# Message test — https://kylon-main-landing-mockup.lesha27109313.chatgpt.site/ After reading your page, only 3 of 15 personas could name a reason to pick you over a similar option. - **Page tested:** https://kylon-main-landing-mockup.lesha27109313.chatgpt.site/ - **Audience tested against:** CEOs, Founders, CMOs from series A start ups - **Personas:** 15 simulated - **Report:** https://grader.wynter.com/r/kylon-your-people-and-ai-finally-on-the-same-t-dVRls8E > These answers are generated by AI, scored on Wynter's B2B Message > Layers framework using behaviorally-diverse simulated personas. The > methodology is real and the critique is directional. What a simulated > persona cannot have is a live budget, a renewal coming up, or a boss > asking about this quarter. --- ## 01 · The scores Every persona answered all four questions. These are four independent proportions of the same panel, not stages of a funnel. | Layer | Question | Cleared the bar | Strength | Of those who passed | | --- | --- | --- | --- | --- | | 1. Clarity | Do they understand what you do? | 15/15 | 78% | all with reservations | | 2. Relevance | Can they tell what it solves, and who it's for? | 15/15 | 91% | 9 without hesitation, 6 with reservations | | 3. Value | Do they actually want it? | 12/15 | 67% | all with reservations | | 4. Differentiation | Is there a reason to pick you over the alternatives? | 3/15 | 35% | all with reservations | **Brand alignment** (a side metric, not one of the four layers) — 8/15, 52% strength (all with reservations). Does the page read like the company you actually are? **Fix first: Differentiation.** Earliest failing layer, walking the sequence in order — not simply the lowest score. --- ## 02 · What to change, layer by layer Ordered worst-first. Specific edits, not a restatement of the score. ### Differentiation **Under 'More control than Slack + Zapier', name what those tools cannot block or log.** The headline asserts more control without saying what control means in practice. Spell out that Zapier fires an action with no named approver and no record of who authorised it, and that Kylon blocks the send until one exists. *effort medium · impact high · tested against Give a reason to choose you* **Move the 'Approval is enforced' point up beside the hero product demo.** The enforced approval gate is the one thing that separates Kylon from Slack, Zapier, and autonomy-first tools, but it sits far down under 'Why Kylon'. Put blocking, named approver, and logged decision in the first screen beside the workspace mockup. *effort low · impact high · tested against Give a reason to choose you* **Replace 'Illustrative integrations' under the Slack and HubSpot logos with the real supported list.** Marking the integration row illustrative leaves a buyer unsure whether Kylon connects to their CRM at all. State which systems are live today and which are on the roadmap. *effort low · impact medium · tested against Proof next to the claim* ### Value **Replace the 'Illustrative workflow impact' 42-min-to-9-min numbers with one measured workflow.** Labelling the before-and-after as illustrative turns the strongest value claim on the page into a hypothetical. Show one real workflow with the firm's size and the measured time before and after. *effort medium · impact high · tested against Specifics beat superlatives* **Add one named customer quote near the demo CTA, from a small professional services firm.** Nothing on the page shows a real company like the reader's using Kylon, so the demo ask rests on assertions alone. One named EU firm of 11-50 people, the workflow it governs, and the result would carry the CTA. *effort medium · impact high · tested against Proof next to the claim* **Delete the 'Mock data · replace before launch' label and the three placeholder customer stats.** The page tells readers its own numbers are fake, so 12 hrs saved, 38% faster, and 2.4× count for nothing. Either publish real figures with the customer named or cut the block entirely. *effort low · impact high · tested against Proof next to the claim* ### Clarity **Rewrite the FAQ heading 'Is Kylon an AI agent or an agent builder?' to carry its answer.** The heading poses the question a reader most wants answered and leaves it inside a collapsed panel. Make the heading state the answer so a scanner gets it without clicking. *effort low · impact medium · tested against Headings stand alone* ### Relevance **Add a second workflow example from a professional services firm beside the hotel one.** The 12-property hotel group is the only scenario shown, and a services firm cannot see its own approval chain in it. Show a client deliverable or invoice going through the same approval gate. *effort medium · impact medium · tested against Name the audience* ### Brand alignment (side metric) **Remove all 'Illustrative' and 'replace before launch' labels across the page.** Build-note language tells an enterprise buyer the page is unfinished and the company is pre-launch, which contradicts a governance positioning. Ship only claims that can stand without a caveat. *effort low · impact high · tested against Answer the live objection* **Add an audit, security, or compliance detail under 'One reviewable record'.** A buyer evaluating a governance layer wants to know how long records are retained, who can export them, and whether access is role-restricted. State retention, export format, and who can view the log. *effort medium · impact medium · tested against Answer the live objection* **Replace the fictional 'Northline Stays' and 'RelayWorks' names with real logos or nothing.** Invented company names read as placeholder art and undercut claims of enterprise readiness. If no customer will go on record yet, remove the logo row rather than fill it. *effort low · impact medium · tested against Proof next to the claim* --- ## 03 · What is working ### The problem and the audience are clear on first read Nine respondents said the problem and intended reader were stated explicitly in the opening lines, with the hotel example doing the work and no inference required. Several also read the product correctly as an approval governance layer rather than a chatbot… > the eyebrow line "For operations teams scaling AI" tells you the reader immediately, and the subhead "Your people and AI. Finally on the same team." plus "approve real work before it runs" gives you the problem in the first two lines > > — Chief Executive Officer, Technology, 201-500 > The hotel example with the Guest Care agent blocked until "the Guest Ops lead approves" made the problem concrete in about two seconds > > — Chief Marketing Officer, Software and SaaS, 51-200 > The eyebrow line "For operations teams scaling AI" plus the subhead "Your people and AI. Finally on the same team." told me the audience before I even hit the fold > > — Chief Executive Officer, Technology, 201-500 > It's a governed workspace for AI agents that sits on top of tools we already have—letting agents draft work from company data (inbox, CRM, Slack etc.) but blocking any external action until a named human approves it, with a log of who approved what. > > — Chief Marketing Officer, Software and SaaS, 51-200 > the eyebrow line "For operations teams scaling AI" plus the H1 "Your people and AI. Finally on the same team." told me the audience, and the subhead spelled the problem > > — Chief Marketing Officer, Software and SaaS, 51-200 > The line "For operations teams scaling AI" is right at the top, and "Your people and AI. Finally on the same team" plus the subhead about building a "governed AI execution workspace" where you "approve real work before it runs" told me exactly what pain this hits > > — Chief Executive Officer, Technology, 201-500 > It's a governed layer sitting on top of AI agents — Kylon connects agents (built or brought-in) to company systems like Slack, Gmail, HubSpot, Salesforce, Notion, Drive, and then forces a human approval gate before any agent action actually executes externally > > — Founder, Professional Services, 11-50 > The hook that stuck with me was "14 manual touches to one human review" — that's the pitch I'd repeat to someone else. > > — Chief Executive Officer, Technology, 201-500 ### The enforced approval gate is the one thing respondents named as different Five respondents pointed to the approval gate — blocking, logging, audit trail — as a concrete, auditable mechanic that separates Kylon from Slack, Zapier, and autonomy-first AI platforms. It was the only differentiator anyone named. > the "1 human review" / enforced approval gate — "sending is blocked until the Guest Ops lead approves" plus "the source, approver, and action are added to the workflow record" is a specific, auditable mechanism > > — Chief Marketing Officer, Software and SaaS, 51-200 > The thing that'd actually tip me toward this one is the "Sending is blocked until the Guest Ops lead approves" mechanic plus the named-approver-and-log detail — that's a concrete enforcement feature, not just a promise of "governance." > > — Founder, Professional Services, 11-50 > The approval-gate mechanic — "Sending is blocked until the Guest Ops lead approves" plus the named-owner, logged-decision detail — is the one concrete differentiator that would pull me toward this over a generic "AI agent platform." > > — Chief Executive Officer, Technology, 201-500 > the approval-gate mechanic is a genuine differentiator if it's real, but until "mock data" becomes a name I can call, this page gives me no reason to pick Kylon over a rival with actual proof > > — Founder, Professional Services, 11-50 --- ## 04 · What the personas said ### The FAQ raises a question it never answers and hides detail Two respondents flagged the FAQ: the title asks whether Kylon is an agent builder without answering, and collapsed answers conceal the granularity needed to judge differentiation. One also found terminology inconsistently stacked across labels. > the FAQ header 'Is Kylon an AI agent or an agent builder?' is the one phrase that flagged ambiguity, because the page poses that as an open question rather than answering it inline > > — Founder, Professional Services, 11-50 > the FAQ answers are all collapsed with no preview text ("How do permissions and human approval work?+"), so I can't actually see if the approval model is granular (per-action, per-agent, per-integration) or just a single blanket toggle > > — Chief Marketing Officer, Software and SaaS, 51-200 > the stacking of category labels - "governed AI execution workspace," "agents," "workflow," "orchestration" all get used almost interchangeably, and none of them is the plain-English sentence I'd use to describe it to my own team > > — Founder, Professional Services, 11-50 ### The hotel example does not map to professional services workflows One respondent said the hotel scenario doesn't match their firm's actual approval workflows, despite others citing the same example as what made the problem clear. > the target customer implied by the example (a 12-property hotel group running guest follow-ups) isn't my world. Nothing about compliance frameworks, audit requirements, or the kind of approval chains a professional services firm actually deals with. > > — Founder, Professional Services, 11-50 ### The mock data label destroys every numeric claim on the page Eight respondents said the illustrative or mock-data labels on the before-and-after metrics, ROI figures, and customer logos invalidate the value proposition outright. They matched the problem but would not credit the numbers. > every number on the page is flagged "illustrative" or "mock data," so I have nothing real to anchor on — no named customer, no third-party audit of the approval log > > — Founder, Professional Services, 11-50 > the "mock data · replace before launch" label under the customer outcomes (12 hrs saved, 38% faster approvals, 2.4x more closed) tells me none of this is proven yet, just illustrative. > > — Chief Marketing Officer, Software and SaaS, 51-200 > the "proof" on the page is explicitly mock data — "Mock data · replace before launch" — so right now there's nothing to believe yet > > — Chief Executive Officer, Technology, 201-500 > the three customer logos (Northline Stays, RelayWorks, Fieldcraft) are flagged the same way - so I have zero actual proof yet, just a well-told hypothetical > > — Chief Executive Officer, Technology, 201-500 > The "12 hrs saved," "38% faster," "2.4x" stats are flagged as mock data on the page itself, so none of that is real proof yet > > — Chief Marketing Officer, Software and SaaS, 51-200 > the 42min-to-9min and "12 hrs saved" numbers are flagged as mock data, so I'd want real ones before I believed the ROI. > > — Chief Marketing Officer, Software and SaaS, 51-200 > One verified customer reference—an actual logged example of an approval workflow going from a messy multi-step process to a single human review, with real before/after times, not the mock hotel-group numbers—that's what would get this past a discovery call into an actual pilot. > > — Chief Marketing Officer, Software and SaaS, 51-200 ### Mock placeholders also cancel the differentiation the approval gate earns Three respondents said the placeholder customer examples disqualify the proof points, and one explicitly called the approval-gate mechanic a genuine differentiator undermined by mock data. The mechanic lands; the evidence behind it does not. > "Northline Stays," "RelayWorks," "Fieldcraft" with "12 hrs saved," "38% faster approval," "2.4x more requests closed" are flagged as placeholders, not real logos > > — Chief Executive Officer, Technology, 201-500 > the proof points sitting right next to it are flagged "Mock data · replace before launch," so I have zero evidence the 42-min-to-9-min or 14-to-1 numbers are real anywhere > > — Founder, Professional Services, 11-50 > the approval-gate mechanic is a genuine differentiator if it's real, but until "mock data" becomes a name I can call, this page gives me no reason to pick Kylon over a rival with actual proof > > — Founder, Professional Services, 11-50 ### The page reads as pre-launch rather than enterprise-ready Three respondents said the mock-data label and absence of named logos or sourced numbers signal an early-stage startup, which conflicts with an enterprise governance positioning. > a company with real customers wouldn't ship that to a live page, so this reads like a pre-launch or just-launched site > > — Chief Executive Officer, Technology, 201-500 > Early-stage, maybe seed to Series A, B2B startup — the single polished hotel-industry example plus "mock data, replace before launch" sitting right in the copy tells me this is pre-revenue-proof or very early customers > > — Chief Marketing Officer, Software and SaaS, 51-200 > The "12 hrs saved," "38% faster," "2.4x" stats are flagged as mock data on the page itself, so none of that is real proof yet > > — Chief Marketing Officer, Software and SaaS, 51-200 ### Respondents want a named customer like themselves before advancing Four respondents conditioned any next step on real evidence: a named EU professional services firm of 11-50 employees, verified workflow improvement, or real metrics replacing mock ones. One said that alone would justify a demo call. > I'd need a named customer in our size band (11-50 employees, EU, professional services) describing the exact bottleneck we have — manager-approval chasing — with a before/after number that isn't labeled illustrative > > — Founder, Professional Services, 11-50 > One verified customer reference—an actual logged example of an approval workflow going from a messy multi-step process to a single human review, with real before/after times, not the mock hotel-group numbers—that's what would get this past a discovery call into an actual pilot. > > — Chief Marketing Officer, Software and SaaS, 51-200 > That's worth a short demo call, but not worth a real commitment yet — those stats are flagged "mock data, replace before launch," so I'd take the meeting specifically to see our own workflow run through it and get real numbers, not their illustrative ones > > — Founder, Professional Services, 11-50 > Early-stage, maybe seed to Series A, B2B startup — the single polished hotel-industry example plus "mock data, replace before launch" sitting right in the copy tells me this is pre-revenue-proof or very early customers > > — Chief Marketing Officer, Software and SaaS, 51-200 --- ## 05 · The hardest read An adversarial pass over the findings. Every claim below was checked against the panel's own answers; unsupported ones were dropped. - **The page's only working asset is its clarity, and clarity is not a reason to buy.** *(high)* Eight respondents understood the problem and audience on first read, yet eight separately said mock-data labels invalidated the metrics, ROI, and logos. Comprehension is high and credibility is zero, so the page converts nothing. - **Placeholder proof costs more than it saves: it actively destroys the one differentiator the page earns.** *(high)* Five respondents named the approval gate as the only differentiator anyone could identify, and three said placeholder customer examples disqualify the proof behind it. The strongest idea on the page is neutralised by its own evidence. - **The page disqualifies itself from the enterprise governance category it claims to own.** *(high)* Three respondents read the mock label and missing named logos as early-stage startup signals conflicting with enterprise positioning. A governance product that cannot show audited, sourced numbers contradicts its own pitch. - **Next steps are blocked on a single missing artefact the page could supply today.** *(high)* Four respondents conditioned any advance on one named customer matching their profile or verified real metrics, with one saying that alone would justify a demo call. Everything else is in place and nothing moves. - **The FAQ actively subtracts from the page by posing the differentiation question and refusing to answer it.** *(medium)* Three respondents flagged that the FAQ title asks whether Kylon is an agent builder without answering, and collapsed answers hide the granularity needed to judge differentiation. It plants doubt in the exact area already weakest. - **The page rests its entire narrative on one example that breaks for the stated audience.** *(medium)* The hotel scenario did the clarifying work for most readers, but one respondent said it does not map to their firm's approval workflows. A single vertical anecdote carrying the whole problem statement fails the professional-services reader it targets. --- ## 06 · Who answered | # | Role | Industry | Company size | | --- | --- | --- | --- | | 1 | Chief Executive Officer | Technology | 201-500 | | 2 | Founder | Professional Services | 11-50 | | 3 | Chief Marketing Officer | Software and SaaS | 51-200 | | 4 | Chief Executive Officer | Technology | 201-500 | | 5 | Founder | Professional Services | 11-50 | | 6 | Chief Marketing Officer | Software and SaaS | 51-200 | | 7 | Chief Executive Officer | Technology | 201-500 | | 8 | Founder | Professional Services | 11-50 | | 9 | Chief Marketing Officer | Software and SaaS | 51-200 | | 10 | Chief Executive Officer | Technology | 201-500 | | 11 | Founder | Professional Services | 11-50 | | 12 | Chief Marketing Officer | Software and SaaS | 51-200 | | 13 | Chief Executive Officer | Technology | 201-500 | | 14 | Founder | Professional Services | 11-50 | | 15 | Chief Marketing Officer | Software and SaaS | 51-200 | --- ## 07 · Before you act on this The methodology is real, and the critique is directional. What a simulated persona cannot have is a live budget, a renewal coming up, or a boss asking about this quarter. **Validate anything you're betting on with real ICPs who are actually in-market.** Being wrong is more expensive than you think. Finding out is cheaper than you'd guess. Wynter runs message testing with verified B2B professionals — trusted by HubSpot, RingCentral, Shopify, Cognism, Paddle, Veeam, Rippling and Miro. This report is kept for 60 days from 2026-10-09, then deleted along with the personas and their answers.