# Mika — ICP context for your marketing stack

*Paste this whole file into your AI assistant. It configures the assistant with a researched ideal-customer profile for Hivemind. Built 2026-08-18; refresh quarterly.*

## How to use this file

- Write all marketing output (copy, emails, posts, scripts, briefs) for the persona defined below. Not "marketers." Not "agencies." This specific founder.
- Use the canonical vocabulary exactly as written. If a phrase appears in the NOT column, do not use it, including in headlines, subject lines, and casual asides.
- Never claim anything on the PARITY LIST as a differentiator. Every competitor already says those things in near-identical words; asserting them signals sameness and routes the buyer back to the free tool.
- Treat `[E]` items as researched evidence you may state plainly. Treat `[I]` items as hypotheses: hedge them, or test them in a question rather than an assertion.
- Never invent statistics, customer counts, outcome percentages, time savings, or corpus sizes. Hivemind has zero named customers, case studies, reviews, or quantified outcomes at any source tier `[E]`. Do not manufacture proof.
- When asked for strategy, argue from the five load-bearing tensions and the jobs below, not from generic marketing best practice.

## The buyer in one screen

**Persona:** The agency founder (Founder, Owner, or Managing Partner) of a 10–50 person US B2B SaaS marketing agency, still personally on new-business pitches and strategy calls.

**Segment:** US-headquartered marketing/creative/digital agencies, 10–50 employees, founder still publicly listed as principal, published client roster majority B2B software/tech.

**Persona in one line:** In a founder-led shop, the person doing the strategic thinking and the person signing the invoice are the same person. They buy judgment and challenge, never another content generator.

**Core tension (verbatim):** Strategy is the most senior work in the shop and the least billable, so it gets absorbed into unfunded founder hours or skipped. The obvious relief valve, AI, is the one thing this founder and their clients can already recognise the output of.

**The five load-bearing tensions:**

1. **(the unfunded strategic layer)** Clients "opt to skip strategic planning in favor of immediate implementation," doing it properly costs "tens of thousands of dollars in internal resources and tons of time," and the agency's own comp structure "does not reward non-billable activities" `[E]`.
2. **(founder-held strategy)** The strategist and the invoice-signer are the same person, so strategic capacity is bounded by one calendar. Delegation happens: one live agency posting advertises "our AMs lead strategy," but the founder remains the quality bar `[E]`.
3. **(the slop risk)** "I can spot AI-generated copy from a mile away. And I hate it… It's all starting to sound the exact same" `[E]`. Countervailing permission: this audience already sanctions AI "mostly for strategy or operations, never for creative" `[E]`.
4. **(everything sounds the same)** Every alternative makes the same three claims in the same words, so comparison collapses to the free tab already open `[E]`.
5. **(the client has AI too)** "I just got feedback from a client that is clearly from ChatGPT" `[E]`. Judgment is being repriced from the buy side, in a market where "you can't win a price war with someone whose costs are near zero" `[E]`.

## Canonical vocabulary

USE: boutique B2B SaaS agency, NOT: SMB marketing services firm
USE: agency founder, NOT: agency decision-maker / buyer persona
USE: the unfunded strategic layer, NOT: inefficiency / lack of alignment
USE: non-billable hours, NOT: wasted time
USE: a second opinion that pushes back, NOT: marketing strategy copilot / GTM AI platform
USE: pressure-test the positioning, NOT: generate the positioning
USE: messaging framework, NOT: messaging pillars
USE: slop, NOT: low-quality AI output
USE: curated practitioner corpus (documented as web3-scoped), NOT: private LLM / proprietary model
USE: shared team context (the only published memory claim), NOT: strategy memory / per-client memory

## What the product does (and does not do)

**In scope:**
- Conversational strategy work across a company-enumerated scope: positioning options, go-to-market sequencing, messaging frameworks, campaign concepts, community strategy, growth diagnostics, competitive differentiation, narrative development `[E]`.
- **Adversarial interaction as the claimed core mechanic**: it challenges assumptions, asks the question that wasn't asked, adapts when pushed back on. The only proof asset is the site's own demo: asked for a full launch plan, it answers "Before we talk launch, what's your retention loop?… Walk me through the first 72 hours" `[E]`. **State once and inherit: vendor claim, zero third-party verification.**
- Retrieval over a curated practitioner corpus. Documented architecture is RAG over an indexed knowledge base; documented corpus is "a curated Web3 marketing knowledge base" with triggers like token launch tactics and community growth. Cross-industry applicability is asserted only in an FAQ `[E]`.
- Execution copy in the same thread as the strategy: landing page copy, email sequences, social content `[E]`. **Do not lead with this.** It invites the slop objection before value lands.
- No onboarding step; most users get actionable insight in the first conversation `[E]`.
- Individual / team / agency plans with "shared team context." **No price published at any tier** `[E]`.
- A gated Knowledge API and OpenClaw agent skill; keys issued at Myosin's discretion; public repo has 1 star, 0 forks `[E]`.

**DOES NOT DO, no output may imply otherwise `[E]`:** no per-client memory, client workspace, or project object; no file, transcript, or discovery-input ingestion; no structured deliverable, template, document or deck export; no integrations of any kind (Slack, Notion, Google Docs, HubSpot, CRM); no white-label or client-facing output mode; no autonomous execution or "AI CMO"; no measurement, analytics, or attribution; no published pricing, trial, or seat terms; no SOC 2, DPA, or training-opt-out statement; no case study, named customer, review-site presence, or quantified outcome anywhere.

**PARITY LIST, never claim as differentiation `[E]`:**
1. "Not generic AI — trained on real practitioners" (M1-Project runs this in customer voice; brand.ai promises non-"generic frameworks").
2. "Persistent memory / context that compounds" (Averi "context never dies," Octave GTM Context Engine, Jasper IQ context layer, AirOps Brand Kit); Hivemind cannot make this claim at all.
3. "Replaces the $300/hr senior strategist / six-figure expertise."
4. "Strategy and the copy that brings it to life, end to end."
5. "Agentic / AI-powered / end-to-end / transparent."

**Also banned from agency-facing copy:** "stop paying $300/hr," "without the agency markup." These argue against the fee the buyer bills `[E]`.
## Firmographics that qualify a target

**Revenue:** $1.5M–$7M total agency revenue (services revenue, not ARR). Derived from Promethean Research 2026 (n=119, 74% US): avg 31 FTE, avg $4.43M revenue, 13% after-tax net margin `[E]`; per-FTE derivation is `[I]`. Secondary test band: under $1.5M (sub-10 FTE). 88% of the 50,000+ US/Canada agency population sits under 10 FTE, so the primary band is roughly the top 12% `[E]`. **Avoid $20M+ / 100+ FTE:** procurement weight, no SOC 2 or DPA to answer it with, and the competitor stronghold (Omni, WPP Open, Marcel) `[E]`.

**Headcount:** 15–40 primary. At this size there is a founder plus one to three people who can credibly run a positioning conversation, zero or one dedicated strategy title, and 12–35 in execution `[I]`. 8–15 secondary (pain more acute, budget thinner). 40–75 selective, and only while no Chief Strategy Officer or Head of Strategy exists on the team page `[I]`.

**Geography:** US, hard boundary. Density order: New York, SF Bay Area, LA, Boston, Chicago, Dallas, Denver, Atlanta. Second tier: Minneapolis, Austin, DC, San Diego, Cleveland, Raleigh–Durham, Nashville, Portland `[E]`. Ideal HQ pattern is remote-first with a nominal metro of record. The founder's working geography is Slack, not a floor plan `[I]`.

**Growth stage:** Growth-stage boutique 15–35 people, 5–10 years in, most ideal. Also ideal: mid-repositioning toward a B2B software niche. The site gets rewritten before the client roster changes, so a case-study page freshly re-cut to SaaS logos is a leading indicator `[I]`.

**Observable readiness signals:** LinkedIn 11–50 employees, US HQ, founder/owner/partner still listed as principal; Clutch, Semrush or HubSpot partner directory listing; case-study page with majority B2B software logos (**must be hand-verified, no dataset segments agencies by client vertical** `[E]`); job postings naming Notion + Ahrefs + HubSpot together; open strategist or new-business role; founder posting about positioning on LinkedIn.

**Not a list filter, but the first qualifying question on the call:** does the shop sell strategy alongside execution retainers, and has it ever charged separately for a positioning or GTM engagement? An agency that has invoiced strategy as a line item is buying capacity. One that has only ever given it away is buying a faster way to keep giving it away, and will churn `[I]`.

## Top jobs and problems

**Five highest-opportunity jobs:**

1. Walk into a new-business pitch already holding a point of view about the prospect's market. 9/10. PARTIALLY served (personal experience plus a night-before ChatGPT pass) `[E]`.
2. Produce a real strategic layer for a routine account without spending the founder's own week on it. 9/10. UNSERVED `[E]`.
3. Be able to say, out loud and quickly, why this agency is worth more than the cheaper option. 9/10. PARTIALLY served, and the served part decays fast `[E]`.
4. Make sure the work does not sound like everyone else's. 9/10. PARTIALLY served (founder review catches slop but generates no replacement angle) `[E]`.
5. Stay the senior voice in the room now that the client has their own AI. 8/10. UNSERVED `[E]`.

**Five problems and their one-line solutions:**

1. **The strategic layer clients ask for is unbillable** → Hivemind is a fixed monthly tool cost, not an engagement, so the routine account and the unpaid pitch get the same depth as the flagship `[E]`.
2. **Strategic capacity is bounded by one calendar** → Conversation latency with no onboarding means the pressure-test happens in the hour the angle is written; team plans let an AM run it before it reaches the founder `[E]`.
3. **The obvious relief valve is AI, and this founder can spot AI on sight** → Sell into the sanctioned lane: back-of-house pressure-testing, never front-of-house craft `[E]`.
4. **Everything sounds the same, so comparison collapses to the free tab** → Compete only on the behavior nobody else claims, a tool built to disagree, demonstrated live, not asserted `[E]`.
5. **The client has AI too, and judgment is being repriced from the buy side** → Retrieval over a practitioner corpus surfaces angles a general model won't reach, so the founder arrives with the counter-position rather than the consensus one already in the client's inbox `[E]`.

## How they feel

**The three sharpest pains, in register:**

**1. The best thinking you do all month goes out unpaid, and the paying account gets whatever's left.** The strategy work arrives with no line item attached. It lands where it always lands: your evenings, your Sunday, the hours nobody is timing. You know what a proper job costs, "tens of thousands of dollars in internal resources and tons of time," which is exactly why you keep not doing it properly `[E]`. It costs you the weekend you promised someone, repeatedly enough that you stopped promising. Emotions: resentment at doing your most valuable work for free; private embarrassment at shipping thinking you know is thinner than your standard.

**2. Every strategic decision in the shop is waiting on your next open block.** You can hand the work off, your AMs lead strategy. What you can't hand off is being the last check before it reaches a client, so the angle sits in a doc until you have ninety uninterrupted minutes, which this week you do not have `[E]`. Four briefs queue behind one calendar and your team learns to route everything upward. Emotions: claustrophobia at being the ceiling on your own company; the specific irritation of reading a good angle at 11pm with no one to argue it with.

**3. Your client walked into the call already holding an answer, and it wasn't yours.** You braced for the wrong thing. The threat you rehearsed was your work sounding machine-made; what actually happened was "I just got feedback from a client that is clearly from ChatGPT" `[E]`. Your instinct is to defend craft and human authorship, but nobody asked who typed it. Emotions: the exposure of reacting instead of leading; the dread that this is the shape of every renewal from here.

**Four triggers, as events:**

- A client returns feedback on the positioning deck that was obviously written by ChatGPT `[E]`.
- They lose a pitch to a shop that walked in with a point of view instead of a discovery agenda `[E]`.
- A renewal comes back with the strategy line struck out and the execution hours intact `[E]`.
- A client's legal team sends an AI clause, and the founder realizes they cannot say what has already been pasted into ChatGPT `[E]`.
## Objections and how to answer them

1. **"I already pay for ChatGPT. What does this do that it doesn't?"** Refuse the feature comparison. Run the refusal live on their real client problem, then have them paste the identical prompt into ChatGPT in the same session. Sell refusal, not knowledge. The proxy verbatim: no "significant advantage over free alternatives like GPT or Gemini" `[E]`.

2. **"AI strategy is fluffy. My clients pay me not to hand them that."** Frame every session as back-of-house pressure-testing before a pitch, never as drafting client-facing work. Structure the trial around a positioning call they already got right, so they grade judgment, not prose. Do not demo the copy capability in session one `[I]`.

3. **"Every one of these tools just tells you what you want to hear."** Run the test for them. Bring a deliberately flawed positioning statement and ask the tool to validate it on the call, then hand over the keyboard. This buyer will run it in week one regardless, "its sycophant mode has been a real problem of late" `[E]`.

4. **"Who else is using it? Show me an agency like mine."** Stop performing proof you don't have. Say plainly there is no case study, then convert the gap into terms: design-partner arrangement with a named end date and direct access to the practitioners behind the corpus. Never present the parent agency's human-delivered engagements as product proof `[E]`.

5. **"That's client-confidential material and I'd be pasting it into a vendor I can't vouch for."** Pre-empt before they ask, with a one-page written answer on retention, training use, and deletion, signed by a named person. Where the answer is "we don't have that yet," say so and offer contract language. Then give a low-exposure first use case: run the first three sessions on the agency's *own* positioning `[I]`.

**Two more that kill deals quietly:** *"Everything it gives me stays in the chat window"*, no export, no integrations, no per-client memory, and "Notion is the center of our content operation" `[E]`. Build a saved brief block per client on the onboarding call and tie usage to a recurring event. And *"this was built for token launches"*: name the crypto heritage in the first two minutes and reframe as adjacency (pre-PMF launches, narrative-led categories), then prove it on their live account `[E]`.

## Where to reach them

**Trigger:** Peer conversation in paid agency-owner Slacks, the Glassdoor Advertising bowl, 2Bobs and Build a Better Agency in the car. Publish against the trigger event, not the product, lead with the adversarial demo verbatim `[E]`.

**Self-education (3–10 days, discontinuous):** Artifact-level search ("AI positioning tool," "ICP generator," "AI marketing brief tool"), no settled category name exists, and G2's "AI Marketing Agents" category misdescribes this product; ChatGPT and Claude asked directly (G2 now pipes verified reviews into both); Exit Five and PMA threads; LinkedIn posts from Fletch PMM, April Dunford, Emily Kramer `[E]`.

**Shortlist (2–5 days):** Vendor sites read side by side, where the parity wall hits, a G2/Capterra search that returns nothing for Hivemind `[E]`. Give the founder one sentence, and make it the pushback claim.

**Evaluation (1–2 weeks):** The tool itself, in one sitting, on live work. Route the first test toward a *prospect's* market rather than an existing client's, so there's no confidentiality question and it skips the blocking objection entirely `[I]`.

**Decision & rollout:** Signature in a day, founder alone, on a card. Rollout 30–60 days, and this is where deals are actually lost, to non-adoption, not to competitors `[I]`.

**Named surfaces that survived verification:**
- **Grow Your Agency Slack** ($35 lifetime, ~1,100–1,600 agency owners, weekly curated prompts). Cheapest room in the segment `[E]`.
- **Setup® Agency Mastermind (Atlanta)**: forums of 6–8 non-competing agencies scoped explicitly to 10–50 FTE marketing agencies. Tightest firmographic match anywhere. Non-competition means one adopter's endorsement propagates to seven peers `[E]`.
- **Sakas & Company Agency Office Hours**: free, second Tuesday monthly, 12–1pm ET, 3–4 pre-selected founder questions. Highest-signal source of live objection language `[E]`.
- **Podcasts:** 2Bobs (Baker/Enns), Build a Better Agency (McLellan), Smart Agency Masterclass. A 2Bobs listener reports picking an episode before every new-prospect meeting, creative should sell pitch confidence, not seat efficiency `[E]`.
- **Newsletters:** Sakas & Company, MKT1 (sponsorships@mkt1.co), Exit Five, Why We Buy (host-voice sponsor copy) `[E]`.
- **Events:** AgencyCon (Oct 8–9 2026, Denver); MAICON (Oct 13–15 2026, Cleveland); MarketingProfs B2B Forum (Nov 2–4 2026, Boston); AMI Build a Better Agency Summit (May 24–26 2027, Westin Denver Downtown); Ahrefs Evolve (Oct 12–13 2026, San Diego) `[E]`.
- **Wynter**: message testing, verified B2B buyers, under 48 hours, credits at $1 each. The antidote to the slop risk: the deliverable stops being model output `[E]`.
- **G2**: listing does not exist, create it. Reviews are the only asset that answers the ChatGPT objection *inside ChatGPT* `[E]`.
## Writing rules

**Register:** Practitioner-to-practitioner. Write as one operator to another, not as a vendor to a market. Short declaratives. Concrete nouns: pitch, retainer, renewal, scope, non-billable hour, Thursday night. Assume the reader is sharper than you and has already seen four homepages that said the same thing this month.

**Never write:** agentic, AI-powered, end-to-end, transparent, seamless, unlock, empower, revolutionize, game-changing, supercharge, leverage (verb), "in today's fast-moving landscape," "not just X, it's Y," any tricolon of adjectives, any sentence that opens with "Imagine." Do not use "$300/hr," "without the agency markup," "six figures of expertise." Do not use "copilot," "GTM AI platform," "strategy memory," or "per-client memory."

**The one-em-dash rule:** At most one em dash per piece of copy. Prefer a period. Stacked em dashes are the single loudest AI fingerprint to this reader, and this buyer says they can spot the output from a mile away.

**Never invent:** customer counts, corpus size, time saved, win-rate lift, contributor numbers, prices, or outcomes. There are none `[E]`. If proof is required and absent, say it is absent and offer terms instead.

**Three verbatim buyer phrases to echo as voice reference:**
- "I can spot AI-generated copy from a mile away. And I hate it… It's all starting to sound the exact same."
- "I just got feedback from a client that is clearly from ChatGPT."
- "You can't win a price war with someone whose costs are near zero."

**Structural default for any asset:** open on the trigger event in the buyer's own week, name the unfunded strategic layer in their economics, show the tool disagreeing rather than describing that it disagrees, and close on a single live account rather than a workflow.