research.evertech
04 — Executive summary · draft scaffold

Executive summary

Deliberately left as a scaffold, as agreed — this gets written once the European pipeline is researched and we know what the recommendation actually is. What follows is the raw material: the findings that would go in it, and the questions that have to be answered before it can be written honestly.

Holding position

The three findings that would lead this page

Finding 01

Only one of the five companies runs a motion we could actually copy

fonio is the only business in the set built on paid performance marketing into an SMB base — 24.2% of its traffic is paid, it is the only one that is Facebook-first, and it made its ad spend self-funding through €300 prepaid credit packs before raising anything. Sierra, Fin and Ada are top-down enterprise motions where CAC is a salesperson and a nine-month cycle. Their playbooks are interesting; fonio's is transferable.

Finding 02

The whole category is fighting a competitor-conquest keyword war — and fonio found the flank

Ada, Fin and Sierra all buy each other's product names on paid search. fonio buys twilio, vapi, retelldeveloper infrastructure — intercepting buyers at the moment their alternative is building it themselves rather than buying a rival. Cheaper traffic, weaker competing offer, higher intent. This is the single most directly copyable mechanic in the research.

Finding 03

Pricing model is the category's real weapon, and it has a correct moment

Every player moved pricing inside a 26-month window, and the moves are not converging: Fin holds $0.99/outcome and publishes rivals' prices as marketing; Sierra charges nothing when the agent fails; Ada tried outcome pricing and publicly reversed, arguing it punishes success; fonio moved credits → subscription in Feb 2026 and got 30%+ MoM growth out of it. What settles the argument is the definition of the billable event — and Fin's "assumed resolution" bills customer silence as success, which is the cautionary example.

Before this page can be written

Open questions

Blocking — research
  • Quidget / SupportYourApp — now the top research target, ahead of everything else. The closest structural peer to Everhelp in the entire dataset, and the only one that has made the decision we are weighing.
  • Three more European companies. The brief asks for at least five. With Hoory (Armenia, $2M/26) and Quidget added from the deck we are at four; aaron.ai, Cognigy, Parloa remain the evidence-backed priorities — see the pipeline.
  • Fix the deck's "Fin AI" rows before it circulates — headcount, revenue, traffic and LinkedIn tone are measuring The Fin AI, a financial-LLM research group, not Intercom's Fin.
  • ARR band coverage. Only fonio sits in the $2–30M target band today. Sierra and Fin crossed it recently enough to be documented; Ada crossed it in 2019–20; My AskAI is below it.
  • fonio's −23.4% traffic month. Needs a September re-pull before anyone repeats it as a trend.
  • The creatives themselves. Ad-library video can't be embedded; capturing and self-hosting fonio's actual ads is a ~1 hour pass and would make the paid section concrete rather than described.
Blocking — decisions we don't have yet
  • What is this research for? Benchmarking Everhelp's own GTM, evaluating entry into AI support as a product line, or competitive defence against AI support eating outsourced support demand. The recommendation is different in each case, and this page can't be written without knowing which.
  • Which metric are we optimising? If it's EGP on a paid motion, fonio's loop is the whole answer. If it's category position, Parloa and Cognigy matter more than anything on pages 01–03.
  • Is the AI-support angle offence or defence for Everhelp? Every company in this research sells the automation of work Everhelp currently performs. A peer has already answered this: SupportYourApp — a Ukrainian support-outsourcing company, 16 years, 250+ clients, Inc. 5000 2026 — built and shipped Quidget AI, a productised AI support agent, ~$1M revenue on 8 people. Offence, as a separate brand, spun out of the services business. The open question is no longer whether but whether we can do it better than they did.
One thing worth saying out loud on Friday

Four of the five companies researched are selling the automation of human support work. Ada quantifies it as "one AI agent does the work of 10+ human agents at less than the cost of one"; Fin sells against a fully-loaded human cost of $5–20 per ticket at $0.99. Whatever we take from their GTM playbooks, the demand-side story they are all telling is a story about replacing the service Everhelp provides. That deserves to be a deliberate strategic conversation rather than a footnote in a marketing teardown.