The blind spot in outbound sales
Throxy crawled the UK's Companies House registry, 6 million businesses, Y Combinator's data shows, that LinkedIn's 1.1 billion users, Y Combinator reported, never see, and turned it into a 4-billion-contact database that now powers a $6.2 million seed round, StartupHub.ai's figures put, led by Base10 Partners. The London startup, founded in 2023 by Pablo Jiménez de Parga Ramos, Bergen Merey, and Arnau Ayerbe, graduated from Y Combinator's Spring 2025 cohort with $1.5 million in ARR, according to StartupHub.ai, built by three people. Today it employs 55 across London and Cape Town, running 60-plus human SDRs who place calls on behalf of clients selling into manufacturing, logistics, construction, and healthcare. Average output: 8 to 12 qualified meetings per month per client, with early cohorts seeing 20-plus in their first weeks. One client, Imnoo, AI quoting software for CNC manufacturers, reports three to five qualified meetings weekly. Across verticals, customers cite 84 percent positive reply rates and 15-plus hours saved per week on prospecting.
The blind spot Throxy exploits is structural. The tools that dominate outbound (LinkedIn, Apollo, ZoomInfo) were built for knowledge workers who live online. They miss decision-makers in legacy sectors who don't post on LinkedIn and don't reply to automated sequences. Sales teams chasing them burn 70 percent of their time scraping lists and writing emails that never land. Vertical AI agents are the first technology designed for that gap. Unlike horizontal sales automation, which sprays generic messaging across digital channels, these agents train on the data, language, and buying motions of a single industry. They mine public registries, government filings, and trade directories that conventional platforms ignore. They pair that intelligence with human sales development representatives who make the actual calls and write the actual emails. The hybrid model books meetings in sectors where pure-play AI has failed.
"We think the best messages, at least as in today, are generated by humans." (Pablo Jiménez de Parga, Throxy co-founder, VentureBeat, Sept 2025)
That philosophy — LLMs for data enrichment and campaign orchestration, humans for copywriting and relationship management — defines the product and the hiring plan. Throxy's pay-per-meeting service replaces internal SDR teams and tool stacks for B2B software companies selling into offline-heavy verticals. The model demands engineers who can build agents that process massive offline datasets while leaving the actual outreach to humans. The next section examines how that split is reshaping hiring, and what it pays.
Hiring for the seam between agent and human
| Role | Throxy Posted Range (USD) | London Market Avg (Glassdoor) | Location |
|---|---|---|---|
| AI Engineer | $80K–$100K base | $73,293 | London / Remote (UK, ZA) |
| Staff Engineer | $80K–$100K base | — | London |
| SDR | $80K–$100K base | — | Remote Global |
Throxy's disclosed salary band runs $71,666–$106,666 across all open roles, with the three most recent postings (AI Engineer (May 13, 2026), Staff Engineer (May 18), SDR (May 19)) clustering at $80K–$100K base. A separate UK listing pegs the AI Engineer role at £28,800–£48,000 (roughly $37K–$62K), suggesting the higher dollar figures reflect the company's Y Combinator backing and global remote policy. The average London AI engineer earns $73,293 per Glassdoor; Throxy pays a premium for sales-domain fluency.
The company's careers page describes itself as "a small applied research lab working on the future of sales" where researchers, engineers, and technologists "invent at the edge of what's useful and possible." The phrase signals a hiring bar closer to a research lab than a typical SaaS shop. The AI Engineer role requires fluency in retrieval-augmented generation, structured extraction from messy public records, and evaluation loops that measure meeting quality rather than email volume. Scoutify's 90-day hiring log shows one AI Engineer, one SDR, and one general Engineer hired: a ratio that mirrors the product's human-in-the-loop architecture.
Following its Spring 2025 YC graduation, Throxy plans to quintuple headcount from three to roughly 15. The new capital targets "additional human experts who can manage larger portfolios of clients while maintaining personalization quality," per the company's own statement. Open roles as of May 2026 include a Software Engineer Fullstack (London), Product Manager (London), and a Fractional SDR Programme (remote global). The SDR hire is notable: Throxy sells outcomes, not seats, so its own SDRs function as quality controllers for the agent pipeline, not volume dialers.
Contrast this with 11x, Throxy's YC cohort mate, which raised $50M to build fully autonomous AI SDRs. Throxy hires for the seam between agent and human. The market signal is clear: companies selling into manufacturing, logistics, and medical equipment — sectors where buyers lack LinkedIn profiles and tenure spans decades — need engineers who understand that data scarcity changes the model architecture. You don't fine-tune on email open rates when the ground truth is a Companies House filing from 2011. The engineer who can turn that into a qualified meeting pipeline — and knows when to hand off to a human — sets the new salary floor.
Incumbents embed agents, open marketplaces, price the uplift
Gong's partner ecosystem hit 250 technology and services integrations in November 2024, double the growth rate of the prior year, and the company now serves more than 4,000 customers worldwide. ARR crossed $300 million in March 2025 and topped $500 million by May 2026, a 55 percent year-over-year clip that suggests enterprises are buying the platform bet. The integrations are not cosmetic. LinkedIn Sales Navigator now surfaces inside Gong workflows, letting reps customize outreach without leaving the call-review screen. Gong landed on AWS Marketplace in late 2024, and a bidirectional sync for Microsoft Dynamics CRM is rolling out so Dynamics customers can capture communication activity, export Gong's AI-generated insights, and edit CRM fields in line.
"A strong ecosystem is essential to helping give our customers a competitive advantage in the market." (Amit Bendov, Gong co-founder and CEO, November 2024)
Customer logos validate the motion: DocuSign, Zendesk, Ironclad, and Exclaimer all selected Gong's Revenue AI Platform in the second half of 2025. LinkedIn's VP of Product, Monica Lewis, said the joint experience helps sellers "identify prospects that are most likely to make a purchase."
"Incumbents like Salesforce and HubSpot are wed to their seat-based pricing model… they are building AI features on top of their core, but have not built the product to support that from the ground up." (Caroline Broder, Base10 Partners, VentureBeat, Sept 2025)
Throxy's counter-move, a pay-per-meeting service backed by a 4-billion-contact database mined from public registries, targets the 99 percent of the economy those platforms overlook. That clash will shape the next hiring wave and the market ceiling.
Market outlook: a $13 trillion labor prize with hard ceilings
Worldwide AI spending is forecast to reach $2.59 trillion in 2026, a 47% year-over-year increase, with AI infrastructure alone accounting for over 45% of that outlay, said Gartner's John-David Lovelock in May 2026. Enterprises have yet to really flex their spending potential, Lovelock added; 2026 is the inflection year.
The total addressable market narrative has shifted. The opportunity used to be the $300–400 billion of annual global software spend. Now the frame is the $13 trillion of U.S. labor spend: a 30x expansion of software's TAM, as one investor put it in January 2026. Vertical outbound agents target a sliver of that labor pool: SDR and BDR roles. But the same dynamics (high turnover, staffing difficulty, compliance pressure) are pulling voice agents into healthcare (insurer calls, pharmacy coordination, post-surgery follow-ups, psychiatry intake), banking and financial services, and recruiting. Those verticals are explicitly outside this article's scope, which excludes manufacturing automation, healthcare diagnostics, and consumer-facing chatbots. The distinction matters: a voice agent that navigates a prior-authorization call with a payer is solving a different problem than one that drafts a cold email to a logistics VP. The former lives in regulated, high-stakes conversations; the latter lives in sequence optimization and deliverability.
Adoption signals are mixed. As of August 2026, 88% of companies use AI in at least one business function, up from 78% a year earlier. Ninety percent of UK retailers are actively exploring AI agents, with a third already implementing them. Nearly 68% of retail executives expect to deploy agentic AI for key operational and enterprise activities within 12 to 24 months. Deloitte estimates AI agents could handle up to 25% of global e-commerce sales by 2030, and referral traffic from ChatGPT and other AI chats already accounts for 15–20% of total referrals for some retailers. But trust remains thin: only 8.5% of people say they can always trust AI Overviews in search. Legacy systems are a brake: 44% of respondents say theirs slow innovation, demanding clean, connected data architectures before agents can plug in. And in high-stakes contexts, users still want a human in the loop at the end to approve, a pattern that shows up in both healthcare voice and B2B outbound where a rep reviews the booked meeting before it hits the calendar.
Geography introduces another limit. There are countries where humans are still cheaper on a permanent basis than best-in-class voice AI, which caps the addressable market for fully automated outbound in those regions. Multilingual capability is a strength: agents handle heavy accents and language switching well, but cost parity isn't universal. Meanwhile, the buyer preference for "solution, not technology" means many prospects want a managed service (Throxy's pay-per-meeting model) rather than a platform they must configure and monitor. That preference favors the vertical-agent-as-service model over pure software seats, but it also limits the total software revenue capture.
The vertical outbound agent market will grow, but its ceiling is set by the subset of B2B sales that are high-volume, scriptable, and tolerant of occasional hallucination — not by the total AI agent TAM. The factory owner in the UK still doesn't have a LinkedIn profile. The Companies House filing from 2011 still sits in a public registry, waiting for an engineer who knows how to read it. Throxy's database already has him. The question is whether the next hire can turn that record into a meeting, and whether the platform giants will buy the startup that does, or build the capability themselves.
Working in frontier tech? Zero G Talent tracks the openings: see every open ASML role, browse frontier tech jobs, openings at Stripe, and the people building the field.