enso's $15M Round Backs Human Operators, Two Years After It Sold $29 AI Agents

On October 1, Tel Aviv's enso raised a $15 million Series A led by MoreTech Ventures, with NFX, Phenomen VC and Gil Hirsch joining. According to Calcalist, that brings its total funding to $23 million. The 26-person company sells AI agents that work marketing and sales channels for companies including Papaya Global and MyHeritage, and Calcalist reports it embeds engineers and marketers with its clients.

The round is paying for a different product. Two years ago enso sold cheap agents as a substitute for an agency. Today its flagship service has no self-serve plan, and a named person runs each customer's agents. For anyone buying "agentic" growth, that path is the useful part of the story.

Two years from $29 agents to a named operator

When enso came out of stealth in July 2024 with a $6 million seed round led by NFX, it offered more than 1,000 prebuilt agents across roughly 70 industries, priced at $29 to $79 a month each. Founder Mickey Haslavsky, a co-founder of RapidAPI, pitched them against agency fees, telling TechCrunch that most of the services they automated would cost about $1,000 a month from an agency. In February 2025 the company launched an agent marketplace for small businesses at a flat $49 a month.

The scope widened as the price disappeared. Calcalist describes agents across search visibility, visibility in AI answers, sales outreach, online communities, newsletters and social media. They’re built around each customer's existing workflows, and people keep control of strategy and sensitive decisions. Together they cover what a growth agency's org chart would, with software doing the repetitive work in between.

The service it now leads with reads differently. enso's Agentic SEO page says "enso is not software you sign up for” and lists no price. It describes a forward-deployed team that runs each agent while a named operator tunes it every week against the customer's numbers. On what enso calls a six-hour loop, the agents probe ChatGPT, Perplexity and Google for gaps and draft sourced answers in the brand's voice. They then publish through the customer's CMS and re-check until a page is cited. Publishing runs "at the volume you approve."

The company's best evidence is a controlled test. In a May write-up, enso restructured half of 142 existing pages that ranked in positions 6 to 25 and left the other half as controls. Over 60 days, it reports 6.4 times the cited mentions across 11 answer engines versus the classic-SEO baseline. That figure is vendor-reported and self-run, so treat the multiple with care. The design is what's worth copying.

The agents still ship without a human, and the category is betting on software

enso hasn't left self-serve automation behind. Its Autoblog page promises a blog that researches, writes, designs and publishes optimized content every day, automatically, with no review step described. Haslavsky's own framing of the round is about machines, too: agents, he told Calcalist, "make it possible to monitor multiple platforms continuously, test where opportunities are emerging and act while those opportunities still matter."

The category's biggest money is also going to software. Profound raised a $180 million Series D in September at a $1.8 billion valuation to build an "AI Marketer" for more than 1,000 enterprise brands. One reading of enso is that the operators are a sales layer for large accounts, which fades as the agents improve.

That may prove right, but it isn't how enso is selling today. Its enterprise customers are buying a team with agents, and the price of that team isn't on the website. The 2024 pitch compared a $29 agent with a $1,000 agency. The 2026 offer puts a named human back in the loop, which means the real comparison is now with the agency itself, on cost per result rather than cost per seat. For a buyer, that makes it a services contract, whatever the pitch deck says.

What this means for buyers of AI growth agents

Before signing with any vendor that runs agents on your channels, ask who the operator is, how many accounts that person covers, and what share of the fee pays for people rather than software. Those three answers tell you whether you're buying a tool or an agency.

Then write the measurement into the contract. Ask for treated and control pages, a fixed prompt set re-run across the answer engines, and citation lift reported per page. enso's own experiment shows the format, so a vendor in this category has no reason to refuse it.

Last, decide who on your side approves what goes live. "At the volume you approve" only works if someone owns the approving.

My bet is that within a year, the AI growth vendors closing enterprise deals will publish their operator-to-account ratios the way agencies publish team structures, because that's what buyers will ask to see first.

The 2026 offer puts a named human back in the loop, which means the real comparison is now with the agency itself, on cost per result rather than cost per seat.

Write the measurement into any AI growth agent contract: treated and control pages, a fixed prompt set re-run across the answer engines, and citation lift reported per page.

By Business of GTM