The Plan Isn't to Beat Salesforce. It's to Sell to Them.

Leadpoet (SN71), from a Beyond Finance community call with its co-founders, Gavin Zaentz and Pranav Ramesh.

Leadpoet is a sales intelligence platform. It finds companies that might buy what you're selling, works out which of them are ready to hear from you, and shows you why. Gavin Zaentz and Pranav Ramesh have been building it since early last year, both of them out of NASDAQ and have extensive sales experience. Subnet 71 is the subnet that powers the sales intelligence.

Salesforce and Hubspot are the customers, not the competition

Every subnet with an enterprise story eventually meets the incumbent question, and the usual answer is displacement. This one isn't.

Gavin splits his market in two. There are the databases, ZoomInfo and Apollo, which in his account just own a lot of data and sell it cheaply and unqualified, so you buy a hundred thousand contacts and a lot of them are wrong or cold. Then there are the newer point-solution agents, Gojiberry being the closest thing to Leadpoet, mostly sourcing and outreach on LinkedIn and narrower on the signals they read.

His argument against the databases isn't volume, it's freshness: "all real-time qualification, real-time sourcing, validating the data quality". Underneath that is a claim about how sales teams actually spend their day. Mostly not selling, and a lot of it qualifying. Of a hundred thousand companies that fit on paper, maybe five to ten thousand are in the market, and Leadpoet's job is to hand over that thousand with the hiring, funding and expansion signals already attached.

Pranav highlighted that they're building the "first ever full funnel sales LLM", where something like Salesforce Einstein optimises parts of the funnel rather than the whole of it. If that's right, Salesforce isn't a rival at all. It's a buyer.

The plumbing is already there. @LeadpoetAI has a place on the Launchpad programme, and leads pipe into Salesforce and HubSpot instances.

This is a quieter route into the enterprise than the one Bittensor usually talks about. Not a subnet taking share off an incumbent, but an incumbent buying a subnet's output and selling it on inside its own product.

So what are they actually buying?

Two things, out of one mechanism. Gavin calls it the two-armed approach.

The agent is the body. It sources and qualifies today, and with V2 due shortly it starts doing the outreach as well, watching the signals and running the process up to the point of the meeting. The LLM is the brain, "a sales model that can help reason, help route that body to be doing effective things".

Both come out of the same place, which they call the lab. Each day a benchmark shows where the agent is performing and where it's failing, broken down by industry and by intent signal. Miners pick a weak spot, Gavin's example is healthcare jobs in Australia, and miners suggest a new data source or a better routing decision. Experimental tests run and if the improvement holds up it merges in and that source becomes part of the agent.

The model gets built out of the trail the miners leave doing it. Around 250 of them, running loops at the same time, which is Pranav's answer to why any of this needs a network instead of a team:

"They can all consecutively be running these auto research loops, providing us so much more data and so much more compute as a result of that than a traditional Web2 company." - Pranav Ramesh

So the agent is the product you'd expect, and Gavin means to sell it widely. The sales LLM is the byproduct, and he thinks it's the bigger asset:

"...the agent is something that we'll sell to thousands, tens of thousands, hundreds of thousands of companies. The LLM we can sell it to 10 companies and it'll even exceed the value of the agent..." - Gavin Zaentz

Ten buyers. And some of the ten he has in mind range from Apollo, ZoomInfo, HubSpot and GoHighLevel, companies carrying hundreds of millions of dollars of pent-up demand for inference and paying frontier prices for a generalist model. "We're able to provide them a more specialized model for sales and also at a cheaper rate than these frontier models. So for them, it's a no-brainer." Cheaper and better on a narrow domain is the specialist's whole case against the frontier labs, and it's a much easier sale than asking anyone to switch platforms.

The byproduct of their network may turn out to be worth more than the job.

And how does any of that reach the token?

This is where most enterprise stories in Bittensor go quiet. You have a product, you have customers, and no way to explain why a single dollar of it touches alpha, because your customers don't hold alpha and aren't going to start.

@gavinzaentz doesn't argue the point. Ask an enterprise buyer to acquire a token before they're allowed to use your software and you lose them instantly, and a client like Dropbox would be gone. So the customer never touches it. What he says instead is the load-bearing sentence of the whole call:

"The idea is that the Subnet is a commodity producer. It produces this agent, produces the LLM as that's ready, and we are able to tap into that. We're able to pay for that." - Gavin Zaentz

And the same goes for anyone else. Spin up a sales agency, or your own sales platform doing nothing but UK landlord and tenant leads, and you're in the same queue as Leadpoet is. Alpha, as Gavin puts it, is "the keys to get in and access the commodities".

So the customer pays dollars to the company, and the company pays alpha to the network. Leadpoet is a customer of its own subnet.

It isn't switched on yet, and that matters. The plan is that alpha-gated access starts once the system has exceeded the team's burn: the system pays for the team first, then gets made sustainable, then gets scaled. So this is a design rather than a mechanism you can go and inspect.

What Leadpoet is asking you to read differently

Most subnets are read as companies with a token bolted on, and judged on whether the company has customers. Leadpoet wants it the other way round: a network that produces a commodity, and a company that happens to be its first and largest buyer.

The difference changes what counts as evidence. A company with customers tells you the product works. It tells you nothing at all about whether the token does anything. What you actually want to know is whether there's something in the design that forces a buyer through the token to get what they came for, and whether that buyer has a cheaper way round.

Follow it out and adoption stops meaning a subnet displaced an incumbent and starts meaning an incumbent needing the subnets outputs, which is a larger win, and a much shorter conversation. The token becomes access to the commodities produced by the network.

What to ask of the next subnet you look at

What does this network produce, separately from whatever the company sells? Who has to spend alpha to get at it, and could they get the same thing another way? And is that gate live, or is it a plan for once the revenue shows up?

Leadpoet has clear answers for all of these and positions it to completely disrupt the sales technology industry.

leadpoet.com

https://beyondfinance.pro/