Recommendations — research.everhelp

Verdicts are our judgement on the evidence, not the companies’ claims. “Free” means it costs time only; “needs investment” means a capability or price change we do not have today.
#	Recommendation	Verdict for Everhelp	What we copy	The one thing that could go wrong
01	Sell before you build	Adopt now	Paid pilots sold to our own support clients, before the product is finished	A slipped pilot damages a paying services retainer, not just a SaaS deal
02	Founder closes the first hundred	Adopt	One named owner with the pen, and a searchable objection log	Account managers hear “yes” too easily; a warm base gives polite, useless objections
03	Run the engine you already know	Adapt — don’t copy	The honesty, not the conclusion: name our engine correctly before betting on it	Our inherited engine is the anchor base — BlueTweak’s, and it rolled off into a loss
04	Meta creatives + prepaid credits	Highest upside, needs investment	fonio’s exact pairing — SMB creatives plus cash-before-cost packaging	We don’t have this muscle, and prepaid billing is a compliance surface
05	Publish your price	Adopt — product tier only	Public price and self-serve signup for the product, no mandatory demo	A public €99 anchors our services retainers downward
06	Win one language market	Adopt	One language, one owner holding the ad account, no second market until won	The good markets are occupied, and our own client base is English-speaking
07	Remove buyer risk, don’t discount	Adopt — our unfair advantage	Money-back and outcome guarantees, backstopped by our own human agents	The backstop is real COGS — it inverts margin if the resolution rate underperforms
08	“% of tickets solved automatically”	Adopt — holding-level decision	Lead with an audited resolution rate; price and guarantee on the same number	It cannibalises the services P&L by design — product vs parent margin

The recommendations

Recommendation 01

Sell the product before you build it

Take real money for the thing before the thing exists. You are not testing an idea — you are finding out whether a real person will pay, and hearing exactly why they will not.

My AskAI pre-sold ~$5,000 to ~50 customers at $99 each — via a “coming soon” pop-up inside their previous, struggling product, before writing any code. The pre-sale made 2× what that product had earned in its lifetime; they killed it and built the new one.
Ada’s founders took frontline support jobs at seven companies for a year, handled thousands of tickets, and secretly tested v1 inside one of those employers before building the company.
Sierra signed four Fortune-scale design partners on the founders’ track record alone, then built the agents for them by hand.
It costs time only — no product, no capital — so there is no good reason to skip it. Skipping it is how you build the wrong product confidently.

For Everhelp Adopt now

What we copy: package the AI agent as a paid pilot and sell it to our own existing support clients before it is finished — one page, one price, one invoice. We are the only company in this comparison that can run My AskAI’s test without having to find an audience first.

Pros

Zero CAC for the test. The 50 pre-sale conversations My AskAI had to earn on Reddit are already in our CRM.
The cheapest possible answer to the real question — not “is AI support interesting” but “will a client who already pays us for humans pay us again for software”.
A paid pilot deepens the services relationship instead of competing with it, and gives us reference logos on day one.
Costs time only. No product, no capital, and it produces the objection log for recommendation 02.

Cons & risks

Delivery risk lands on the services account. If the pilot slips, we do not just lose a SaaS deal — we damage a paying retainer. Pick clients who can tolerate a beta and say so in writing.
The services trap: a friendly client will ask for custom work, and “yes” to each request turns the product back into an agency project. Fix the scope before the first invoice.
Biased signal. Our clients buy because they trust us, not because the product wins on merit — this is precisely how BlueTweak mistook its captive base for product-market fit.
Pilot revenue books like services revenue. Decide up front how it is counted, or the SaaS metrics start dishonest.

Recommendation 02

The founder closes the first customers personally — and writes down every objection

No sales team and no agency at the start. The founder does the calls, and logs every question, every push-back, and the exact words customers use.

fonio’s founder ran demo calls 14 hours a day and personally closed ~1,000 clients and ~500 partners. That objection log later became the ad copy, the pricing tiers and the onboarding flow.
Chatarmin and My AskAI show the founder is the early channel: 100% of Chatarmin’s social traffic is the founder’s LinkedIn, 100% of My AskAI’s is the founders answering questions on Reddit.
Skip it and you get revenue but no asset: you can outsource selling later, but nobody else will learn the customer’s language for you.

For Everhelp Adopt — with one named owner

What we copy: one named person closes the first hundred and keeps the objection log. Not a team, not the account-management function — one owner with the pen.

Pros

We already speak the customer’s language. Our agents work these tickets daily — that is the year of frontline work Ada’s founders had to do on purpose, and we get it for free.
The objection log converts directly into ad copy, pricing tiers and onboarding — the same asset fonio built with 14-hour demo days.
It gives one person accountability for a number, which is what turns “an AI initiative” into a business line.

Cons & risks

Account managers are the wrong closers. They are paid to protect retention, so they will hear “yes, interesting” and log a win where a founder would hear “no, because…”.
It consumes our most expensive people for months, with no leverage and no headcount relief.
A warm base produces polite objections. Force at least a third of the hundred to be strangers, or the log is worthless.
If the owner leaves, the asset leaves with them unless the log lives in a system from day one.

Recommendation 03

Run the marketing engine you already know how to run

Be honest about what kind of team you are. Choose the acquisition motion you can execute on Monday — not the one that looks best in a strategy deck.

Six of the eight European companies inherited their engine from the founders’ previous business: fonio from a performance-marketing agency with a ~€3M ad budget, Trengo from a web agency, Parloa from a conversational-AI agency, Chatarmin from a paid-education business.
BlueTweak never had an engine at all — its first customers came free through its parent BPO, and when that ran out it bought cheap comparison-site traffic for an enterprise product. Average visit: 7 seconds; revenue down two years running.
Stonly tried US content marketing with a two-person team out of New York — and has not raised since January 2022.

For Everhelp Adapt — do not copy literally

What we copy: the honesty, not the conclusion. Our inherited engine is relationship-led B2B services selling into a warm base — and this research shows that engine has an expiry date. BlueTweak is our archetype, and it is the cautionary one.

Pros

Naming it correctly is worth the whole slide. Six of eight European companies inherited a marketing engine; we inherited a sales engine — a different thing, and better to know now.
The warm base is genuinely the fastest route to the first ~€500k of product revenue, and it is free.
Support operations, QA and workforce management are a decade of accreted expertise we cannot be undercut on — the same depth BlueTweak had, and the part of it that was real.

Cons & risks

The anchor gets consumed, not compounded. BlueTweak hit €1.92M on its parent’s network, then fell 14% and 19% in consecutive years into a loss, headcount still climbing. That is the exact shape of an anchor rolling off.
We do not currently own the engine this category actually rewards. Recommendation 04 is a capability we would have to build or buy, not one we inherit — treat it as an investment decision with a real number attached.
Services selling is high-touch and low-volume. It will not fill an SMB funnel at a €99 price point, and pretending otherwise wastes a year.
Do not do what BlueTweak did next: bolt cheap competitor-alternative SEO onto an enterprise sales motion. Average visit, 7 seconds.

Recommendation 04

Try Meta creatives + prepaid credits as the paid engine

Buy Facebook/Instagram ads aimed at small businesses, and charge prepaid credits — so the customer’s cash arrives before your costs do, and every sale funds the next ad.

fonio is the only European company in the set with a real paid engine: 24.2% of all its traffic is paid — 2.3× the next company and 13× the median.
68% of fonio’s social traffic is Facebook, and half of it is bought Meta clicks. No other European company in the research spends a euro there — the auction for SMB buyers is uncontested.
Prepaid credits from €79 made the loop self-funding: this carried fonio from zero to ~$10M ARR in about 18 months, before it raised any money.
The natural fit matters: this engine works because the buyer is small enough to purchase from an ad without a sales call.

For Everhelp Highest upside — and our biggest capability gap

What we copy: fonio’s exact pairing — creatives aimed at SMB buyers, and prepaid credits so the customer’s cash arrives before our cost does. This is the single strongest correlation with speed in the whole research, and the one thing no other European company in the set is doing.

Pros

The auction is uncontested. 68% of fonio’s social traffic is Facebook and no other European company here spends a euro on it. Being second into an empty auction is still early.
Prepaid is negative working capital. Cash before delivery means CAC payback stops being a constraint — it is what carried fonio from zero to ~$10M ARR before it raised anything.
It is the only motion in this research that funds its own growth, which means we can test it without a funding round and keep the option to raise on our terms.
Our support operation can absorb the volume of small customers this model creates — a cost problem for a software company, an existing competence for us.

Cons & risks

We do not have this muscle today. fonio’s founder came from a performance-marketing agency with a ~€3M budget. Buying the capability is the real decision here, and it is not cheap.
Compliance surface. Prepaid credits plus auto-topup is exactly where unexpected-charge and refund risk lives. Expiry, top-up and refund terms must be unambiguous before the first ad runs — nothing that could read as a grey mechanic.
It caps ACV. A €79 entry price means thousands of customers to reach $10M — a very different company from the one we run today.
The loop buys acquisition, not retention. fonio’s own NRR is under 100% while it spends seven figures a month. Model payback and ROMI before volume, not after.

Recommendation 05

Publish your price

Put the price on the website and let people buy without talking to anyone. Paid advertising only compounds when the click can convert on its own.

The two fastest-growing companies both publish: fonio at €79–499/month and Fin at $0.99 per resolution — which let 12,000 customers sign themselves up.
Sierra and Ada keep pricing unpublished, so every deal needs a sales team. Sierra can afford that with Fortune 50 buyers; Ada has been stuck at ~350 customers since 2021.
A hidden price sends SMB buyers straight to the competitor whose price they can see.

For Everhelp Adopt for the product tier only

What we copy: a public price and a self-serve signup for the product — no mandatory demo. This is a precondition for recommendation 04, not an independent choice: a bought click cannot convert on a “book a demo” page.

Pros

It is the gate on the whole paid motion. fonio and Fin publish and grow fastest; Ada hides its price and has been flat at ~350 customers since 2021.
Removes sales cost from the small end entirely, which is the only way a €99 customer is profitable.
It forces a productised scope. Publishing a price means deciding what is in and out — the exact discipline that separates a product from a services engagement.

Cons & risks

It anchors our services pricing downward. An enterprise client who sees €99/month will ask what they are paying a retainer for. This is the most under-estimated cost on this page and it needs an answer before launch, not after.
A public price is a public target. BlueTweak’s €65 all-in wedge undercut Zendesk by ~65% on paper and still lost — cheap is a position better-funded rivals erase with one pricing-page edit.
It exposes our margin structure to competitors who currently cannot see it.
Two price lists (product public, services private) is a real operational and messaging burden.

Recommendation 06

Win one language market completely before you touch a second

Pick one language, translate everything — ads, site, support, invoicing — and stay until you are the default answer there.

fonio went German-first: German ads, German site, German support. 63% of its traffic is European and the German-speaking market was won before any English push.
Watermelon proves both directions: winning the Netherlands (68% of traffic is Dutch) made it profitable — and never leaving is why it stopped growing.
Stonly skipped its home market for the US and ended up with neither — its French traffic is mostly its own customers’ help centres, and it never built a US position.

For Everhelp Adopt — the market choice is the actual decision

What we copy: one language market, everything translated — ads, site, support, invoicing — with one owner who holds the ad account, and no second market until the first is won.

Pros

Concentration beats spread at our budget. One language means one auction and one clean ROMI read, which is the only way to learn anything from a first paid test.
fonio’s per-market owner model produced +30–66% growth per country. The mechanism is ownership, not translation.
Watermelon proves the upside honestly: winning the Netherlands (65% of traffic) is what made it profitable at all.

Cons & risks

The good markets are occupied. Germany is fonio’s and VITAS’s home; the Netherlands belongs to Watermelon and Trengo. Entering either means a head-on fight with an incumbent that has the local brand.
Direct conflict with recommendation 01. Our existing client base is international and English-speaking. “Win one language market” may mean building a funnel aimed at people who are not our current customers — decide which of the two comes first.
Watermelon is also the ceiling: 65% single-country concentration made it profitable and stuck at $2–5M after eleven years.
Real localisation is a cost line — native support hours, local invoicing, local ad creative — not a translation project.

Recommendation 07

Remove the buyer’s risk instead of cutting the price

When a buyer hesitates, do not discount. Take the risk onto your side of the table and keep the price intact.

Fin offers a $1M resolution guarantee — miss a 65% resolution rate and they pay you. Sierra makes every escalation to a human free.
fonio stacks three of them: a 100% money-back guarantee, entry at €79 prepaid, and no contract — the buyer risks nothing but an afternoon.
Ada is the only US player with no risk reversal — and the one that is stalled.
Discounts attract the price-sensitive. De-risking attracts the sceptical-but-serious — the ones who renew.

For Everhelp Adopt — and we can go further than any of them

What we copy: fonio’s 100% money-back guarantee and no contract, plus Fin’s outcome guarantee — but backed by something they do not have. If the AI does not resolve it, our humans do.

Pros

This is our one genuinely unfair advantage. No pure-software rival can promise a resolution and staff the fallback. A guarantee we can actually honour is a moat; a discount is not.
It keeps price intact, so it protects ROMI while still converting the sceptical buyer — the opposite of discounting.
It answers the category’s loudest objection (“what if the AI gets it wrong in front of my customer”) with an operational fact rather than a reassurance.
Ada is the only US player with no risk reversal, and the only one stalled. The correlation is worth using.

Cons & risks

The guarantee has real COGS. Every unresolved ticket we backstop with a human is margin, and if the resolution rate underperforms, gross margin inverts. Do not offer it before recommendation 08 is instrumented.
It can be gamed — by customers with unusually hard queues, or by ones who use the human fallback as the actual service. Cap it and define it.
Guarantee wording is a claim-substantiation surface. It must be provable, dated and honoured without argument, or it becomes a brand and compliance problem.
A human backstop weakens the “AI-first” story to a sceptical buyer if it is presented as a crutch rather than a warranty.

Recommendation 08

Make “% of tickets solved automatically” your key metric

The share of tickets the AI resolves with no human touching them is the number this whole market sells, prices and competes on. Lead with it in your marketing — and price on it if you can.

Ada’s homepage headline is the metric itself: “autonomously resolve up to 83% of support issues.” The number is the pitch.
Fin built its entire marketing on it: a live counter of 31.6M conversations resolved, a public chart showing its resolution rate climbing “1% every month” (28%→65%), and a head-to-head — Fin 73% vs rivals’ ~50% — that makes buyers evaluate everyone on Fin’s yardstick.
Sierra and Fin go one step further and price on the metric — per resolution — so the sales pitch, the invoice and the guarantee are all the same number.
The buyer is not buying “AI” — they are buying fewer tickets a human has to touch. One warning: measure it honestly. A deflected ticket is not a resolved one, and Fin already taught the market to ask the difference.

For Everhelp Adopt — but this one needs a holding-level decision

What we copy: lead with the percentage of tickets resolved with no human involved, and price on it if we can. We can measure it more honestly than anyone in this research — and that is both the opportunity and the problem.

Pros

We own the data the software vendors have to fake. Real historical tickets across real clients, with human-verified outcomes — so we can publish an audited resolution rate where Fin’s own chart is footnoted “testing conducted by Fin customers”.
One number does three jobs: the marketing claim (rec 08), the pricing unit (rec 05) and the guarantee (rec 07). That alignment is what makes Fin’s and Sierra’s positioning hard to answer.
It is the honest metric for the AI question too — if the number does not climb, the thesis is wrong, and we find out early and cheaply.
Deflection-vs-resolution is a distinction we can prove and competitors mostly cannot. That is a category-leadership wedge, not a feature.

Cons & risks

It cannibalises the services P&L by design. Every point of automation is billable human hours removed. This recommendation puts the product against the parent business, and it needs an explicit decision at holding level about which margin wins — not a marketing decision.
Publish it and you cannot unpublish it. A weak early number becomes a public liability, and a strong one invites audit.
Measurement first, spend second. fonio is the warning: seven figures a month on “gut feeling”, dashboards that disagree, NRR under 100%. Instrument resolution rate, ROMI and payback before scaling anything.
Pricing on resolutions transfers volume risk to us and makes revenue harder to forecast than a seat licence.

Eleven companies. Eight recommendations.

What Everhelp should copy, and what it costs

Sell the product before you build it

The founder closes the first customers personally — and writes down every objection

Run the marketing engine you already know how to run

Try Meta creatives + prepaid credits as the paid engine

Publish your price

Win one language market completely before you touch a second

Remove the buyer’s risk instead of cutting the price

Make “% of tickets solved automatically” your key metric

Four ways companies in this research lost

Building where you cannot sell

Cheap traffic, wrong buyer

Being slow while well funded

Being early, then re-pricing too late