Skip to content
verastack/labs
Launch kit Start a project
← all notes

18 October 2026 · 4 min read

Funded this fortnight: six AI agents for the back office

$136.8M went to six startups whose AI agents quote, settle invoices and test ERP systems. Who raised from whom, 3 to 17 October 2026, and how AI search reads them.

The loudest money of the last two weeks went to software that does the dull work. Not copilots that suggest, but agents that own a task end to end: pricing a manufacturer's quote, settling an invoice dispute, testing an SAP migration, keeping product videos up to date.

Six of them raised between 3 and 17 October 2026, $136.8M between them, from Benchmark, Index, Balderton, Base10 and a16z speedrun among others.

Startup What its agents do Round Led by
Hone Own a business goal for months $60M seed Benchmark, Index Ventures
Duvo Run approved steps across ERP, portals and email $32M Series A Balderton Capital
Causa Prima Settle invoice disputes, agree early payment $20M seed Base10 Partners
Clueso Make and update product videos and guides $10M Series A Dharana Capital
Forgepoint Price manufacturers' quote requests $9.8M seed Primary, a16z speedrun
Shamrock AI Test ERP migrations such as SAP S/4HANA $5M seed Endiya Partners

The six

Hone raised a $60M seed, announced 8 October, from Benchmark and Index Ventures. Its "Engines" are agents that take on a long-running goal, such as sales conversion or procurement cost, and work at it for weeks or months instead of answering one prompt. CEO Moritz Stephan was chief of staff at Cognition.

Duvo raised a $32M Series A led by Balderton Capital on 14 October, with Index Ventures, Northzone and Credo Ventures, bringing it to $48M. The Prague company calls itself an enterprise execution OS: people and agents run governed processes across ERP systems, supplier portals, spreadsheets and email. Its CEO, Tomáš Čupr, also founded the grocery company Rohlik.

Causa Prima raised a $20M seed led by Base10 Partners, announced 13 October, with QED Investors and FJ Labs joining Creandum, Kfund and Borusan Ventures. Its agents act for buyers and suppliers, settle invoice disputes and negotiate early payment. CEO Maex Ament founded Taulia before it; the team works from Madrid and Munich.

Clueso raised a $10M Series A led by Dharana Capital on 13 October, with Founders Future, Neotribe, Y Combinator and f7 Ventures. Its agents create product videos and step-by-step guides for software companies, and keep them current as the product changes.

Forgepoint raised a $9.8M seed led by Primary Venture Partners and a16z speedrun on 14 October. Its agent, Frank, reads a manufacturer's requests for quote, prices them with that manufacturer's own cost model and drafts the reply for a person to approve. Jack Strauss and Alex Pareto run it from New York.

Shamrock AI raised a $5M seed led by Endiya Partners, with Capria Ventures, on 15 October. It sells "agentic assurance", starting with AI test automation for SAP S/4HANA migrations. Founders Sarat Addanki, Anuraag Nallapati and Mick Gallagher build it from Shillong.

Why a funding round is an AI search moment

A round is the week the most people look a company up at once: buyers checking the vendor, candidates checking the employer, journalists checking the facts. A growing share of them ask ChatGPT, Perplexity or Google's AI answers instead of searching, and the answer is only as good as what the model can read on the company's own site.

So we ran all six homepages through our AI search check on 18 October 2026, the same sixteen checks as our YC Summer 2026 and SaaS studies.

  • Clueso: 100. Every check passes, from plain HTML that says what the product does to a marked-up FAQ and an llms.txt. A video company with a homepage that reads perfectly as text.
  • Duvo: 90. Everything a model needs is there. The one note: robots.txt turns away a training crawler, which is a fair choice, not a mistake.
  • Causa Prima: 82. Organization data, a marked-up FAQ and almost 2,000 words of readable HTML. What's missing is structured data that describes the product itself.

On the other three homepages, the gaps were the same three: no structured data that describes what they sell, no FAQ an engine can quote, and Organization data that is missing or links to no other profiles, so a model has to guess which of several namesakes it is reading about. None of that is hard. For a company that just raised, each is about a day's work.

Check your own

If you raised this month, or plan to, run your homepage through the check below before the announcement goes out. It's free and takes about ten seconds. If you'd rather have us make the fixes, tell us about your launch.

How does your site read to AI search?

Three of these six freshly funded startups scored 82 or more. See where your homepage lands.