🔒

Verity Agentic: Conversion Improvement Plan

The full 10-item plan for verityagentic.ai, with the reasoning behind each call. Enter hub password to continue.

Wrong password

Item 1 COMPLETE 2026-07-31

The site had zero client proof two days ago. Both references are now live.

	Quote	Placed on	Objection it answers
Bob Manthy, LPC, MS Computer Science, MBA. Founder, Rebuilding Seminars	Approved 2026-07-30. 15 years confirmed by Annette.	Homepage (between the proof cards and services) and the coaches page	"AI is hype" and "why wouldn't I just build this myself"
Ben, Founder, HourHound	Approved 2026-07-30 with his own edits	The Bottleneck Audit page (above the price) and pricing	"I'll end up dependent on her"

The relationship is disclosed, by Annette's decision. Bob's attribution carries "and my partner, who gave me the unvarnished version". Anyone who searches either name finds it in seconds, so saying it first reads as confidence while having it discovered later reads as an omission. Do not quietly drop that line.

Attribution rules Annette set for Ben: first name only, no photo, no logo, and the word HourHound links to hourhound.app.

Also shipped in this pass

The Automation Makeover sample on the audit page. Grok called this "the single best trust object you do not have": the site sold a $500 written plan and never showed the object. Now shows the fixed five-part shape with one priority worked through, using a coaching practice as the example.
"What happens after you get in touch", five steps, written against the ACTUAL flow (form, free call, then the paid audit) rather than a checkout that does not exist. Ends on the line the panel said non-technical buyers stall for: nothing else gets sold to you in between.
Boulder geo-targeting removed from the coaches page. Annette does not want that page targeting coaches and therapists in Boulder specifically. Stripped from the title, meta, keywords, og and twitter cards, the Service schema and the visible copy. areaServed changed from a City node to Worldwide, because a City node keeps telling search engines the service is local-only. The footer location line stays: that is where she is, not who she serves. Note the trade-off: this gives up local-intent ranking for terms like "Boulder therapist marketing". The framing was not from this session's work; it came from commit 552628d on 2026-07-13 when the page was created.

Two placement bugs caught by looking rather than by validating

Both would have passed every automated check:

Bob's quote first landed inside .offer-grid-4 on the coaches page, so it would have rendered as a squeezed grid cell instead of a full-width block. Found by measuring the rendered box against the grid, not by reading the HTML.
bottleneck-audit.html shipped 2026-07-29 with no </head>. Browsers auto-close it, so the page rendered correctly and passed the validator twice while still being malformed.

The lesson for whoever picks this up: GREEN on the validator means no overflow, no sticky, no em dashes and links resolving. It does not mean the markup is well formed or that a new component landed in the right container.

SHIPPED 2026-07-30 (second wave)

Deployed and validated GREEN over two runs, 16 pages.

Pricing contradiction resolved. Annette's call: the lower numbers win, matching the small-projects focus. Done-For-You build is From $1,500 (was $3,000 to $8,000 typical) and the AI Ops Retainer is From $500/month (was $1,500 to $3,000). Corrected on the pricing ladder, the Offer schema, offer descriptions, the homepage FAQ and its FAQPage schema, and llms.txt. The coaches page already carried these numbers and was the reference. Article prose still discusses $15,000 market rates, which is correct: that is commentary on what the industry charges, not a Verity offer.
Audit page un-orphaned. It was linked from exactly one page and missing from the nav (panel finding, 4/5 models, confirmed by crawl). Now: a THE AUDIT nav slot on all 36 pages, a homepage section directly under the hero, a linked FAQ mention, and the coaches page offer card.
Ben's testimonial live. Approved 2026-07-30 with his own edits. Placed above the price on the audit page and above the capacity note on pricing. Attribution per Annette: first name only, no photo, no logo, the word HourHound links to hourhound.app. One typo corrected in his text ("do his stuff" to "do this stuff"), flagged for her to confirm with him.
Bug fixed: bottleneck-audit.html shipped on 2026-07-29 with no </head>. The generator sliced the reused shell up to but not including that tag. Browsers auto-close it, so the page rendered correctly and passed validation while still being malformed. Tag inserted, generator corrected.

Deploy note: the wrangler OAuth token was revoked server-side while its local expiration_time still read as valid, so ./deploy.sh failed with "Invalid access token [code: 9109]" and whoami confirmed it. There is no Cloudflare API key in the PowerShell profiles, .bashrc, or any .env. The fix is npx wrangler login, which is Annette's step because it is a browser flow.

Still open from the panel

Not yet built, in the order the panel ranked them: a sample of the Automation Makeover on the audit page (called "the single best trust object you do not have"), a way to actually pay rather than routing every CTA to a free call, "what happens next" copy after booking, a how-it-works for the build tier, an explicit buyer sequence strip, and at least one example of an agentic workflow actually built for someone. Bob's testimonial goes on the homepage between the proof cards and services when it lands.

SHIPPED 2026-07-29

Items 2, 3, 4, 6 and 7 are live on verityagentic.ai, plus the ladder change and the naming. Validator GREEN over two runs, 16 pages.

Item	What shipped
2	`bottleneck-audit.html`, the storefront for the paid entry offer. Built by splicing into the `pricing.html` shell so the design is identical. Problem, three failed fixes, the offer, five named deliverables, price, four-step process, credentials, eight objection FAQs, closing CTA. `Service` + `FAQPage` schema. Linked from the pricing ladder, `sitemap.xml` and `llms.txt`, and added to `validate.config.json`.
3	One call to action. Nav button is the free audit across all 35 pages; `FREE REPORT` de-accented from a pseudo-CTA to a plain nav link (68 instances); the quiz block reframed as step one of the free audit rather than a rival offer.
4	Five pillars now carry a sentence each. Grid switched to `auto-fit` so a sentence has room.
6	Objection FAQ on the homepage beneath the informational set, which stays because it earns AI-search citations. `FAQPage` schema extended to 10 questions.
7	About page gains a method-origin section above the timeline, so Ethiopia, Kenya and Haiti explain the method instead of decorating the history.

Still not done from Phase 1: Item 1 (testimonials, Annette expects approval 2026-07-30) and Item 5 (application gate). Item 11 approved but not built.

Contrast, verified by hand

There is no automated contrast check, so this was measured manually on the live page. Everything passes AA. Two notes for whoever works on this next:

Do not trust a naive contrast script on this site. .paper-band carries its cream base inside a background gradient stack, not background-color, so any checker that only reads backgroundColor falls through to the black body and reports invisible text where there is none. Dark text on .paper-band actually measures about 11.5:1. A first pass at checking this page produced five false failures for exactly that reason.
One real finding, pre-existing and sitewide. .button-pink is white text on #ff1f8f, which measures 3.6:1. That passes AA for large text but fails the 4.5:1 floor at the 14px bold it currently uses, and it appears on the homepage, pricing, the coaches page and now the audit page. Not introduced by this work. Two cheap fixes that keep the brand: raise the button text to 19px bold (large-text threshold, only needs 3:1), or darken the pink for buttons only while leaving accents as-is. Annette's call, not a silent repaint.

The thesis in one paragraph

Steph Nemchik sells an adjacent service to a nearly identical buyer with weaker credentials than Annette's, and does it better because of five structural choices: a real sales page for her paid entry offer, client proof after every section, exactly one call to action, a named framework repeated on every page, and an application gate on the high-ticket tier. Verity Agentic has stronger proof and a better story with none of those five. Everything below is copy and structure. Nothing here requires touching the brand.

Explicitly out of scope: the noir black / yellow / pink / Impact / clip-path visual identity stays exactly as-is. Her cream-and-Cormorant palette is not to be borrowed. See "Do not repaint" in the teardown.

Working notes for whoever picks this up

Source of truth is public/ in this repo. Deploy with ./deploy.sh from repo root. The retired Codex upload folder is not to be touched.
There is no external stylesheet. Every page carries its own inline <style> block. New components must have their CSS added to each page that uses them. Check for an existing class before inventing one.
Existing reusable classes: .section-pad, .section-inner, .eyebrow, .button, .faq-home-list, .faq-home-item, .faq-home-q, .faq-home-a, .stats-heading.
Voice rules apply to every word written here: no em dashes, always use contractions, no "this isn't X, it's Y" constructions, no coaching-industry vocabulary. Load ~/memory/voice-profiles/annette-anti-ai-voice.md before drafting any long-form section, and run /no-ai-slop as the final pass on the sales page.
Run the validator before calling any item done: bash py -3.13 C:\Users\annet\ClaudeProjects\ValidateSite\validate_site.py --config C:\Users\annet\ClaudeProjects\VerityAgentic\validate.config.json
There is no automated contrast check as of 2026-07-29. Check new buttons and dark sections by eye against 4.5:1.

Phase 1: The two changes that move revenue

Item 1. Client testimonials, sitewide

Why first: it is the largest credibility hole and the cheapest fix. Every page of proof on this site is first-person (31,310 keywords, 4,000 dogs, 7 countries). That proves capability. It does not prove anyone has paid for this service and been glad. Those are two different objections and only one is currently answered.

Blocked on Annette: this needs real quotes with real names. Nothing can be fabricated, invented, or written on a client's behalf.

Status 2026-07-29: two references secured. Bob Manthy (Rebuilding Seminars) and Ben (HourHound, hourhound.app) have both agreed to be references and asked Annette to draft something they can approve or edit. Drafts are below. Neither is usable until the named person signs off in writing.

Further sourcing options, in order of likely speed: 1. The Boulder RISE coaches and therapists group (there is already a presentation deliverable aimed at them). 2. Any pilot or discounted build already delivered. 3. A peer or referral-partner vouch. Steph's single strongest quote is from Sarah Noked of The OBM School, who refers clients to her. A peer vouching outperforms a customer thanking. Worth asking who in Annette's network would say something on the record.

What makes a testimonial work for this specific business

Not a chorus, a portfolio. Two quotes that both say "she's great" do the work of one. Two quotes that kill two different fears do the work of four. The live fears for a 1-to-10-person owner spending $500 to $8,000 are:

AI is hype, and 95% of pilots fail (Verity cites this stat itself)
It will break the moment she is gone, and then I am dependent on her
She will bolt technology onto my business without understanding the business
I am not technical enough to own the result
Solo consultant, so will it actually ship

Rules the drafts follow:

One objection per testimonial. Assigned below.
A checkable specific beats any adjective. A number, a named artifact, a real before-state.
The first sentence must be a recognizable before-state, so the reader sees themselves and keeps reading.
It has to sound like the person, not like marketing. Bob and Ben are very different people and the two drafts should not sound alike. That difference is itself a credibility signal.
One concession or mild reservation makes the whole thing more believable than stacked superlatives.
Never "highly recommend." It is the phrase that signals a template.
Full name, business, role, and a link. Initials-only reads as invented.
Never claim a result that has not happened yet. Neither of these people has a traffic or revenue number to point at yet, so neither draft claims one. Both are credibility-of-diagnosis and does-it-run testimonials, which is what is actually true today.
40 to 90 words for a card. A peer vouch can earn more length.

Draft for Bob Manthy, Rebuilding Seminars

Objections it kills: "AI is hype and most of it doesn't work" plus "why wouldn't I just build this myself." Both are top-tier buyer fears and neither was covered before this revision.

Corrected 2026-07-29: Bob is not a non-technical client. An earlier draft opened with "I'm not a technology person." That is false. Bob holds a master's in computer science and an MBA alongside his LPC. Do not reinstate any version that frames him as non-technical; it would be wrong on the facts and a prospect who looks him up would catch it.

The correction makes the quote stronger. Someone qualified to evaluate the build, who could plausibly have specified it himself, choosing to have Annette do it is far better proof than a non-technical owner who had no alternative. It also converts him into the credible voice on AI hype, which is the objection Verity's own homepage raises (the MIT 95%-of-pilots-failed stat) and then leaves a stranger to resolve alone.

"I have a master's in computer science, so I can usually tell when someone's overselling technology. Annette wasn't. She went through 288 of my Google reviews and told me things about my own graduates that [N] years of running these seminars hadn't shown me. Then she built the thing I'd been describing for years: someone takes the assessment, the system scores it and sends that person a personal response, and nobody touches it. I understood exactly what she built, which is why I know how good it is."

Bob Manthy, LPC, MS Computer Science, MBA. Founder, Rebuilding Seminars

Put the credentials in the attribution line rather than stacking them in the first sentence. They do more work there and they keep his voice sounding like speech.

Bob needs to confirm: how many years he has run the seminars; that the assessment scoring email is genuinely running reliably now (there was a low-scorer delivery bug in the GHL automation on 2026-07-09 before the rebuilding-fdas worker was deployed on 2026-07-13, so verify before he signs off on "nobody touches it"); that he is comfortable with the 288-review number being public; and how he wants his credentials listed.

Short pull-quote version, for cards: "I have a master's in computer science, so I can usually tell when someone's overselling technology. Annette wasn't."

That short version may be the single strongest line available to the site. It answers the hype objection in two sentences, from someone with standing to answer it.

Draft for Ben, HourHound

Objection it kills: "I will end up dependent on her" plus "will she understand what I already built." This is the enablement testimonial and it is the more broadly useful of the two, because dependency is the fear that stops the most deals.

What the engagement actually was (corrected 2026-07-29; an earlier version of this plan mis-described it as a search audit). The week-one deliverable was a teaching document, not a findings report. It contains six paste-ready Claude Code prompts each attached to a specific fix, a primer explaining what SEO, GEO and GBP each are and which of the three actually applies to a directory like his, a sequencing judgment call (ship all three page fixes before submitting the sitemap, because Google's first crawl impression is sticky), a step-by-step Google Search Console setup, three analytics options with a recommendation, and a monetization map tagged by what works before traffic exists. One prompt teaches him to have Claude Code write a script that calls the GSC Indexing API to submit new venue URLs automatically.

Why that matters for positioning: Ben built HourHound itself with Claude Code. He was already fluent. Annette still taught him AI automation capability he did not have. That is the single strongest proof point available to Verity Agentic, because it says even competent AI-tool users get more capable after working with her. It also sits squarely inside the Team Enablement pillar, so the service match is direct, not adjacent.

"I built HourHound with Claude Code, so I figured I already knew what these tools could do. Annette sent back an audit where every fix came as a prompt I could paste and run myself, including a couple of things I hadn't realized I could automate. She also caught that the most important page in every city I cover was showing Google almost nothing. Then she added a whole section on how to actually make money from what I'd built, which I never asked for. Learning to find and fix this stuff myself was worth more than the fixes."

Ben [surname], founder, hourhound.app

Short version for cards and tight placements:

"I built it with Claude Code and still learned things I didn't know I could automate. Then she added a section on how to make money from it, which I never asked for."

On the went-beyond-scope clause (added at Annette's direction 2026-07-29). It is true: the audit's monetization section opens "You didn't ask me for a business plan, but you built something with real assets, so here are some ideas," and lays out five revenue paths tagged by whether each works before traffic exists. Two reasons it earns its place: exceeding scope is a strong trust signal, and monetization is the most legible part of the whole engagement to a non-technical reader, which offsets the concern that Ben's quote skews technical.

Deliberately does not say "free." A quote implying extras get thrown in reads as generous to some buyers and as loose boundaries to others, and it quietly undercuts a $3,000 to $8,000 price. "Which I never asked for" keeps the generosity, frames it as Annette's initiative rather than a client entitlement, and sets no expectation for future engagements. Do not swap this back to "free" or "at no charge."

The long version runs about 100 words, over the 40-to-90 guideline above. It earns it because it is one escalating arc rather than a list: I thought I knew this, she taught me more, she caught something real, she gave me more than I came for, and the teaching outlasted the fixes. Keep the enablement line last in any edit; it is the sentence that sells the business.

On the "too technical" question. Annette raised the concern that Ben's quote would go over a non-technical buyer's head. Partly right, but the fix is not to plain it down. Specificity is what makes a testimonial believable even when the reader cannot parse it, and a developer vouching is exactly the proof a nervous non-technical buyer needs. The draft above resolves it by leading with the enablement claim and demoting the technical catch to supporting evidence. The closing line ("learning how to find and fix that kind of thing myself was worth more") requires zero technical knowledge and states Verity's whole value proposition in Ben's own mouth. An earlier draft led with "static pages, schema, sitemap," which put three meaningless-to-the-buyer words in the second sentence, the worst possible place for a dead line. Do not restore that version.

Ben needs to confirm: his surname and how he wants to be titled; that he is fine publishing the 438KB / seventy-character / 2,962-page specifics about his own site, since they are unflattering about the pre-fix state; and whether he has actually shipped the fix.

Verified 2026-07-29: there is no result to point to yet, so do not imply one. The SEO Believer AI-citation tracker has been measuring hourhound.app weekly since 2026-07-13 (runs on 07-13, 07-20, 07-27, 80 queries per run across ChatGPT, Gemini, Perplexity, and Claude). Organic mentions for HourHound: zero in every run. That is expected three weeks in and it is exactly what the tracker exists to measure, but it means the draft above is correctly limited to the quality of the diagnosis. Do not let the quote drift toward "and traffic went up." Re-check after the 2026-07-31 week-two follow-up.

Retracted: the earlier "service-match caveat" in this plan was wrong. It claimed Ben's engagement was adjacent to what Verity Agentic sells, on the grounds that it was an SEO audit delivered under the SEO Believer brand. That mis-read the deliverable. Teaching a Claude Code developer to automate his own fixes is Team Enablement, which is a named Verity pillar, so the match is direct. The tracked-client-domain detail is a measurement artifact, not a statement of scope.

Where that leaves the pair. Both testimonials now match the service, and they cover different objections cleanly:

	Bob Manthy	Ben (HourHound)
Objection killed	AI is hype; and why not build it myself	I will end up dependent on her
Credentials	LPC, MS Computer Science, MBA	Built his own app with Claude Code
Verity pillar	Agentic Workflows / AI Operations	Team Enablement
Arm's length?	No, he is Annette's partner (see below)	Yes
Best placement	Home page hero-adjacent, next to the MIT 95% stat	Pricing page, tech-stack FAQ, build application

The gap this leaves, stated plainly. After the Bob correction, both references are technically sophisticated: a master's in computer science, and a developer who built his own app with Claude Code. That is a strong pattern in one respect (capable people who could have done it themselves chose her, which kills the do-it-myself objection twice over) and a real hole in another: neither quote gives a non-technical buyer anyone to see themselves in. The coaches and therapists page is aimed at exactly that reader, and it currently has no mirror.

So the third reference is no longer a nice-to-have. The priority for reference #3 is someone non-technical, arm's-length, and Verity-scoped, which is precisely the RISE coaches and therapists group. Until that exists, do not place either current quote on for-coaches-and-therapists.html expecting it to do the "someone like me" job, because it cannot.

Stop trying to make one quote work everywhere. Bob answers the hype question for a skeptical buyer who wants a qualified opinion. Ben answers the dependency question. A future RISE quote answers "will this work for someone like me." Steph runs different quotes on different pages, and that is why hers land.

Short pull-quote version, for cards: "I built it myself and thought search was handled. She found that the most valuable page in every market I cover was effectively blank to Google."

Note on the Bob attribution

Bob Manthy is Annette's partner and her co-author on the divorce recovery book, and that relationship is public. A prospect who searches either name will find it. An undisclosed partner testimonial presented as an arm's-length client reference costs more trust than having no testimonial at all.

This does not mean do not use him. The work was real, substantial, and he gave positive feedback on the 88-page audit on 2026-05-10. It means the relationship gets acknowledged rather than hidden. Three workable options, for Annette to pick:

Attribute plainly and let the work speak, accepting that a reader may connect the names later. Weakest option.
Disclose in the attribution line itself, something like "Bob Manthy, founder, Rebuilding Seminars (and my partner, who got the unvarnished version)". Turning it into a small piece of candor converts a liability into a voice note.
Use Bob as a named case study with the relationship disclosed in the body, and rely on Ben plus a third arm's-length reference for the testimonial slots.

Recommend option 2 for the card and option 3 for the eventual sales page. Either way, Ben carries the arm's-length credibility, which is another reason not to let his draft slip.

If either of them edits toward "highly recommend"

Both were asked to approve or edit, so they may soften the specifics into generic praise, which is the normal instinct. That would make each quote materially weaker. Worth Annette gently asking them to keep the concrete detail (the 288 reviews, the 438KB map) even if they rewrite every other word, and to keep their own reservation in.

Build: - New component .testimonial-band, styled in the existing brand (noir card, yellow or pink rule, no bracket ornaments per the global rule). - Required fields per quote: quote text, full name, business name. Never initials-only. Steph uses full names on primary pages and initials only on the repeat carousel, which is the right way round. - Placement, in priority order: - index.html: one after the "Systems I've already shipped" case-study section, one immediately before the closing CTA. - pricing.html: one after the ladder. - for-coaches-and-therapists.html: one high, since it is the coldest traffic. - The new audit sales page (Item 2): at least three, spaced between sections.

Done when: at least two real attributed quotes are live on index.html, and the component exists on the other three pages ready to populate.

Item 2. Sales page for the Revenue Leak Audit

Why: this is the single biggest gap on the site. The Revenue Leak Audit is the rung that converts a stranger into a paying client, and it currently gets one 44-word paragraph on pricing.html. Steph's equivalent $997 offer has a dedicated ~1,800-word page that repeats one CTA seven times.

New file: public/busywork-audit.html

Naming settled 2026-07-29. The paid offer is The Busywork Audit, delivering the Automation Makeover. The free download stays the Revenue Leak Report, unchanged. The old collision is gone, because the two offers no longer share a word.

The timeframe was dropped from the deliverable name, and that was Annette's call. An earlier draft called it the 90-Day Automation Makeover, borrowed from Steph's "Custom 90-Day Action Plan." Annette pointed out she can complete and implement an automation in well under 90 days, and she is right that the number was working against her:

She has now committed to findings in 3 to 5 business days and a build in 1 to 2 weeks. A deliverable labeled "90-Day" contradicts the speed that is her actual differentiator.
Steph needs 90 days because her engagements are quarterly retainers. Verity's positioning is the opposite of quarterly.
30-Day is not available as a substitute. pricing.html already promises "your business running without you in 30 days" for the $15,000 AI Ops Blueprint. Reusing 30 days on a $500 deliverable recreates exactly the naming collision this decision just fixed.

So speed lives in the promises, not the label. The turnaround (3 to 5 business days) and the build window (1 to 2 weeks) both appear as stated commitments on the sales page, where they function as promises. A number inside an artifact's name is decoration and boxes her in for no gain.

Stated turnaround: findings delivered in 3 to 5 business days. Goes in the deliverables list below as a promise, not a hope. Ben's came back in 24 hours, so this is comfortably beatable, which is the right side of the line to be on.

One copy risk, and the safety net has been removed on purpose. "Makeover" can read as the work being done rather than planned, so a buyer paying $500 could reasonably expect built automations. Adding "Plan" to the name would have handled that quietly. Annette chose the shorter name (2026-07-29), which means the surrounding copy now has to carry it explicitly. This is a requirement, not a nice-to-have:

The deliverables list must say in plain words that the Makeover is the roadmap: what to fix, in what order, and why.
The page must name the next rung, so the reader sees where building actually happens and can choose it.
Do not let a headline imply the automations arrive with the audit. That mismatch is what produces refund conversations, and without "Plan" in the name there is nothing else stopping it.

Worth noting the upside of her call: "the Automation Makeover" is a better name, more confident and less bureaucratic than "Makeover Plan." It just shifts the work onto the copy.

Section sequence (adapted from her eleven-part structure, in Annette's voice, not hers):

Hero. Eyebrow, headline naming the outcome, one-line subhead, primary CTA.
The problem, in the reader's own words. Not categories. Lived moments: the intake form that still needs you to read it, the follow-up you meant to send Tuesday, the client who slipped because the handoff lived in your head. See Item 8 for the rewrite rule.
The three fixes they have already tried and why each failed. Steph's version: the funnels that were supposed to bring people in, the automations that were supposed to run without them, the team that was supposed to handle things. Verity's version should name the no-code tool that still needed a human, the freelancer who built one thing and left, and the ChatGPT subscription nobody operationalized.
The after-state. What the week looks like when the leak is closed.
Testimonial.
Introducing the offer. What it is, plainly.
What is included: six named deliverables. This is the highest-value part of the whole page and it is mostly a matter of writing down what already happens. Draft list to confirm with Annette: - Pre-session intake questionnaire and backend review - 60-minute working session - the Automation Makeover, the written findings document. Named, because a named artifact is worth materially more than "written findings." Settled 2026-07-29: no timeframe in the name, and not "Plan" either. - Recorded walkthrough of the findings - Full session recording - The fee credits toward a build if they continue
Price, stated plainly. From $500, credited toward a build. The credit is a strong risk-reversal and currently gets one clause on the pricing page.
CTA repeat.
How it works, four numbered steps. With a concrete turnaround number, not "within weeks."
Who Annette is, short, and pointed at this offer specifically rather than a general bio.
Objection FAQ, eight to ten questions. See Item 6 for the list.
Closing CTA.

CTA discipline: one button, identical text every time, repeated at least five times down the page. Per Item 3, that text is the free 15-minute audit everywhere else on the site, but on this page the CTA is the paid booking. This is the one page where a different primary CTA is correct, because the reader arrived already qualified.

Also required: - Link it from the pricing.html ladder (rung 01 becomes a link, not a paragraph). - Add to sitemap.xml. - Add Service or Offer schema with the price. - Add to llms.txt.

Done when: the page is live, linked from pricing and nav, validator green, and /no-ai-slop has been run on the full draft.

Phase 2: Structural discipline

Item 3. Collapse to one call to action

The problem: index.html currently asks for four different things. Nav carries "FREE REPORT" and "BUILD YOUR SYSTEM". Body carries "GET YOUR FREE 15-MIN AI AUDIT" three times plus "TAKE THE QUIZ" plus two "Read the case study" links. Each is defensible alone. Together they split the decision. Every button on Steph's site says the same five words.

Decided 2026-07-29, and it is better than what this item originally proposed. Annette's answer: the primary CTA is the free 15-minute audit, and that audit should consist of a detailed questionnaire completed first, then a 15-minute Zoom call.

That second half changes the whole item. The original plan was to demote the quiz into a quiet secondary band, which treated it as clutter to be tidied away. But the quiz already IS the detailed questionnaire. quiz.html asks ten questions scoring the reader across Operations, Data, People, Strategy and Execution. That is exactly the intake instrument the free audit now requires.

So the quiz stops being a competing call to action and becomes the on-ramp to the only one:

Take the quiz and get your score → book the 15-minute call to go through it with Annette

Why this is a genuine upgrade, not just tidier: - It removes a competing CTA by promoting it rather than burying it, so nothing of value gets thrown away. - Annette arrives at every call already knowing the person's operation, which makes a 15-minute call actually sufficient. - It qualifies the caller before the call, the same mechanism Steph gets from her paid intake, except Verity gets it for free. - The quiz gains a reason to exist. Right now it produces a score and then asks nothing of the reader. - It mirrors the paid ladder, so the free tier previews how the paid work feels.

Changes: - Primary CTA everywhere: the free 15-minute audit. One wording, used verbatim in every instance. - Rewire the quiz result screen so completing it leads directly to booking the call, carrying the score through. This is the highest-value single edit in the item. - Nav: drop to one primary button. BUILD YOUR SYSTEM and FREE REPORT currently compete; keep one button and move the other into the plain link list. Apply to both .nav-links and .mobile-menu in every page, since the nav is duplicated inline per file. - Free report: demote to a single quiet band low on the home page, below the FAQ. It is a lead magnet for people not ready to talk, which is a real but secondary job. - Booking flow: Calendly (confirmed by Annette 2026-07-29). Mechanism verified against Calendly's current docs, not assumed: - Calendly accepts name and email (or firstName / lastName when the event splits them), plus custom invitee answers as a1 through a10. Both work either as URL query-string parameters or via the prefill object on a JS embed. - Use the inline JS embed with the prefill object, not a query string. Same result, but the reader's answers never end up in a shareable, loggable, referrer-leaking URL. Their business's weak points are their information, and there is no reason to publish it in an address bar. - Do not map ten quiz questions to the ten a slots. It burns the entire budget and produces an unreadable notification email. Pack it instead: a1 = overall score plus the five section scores as one compact line, a2 = the weakest section named. That is what Annette actually needs to see before a 15-minute call, and it leaves eight slots spare. - Example shape (confirm the event's own question order at build time, since a1 maps to whatever question sits first): js Calendly.initInlineWidget({ url: 'https://calendly.com/<annette>/15min', parentElement: document.getElementById('calendly-embed'), prefill: { name: quizName, email: quizEmail, customAnswers: { a1: scoreSummary, a2: weakestSection } } }) - Verify by booking a real test slot and confirming the score line arrives in the Calendly notification. Per the standing rule, no half-working form ships. - Calendly also has Routing Forms, which can ask questions before routing to a booking. Not the right fit here: the quiz already exists, it lives on Annette's domain where it earns search traffic, and moving it into Calendly would give that away. - Sweep all eight pages for stray CTA variants. The wording must be identical, not merely similar.

Done when: a full-text grep of public/*.html shows exactly one primary CTA string, finishing the quiz leads straight into booking, and a test booking arrives with the quiz answers attached.

Item 4. Define and repeat the framework

The problem: "The Verity Agentic System" resolves to five bare chips (Agentic Workflows, AI Operations, Custom Automations, Team Enablement, Governance and Guardrails) with no definitions, on one page. Steph's six-pillar "The Foundation I Build" block carries one sentence each and appears identically on three pages. The repetition is what converts a list into a methodology.

Changes: - Write one sentence per pillar, describing what it covers in the client's terms, not the tool's. Roughly 20 to 30 words each, matching the density of the existing "What We Build" bullets. - Consider whether five is the right number or whether Governance and Guardrails plus Team Enablement should split into six to match the natural coverage. Do not pad to six for its own sake. - Extract into a single block and repeat verbatim on: index.html, about.html, pricing.html, for-coaches-and-therapists.html, and the new audit page. - Because CSS is inline per page, the block's styles must be copied into each page's <style>. Keep the class names identical across pages so a future extraction to a shared stylesheet is trivial.

Done when: the block appears on five pages with identical copy and identical rendering.

Item 5. Application gate for the high-ticket tiers

The problem: the $3,000 to $8,000 build and the $1,500 to $3,000/month retainer both route to the same generic contact form as everything else. Steph gates her project tier behind a 10-minute application.

Why it works: it filters tire-kickers, it makes the offer feel selective rather than available, and it means the first call starts with the diagnostic half-finished.

New file: public/build-application.html

Questions (adapted; hers are a good model and the last one is the sharpest): - Name, email, business name, website - What the business does and who it serves - Annual revenue range (select) - Which tier they are applying for - The full current tech stack, free text. This is the single most useful field. - What breaks today, in their words - What they will be able to stop doing when this is built - Timeline, and any launch or deadline - Have they hired someone for this before, and what happened? A screening instrument disguised as curiosity. Keep it.

Implementation: Web3Forms, same key as the existing contact form (3c8b2baa-2dec-4e05-88c2-1a9156b1221b). Test end-to-end before the page ships. Per the standing rule, no broken form is ever published.

Routing: pricing.html rungs 02 and 03 point here. contact.html stays as-is for everything else.

Done when: the form submits successfully end-to-end and a test submission has actually arrived at [email protected].

Phase 3: Copy

Item 6. Objection FAQ

The problem: the home FAQ answers What does Verity Agentic do? Who is Annette Thompson? How much does it cost? Those are written for a search engine, which is a legitimate purpose. They are not written for a hesitating buyer. Steph's ten FAQs are almost entirely objections, and each concedes something real before turning it. Her answer to "couldn't I figure this out myself" opens with the word "Maybe."

Note: pricing.html already does this well in four questions. That is proof the voice exists. It just needs to appear where the buyer actually hesitates.

Changes: - Keep the five existing informational Q&As and their FAQPage schema. They earn AI-search citations and should not be removed. - Add a second set beneath, five questions minimum: - What if AI is not actually the answer for my business? - My process is a mess. Is it too early to automate it? - What happens when it breaks, or when a model changes underneath it? - What if my data is not clean or not in one place? - Am I too small for this? - Rule for every answer: concede the real thing first, then turn it. An answer that does not concede reads as sales copy and gets skipped. - Extend the FAQPage schema to cover the new questions too. - Put the audit-specific version of this set on the new sales page (Item 2, section 12), not a copy of the home set.

Item 7. Rewrite the about page around the method

The problem: the about page is a stronger set of facts than hers, told as a résumé. 1994 online, 1995 adoption.com, Capgemini, Ethiopia, Kenya, Haiti, E-Teach Abroad, Bone Voyage. The reader admires it and moves on, because a timeline does not explain a method.

The line that actually sells is already on the page, buried mid-paragraph: building an adoption program in Ethiopia with no budget, no team, and no playbook is exactly the condition under which most AI implementations fail. Steph has a section headed What changed how I think about this work that does this job explicitly.

Changes: - Promote that insight into its own headed section, positioned above the timeline. - The section answers one question: why does thirty years of building in unstructured conditions make her the right person to build your system? Connect the history to the method, not to her impressiveness. - Keep the timeline. It is good supporting evidence. It just should not be carrying the argument. - Do not add trauma-narrative framing. Steph's version works for her buyer and her positioning. Annette's equivalent is competence under conditions with real stakes, which is a different and stronger claim for this audience.

Item 8. Symptoms in the buyer's own words

The problem: the before/after table on index.html reads "Manual data entry / 5 follow-up emails / Missed handoffs / You doing the work." Those are categories. Steph writes lived moments: "The funnels that were supposed to bring people in consistently… don't."

Changes: - Rewrite the BEFORE column as specific recognizable moments. The test: could the reader picture the exact Tuesday it happened? - Do not copy her trailing-ellipsis rhythm. It is on the banned list. - Same rewrite applies to section 2 of the new sales page.

Item 9. Concrete containers

The problem: "Most first automations ship in weeks" is vague where Steph is specific: three weeks, ninety days, weeks 1-2 audit, weeks 3-10 build, weeks 11-12 test. A named duration reduces perceived risk more than any adjective.

Numbers decided 2026-07-29:

Offer	Published commitment
The Busywork Audit, written findings	3 to 5 business days
Done-For-You Agentic Workflow	1 to 2 weeks from kick-off to running

Changes: - Replace every vague duration on the site with the numbers above. "Most first automations ship in weeks" becomes the specific figure. - Put the 3-to-5-day findings turnaround in the Item 2 deliverables list. - Consider a week-by-week breakdown for the retainer's first 90 days on pricing.html.

One flag on the 1-to-2-week build, recorded rather than argued: it is aggressive for a solo operator taking limited clients, and once published it is the number every prospect anchors on. The competitive claim that matters ("weeks, not the six-month timelines big vendors quote") holds just as well at three weeks. If a build ever looks like slipping past two weeks, the right move is to revise the published number, not to miss it once. Annette made this call with the under-promise argument in front of her.

Why turnaround lives here and not in Ben's testimonial (decided 2026-07-29): a promise the business makes is stronger than a compliment a client pays, because it is a commitment rather than a data point. Two further reasons to keep it out of the quote: Ben's quote is already at 100 words doing three jobs, and a fourth element turns a coherent arc into an inventory. And advertising that a research-heavy deliverable took a day invites the buyer to price it as a day of work, which is the same trap as saying "free." If Annette still wants it in the quote, the cheapest form is a five-word insertion reading as speech, not a service level: "Annette sent it back the next day, and every fix came as a prompt I could paste and run myself." Do not use "in 24 hours" inside a testimonial.

Item 10. Move the tech stack down the ladder

Minor but real: pricing.html names Claude Code, Codex, and LangGraph in the FAQ. Steph does not name a single tool until deep inside her project tier, because tool names invite tool comparisons at the exact moment the reader should be evaluating the outcome.

The answer itself is good and worth keeping, since "not a black box only I can touch" is a genuine objection-killer. Consider moving it below the ladder rather than into it, or onto the application page where the reader is already committed.

Explicitly not doing

Thing	Why not
Repainting to cream, sage, and Cormorant Garamond	Her palette sits almost exactly on Annette's documented visual taste, which makes it tempting. But noir black with yellow and pink is genuinely differentiated in a category drowning in soft beige consultant sites. Trading it away would make the site look like everyone else's. All the value in this teardown is structural and palette-independent.
"This isn't X, it's Y" constructions	Banned in the voice rules and all over her pages.
Coaching vocabulary: "containers", "a little client love", "a space where you get to be your full self"	Fits her buyer. Reads as soft in front of an operator deciding whether to spend $8,000.
Trailing ellipses as a rhythm device	Reads as AI-adjacent. Not Annette's voice.
Two service pages with identical content	Her `/services2` and `/work-with-me` are the same page at two URLs, a live duplicate-content problem. Do not replicate.

Suggested order

Item 1 (testimonials) starts immediately, because it is blocked on Annette collecting real quotes and that has the longest lead time.
Item 2 (audit sales page) is the highest-value build and is not blocked on anything.
Item 3 (one CTA) is quick and makes items 1 and 2 land harder.
Items 4, 5 structural.
Items 6 through 10 copy passes.

Items 3, 4, 8, 9, and 10 are all editing text that already exists and could reasonably be done in a single sweep.

Decisions made (answered by Annette 2026-07-29)

All six are settled. Two of the answers changed the plan rather than just filling a blank, and those are flagged.

1. Naming the two offers

The free one stays the Revenue Leak Report. Annette's note: it is the shorter document. Once the paid offer is renamed, the collision disappears on its own and the free download needs no change at all.

The paid offer becomes The Busywork Audit. Recommended over the alternatives because "busywork" is already Annette's own word on the live homepage ("the busywork that eats your week"), it is instantly legible to a non-technical buyer, and no competitor in AI consulting is using it. "Automation Audit" and "Revenue Leak Audit" both test as generic. It also pairs cleanly with the deliverable name in #5: the audit names the problem, the makeover names the fix.

Resulting file names: paid offer at public/busywork-audit.html, free download unchanged at public/revenue-leak-report.html. Nav labels: "THE BUSYWORK AUDIT" and "FREE REPORT" respectively, and they can no longer be mistaken for each other.

2. Findings turnaround: 3 to 5 business days

Goes in the Item 2 deliverables list as a stated promise. Comfortably beatable, since Ben's came back in 24 hours.

3. Done-For-You build: 1 to 2 weeks

Recorded as decided. Worth one note for whoever ships the sales page: this is aggressive for a solo operator, and it becomes the number every prospect anchors on. If a build ever looks like slipping past two weeks, the honest move is to change the published number rather than to miss it once. The competitive claim ("weeks, not the six-month timelines big vendors quote") survives intact at three weeks too, if she ever wants the headroom.

4. Primary CTA, and the free audit gets a real structure

Primary CTA everywhere: the free 15-minute audit. Settled.

But the free audit is now a two-part offer: a detailed questionnaire completed first, then a 15-minute Zoom call. That is a genuine upgrade rather than a naming decision, and it changes Item 3. See the rewritten Item 3 for what it unlocks.

5. The findings document is the Automation Makeover

Annette first chose "90-Day Automation Makeover", then corrected it the same day: she can complete and implement an automation in well under 90 days, so the number undersold her. The timeframe is dropped from the name. Full reasoning is in Item 2, but in short: 90 days contradicts her own 1-to-2-week build commitment, and 30 days is already claimed by the $15,000 AI Ops Blueprint on the pricing page.

Speed now lives where it belongs, as a stated promise (3 to 5 business days for findings, 1 to 2 weeks for a build) rather than as decoration inside an artifact name.

Settled 2026-07-29: "the Automation Makeover", not "Makeover Plan". Annette's call. So the job of preventing a $500 buyer from expecting built automations moves entirely into the surrounding copy, which now has to do it explicitly rather than leaning on the word "Plan" to do it quietly. See the copy-risk note in Item 2, which is no longer optional.

6. Yes, spec the enablement document as a new rung

See the new Item 11.

The ladder as it now stands (shipped 2026-07-29)

The $15,000 AI Ops Blueprint has been removed from the live site. Annette's call: focus on small projects first. Removed from the pricing ladder, the Offer schema, all three pricing meta descriptions, the homepage FAQ answer and its FAQPage schema, and llms.txt. The retainer renumbered from 04 to 03 and the two-column grid holds fine with three cards. Deployed and validated GREEN over two runs.

Industry-rate mentions of $15,000 inside the articles were left alone. Those are editorial commentary about what the market charges, not Verity offers.

The settled naming also went live in the same pass: the paid audit is now The Busywork Audit on the pricing page, the coaches page, the homepage FAQ and llms.txt. The free download is untouched and stays the Revenue Leak Report.

#	Offer	Price	Status
0	Free 15-minute audit	Free	Live. Becomes questionnaire plus Zoom per Item 3.
1	The Busywork Audit	From $500	Live, name updated. Needs the sales page (Item 2).
2	The Automation Playbook	$1,200 to $1,500 proposed	Not built. See Item 11.
3	Done-For-You Agentic Workflow	$3,000 to $8,000	Live. Needs the 1-to-2-week commitment stated.
4	AI Ops Retainer	$1,500 to $3,000/mo	Live, renumbered.

Why removing the top rung helps rather than hurts. The ladder now runs from free to $8,000 with no jump larger than roughly 4x, and every rung is something a 1-to-10-person business can actually say yes to. The $15,000 tier was aimed at a buyer Verity was not yet set up to serve, and its presence invited the "is this really for businesses my size" doubt on the page whose whole job is removing friction. If a Blueprint-sized engagement ever walks in, it can be quoted without being advertised.

One consequence for the naming reasoning. Item 2 originally ruled out "30-Day Automation Makeover" partly because the Blueprint already claimed "running without you in 30 days." That collision is now gone. The decision still stands unchanged, because the stronger reason was always the first one: a timeframe in the artifact name fights a 1-to-2-week build commitment. Do not reopen it.

Item 11. The enablement rung (new, approved 2026-07-29)

The gap it fills. The ladder currently jumps from a $500 audit straight to a $3,000 to $8,000 build. There is nothing for the buyer who wants to be taught to do it rather than have it done for them, which is a real and growing segment now that capable people are already using these tools.

The template already exists. The HourHound week-one document is the prototype: six ready-to-run Claude Code prompts each tied to a specific fix, a primer on the concepts that actually applied to him, a sequencing judgment call, a Search Console walkthrough, analytics options with a recommendation, and a monetization map. Ben was already fluent with Claude Code and still gained capability he did not have. That is the proof the offer works.

Proposed shape, for Annette to confirm:

Name: working title The Automation Playbook. Wants to sound like a thing you receive and keep.
Price: $1,200 to $1,500. Sits deliberately between the $500 audit and the $3,000 build, and is defensible because the buyer leaves able to do the work repeatedly rather than owning one finished workflow.
Format: the document plus one 60-minute working session where Annette runs the first prompt alongside them. The session is what makes it stick and what justifies the price over a PDF. Ben effectively got the document alone; adding the session is the productized upgrade.
Deliverables: five to eight prompts tailored to their actual stack, the concept primer scoped to what applies to them, the shipping order, and the working session recorded.
Positioning line to test: the audit tells you what is broken, the playbook teaches you to fix it yourself, the build has it fixed for you.

Open sub-questions: final name, final price, and whether the working session is included or a paid add-on.

Why this rung is strategically interesting. It is the only offer on the ladder that scales without Annette's delivery time growing linearly, because the document is reusable across clients in the same stack. It is also the natural upsell path from the free audit for anyone who says "I would rather learn this."

Still open

Only three things, all small:

Confirm the Item 11 name, price, and whether the working session is bundled.
Bob and Ben need to approve their quotes. Emails are ready at the testimonial kit.

Naming is now fully settled: The Busywork Audit (the $500 offer), delivering the Automation Makeover (the findings document), with the Revenue Leak Report staying as the free download.

Rendered from PLAN.md in the VerityAgentic repo, which stays the source of truth. Regenerate with scripts/build-verity-plan-page.py after editing it.