid as well as a willingness to adopt new technology). Even energy-storage-friendly utilities are only procuring new assets at specific times under specific criteria.

Through Halcyon, I can check a utility's filings with its state public utility commission (or equivalent), its company outlooks, and its public statements to find the most promising partners. Let’s say we decide to start by researching Indiana’s CenterPoint South:

Success! This suggests that CenterPoint South has a publicly stated interest in grid-scale energy storage. Let’s dig deeper to see if the utility has issued a request for proposals that we might apply to now:

They do! Next up, we could dig deeper to better understand the formal application process. However, even in the utility space, you can’t sell your product to a corporate entity alone; ultimately, you must sell to people. So, we need to identify some CenterPoint South employees who work in this area so we can start building a relationship, understand their vision, and learn about the utility’s proposal evaluation criteria.

Now, we are in a place where we can write a cold email that makes the world better, not worse (as did the ill-fated emails for Matthias). We have found a customer we know is interested in our product, confirmed it is actively seeking a partner, and identified two specific technical leaders working on issues related to the RFP. Now, it’s on us to explain why what we do is relevant to their specifications; from here, we could take to LinkedIn or email and reach out.

Thoughtful Outreach at Scale

Now, let’s imagine I want to find more utilities like CenterPoint South that are in the market for grid-scale storage. Using Halcyon’s internal API, I can ask these same three questions of every utility in America and focus only on those that have active dockets or RFPs.

For our more technical readers, imagine a query loop that can instruct an API to return structured JSON responses rather than paragraph answers, returning a data class like this:

Asking these three questions across Halcyon’s data catalog of more than 800,000+ (and growing!) authoritative documents, our hypothetical developer could qualify all 3,000 US electric utilities with only 9,000 total LLM calls, providing a first-pass national lead list for less than $1,000, in under an hour (assuming some straightforward batch or parallel query running). The structured output can easily be stored in a spreadsheet or CRM and handed over to salespeople who will only focus on buyers who are happy to hear from them.

Halcyon is doing exactly this type of work today. If this sounds like a fit for you, or you are fellow developers and LLM-enthusiast exploring these projects, we’d love to talk!

(And if you’re attending RE+, I will be there doing thoughtful outreach of my own and I’d love to chat - drop me a note at sayhi@halcyon.eco. Add a personalized touch - if you’ve read this and see me, ask if my name is actually Matthias 😁 )

Comments or questions? We’d love to hear from you - sayhi@halcyon.eco, or find us on LinkedIn and Twitter