Under the hood
Use the right engine for the job
OpenClaw, Hermes, Claude Code, Codex, local models, and future tools are infrastructure. ProductiveBot is not a reseller of any one engine.
00 · Start here
If you read one screen before your first meeting, read this one.
Everything below is the short version. The numbered sections expand on it. New to the team? This page plus §1 and §2 is enough to hold a real conversation.
Last verified 2026-08-19
What we sell
An AI agent workstation the business owns — a Mac mini that arrives set up, with the support, memory, permissions and workflows around it. $1,499, ships within 1–2 weeks.
Who buys it
Owners of small businesses — roofing, restaurants, agencies, real estate — who believe AI can help but do not want a technical project. They want time back and a useful first result.
Why we win
Not the engine. Anyone can run an agent. We do guided setup, repair when it breaks, durable memory, and permissions — the things a buyer cannot judge from a demo video.
What we are not
Not a chatbot, not a reseller of one AI vendor, not a replacement for staff, and not a Mac mini with software dropped on it.
Three things to never say
- Never claim HIPAA compliance. We are not certified. Describe the architecture instead — data on hardware the customer owns, approval required for sensitive actions.
- Never invent a number, testimonial, customer name or rating. If a claim is not in Proof Points with a green status, it does not go in front of anyone.
- Never promise a partner commission, discount, or territory that is not in §6 Partnerships. Rates and tiers are set; anything custom needs Alex.
Who owns what
| Question | Ask |
|---|---|
| Pricing, discounts, anything contractual | Alex |
| Can I say this publicly? / claim status | Alex — check §5 first |
| Partner terms, commissions, payouts | Alex — terms in §6 |
| Product behaviour, setup, what ships | Scout, then Alex |
| Brand, visuals, tone | Brand Guide |
Part one — what you can say externally
SettledSections 1–6 are agreed and safe to repeat in customer, partner and public conversations, subject to the claim statuses in §5. If something here is wrong, it is a bug — tell Alex.
01 · The thesis
ProductiveBot helps businesses use AI to finish real work on hardware they own.
This is the core message. Use it to check ads, web copy, and sales answers. If a message conflicts with it, rewrite the message.
Last verified 2026-08-19 · stable since June
The customer system
ProductiveBot provides the practical support around AI
Setup, onboarding, Scout and Doctor, memory, permissions, skills, workflows, diagnostics, repair, and managed updates.
The outcome
Completed business work
Support, operations, research, scheduling, content, lead follow-up, admin, reporting, and internal workflows—not “an AI.”
Why this framing holds: a better engine can arrive without forcing the customer to rebuild their working system. See Competitive Landscape for the current market context.
| ProductiveBot is | ProductiveBot is not |
|---|---|
| An AI agent workstation your business owns and controls. | Just a chatbot. |
| The setup, support, memory, permissions, diagnostics, and workflow layer around practical agents. | Just a Mac mini with OpenClaw installed. |
| A way for teams to turn conversations into completed work. | A generic AI productivity companion. |
| A controlled business system for choosing agents, models, tools, files, and workflows. | A fully autonomous employee replacement. |
02 · The buyer
Buyers want a useful first result, not a technical setup project.
The clearest recurring buyer is an owner who believes AI can help but does not want to assemble and maintain a fragile technical system.
Last verified 2026-08-19
Who they are
- Owners and operators of small businesses.
- Examples include roofing, restaurants, stores, agencies, and real estate.
- They think AI could change how they work.
- They need help before they can hire more people.
What they actually care about
- They care less about which AI model is underneath.
- They care whether it helps their business.
- Setup must feel complete, not like another technical task.
- They want time back and a useful first result.
First run is the highest-trust moment
In the first session, customers are ready to share context and explain what success looks like. Do not spend all of that time on technical setup.
Scout is a role, not a chat box
Scout helps with setup, support, and fixing problems. The customer’s agent helps with their business work. Use “Ask Scout” and “Open My Agent,” not one vague “Open Chat” button.
Getting started is the biggest challenge
Capable technology still needs coaching. One roadblock can make an owner give up. Treat a confident setup and first result as part of the product.
03 · Language guardrails
Say this. Not that.
These are not just style preferences. Most exist because a specific phrase created confusion, overpromised, or caused another real problem.
Last verified 2026-08-19 · one known conflict with the live homepage — see "AI employee"
Use often
Approved- AI assistants that turn conversations into completed work
- Practical AI agents for real business workflows
- An AI agent workstation your business owns and controls
- Mac mini included
- Own the workstation. Choose the agents.
- Your computer. Your agents. Your control.
- Human-in-the-loop AI systems that keep people in control
- Runs on your own hardware, not in someone else's cloud
- Your files, passwords, and sensitive data stay on your device
- Setup, support, memory, permissions, diagnostics, and workflows included
Avoid
Do not state- "Just a chatbot" / "just a Mac mini with OpenClaw installed"
- "A generic AI employee on your desk"
- "Local AI" as a headline — Alex found it confusing
- "AI workstation" alone when it is unclear what is being bought
- "Mac Mini" — the correct capitalization is Mac mini
- "No monthly fee" as broad public copy — use "no required cloud agent subscription"
- Fully autonomous / 24÷7 / remembers everything / replaces people, unless the workflow supports it
- ProductiveBot.com — the domain is productivebot.ai
- Invented metrics, testimonials, named customers, ratings, or review counts
Use carefully
Conditions attached04 · The hard questions
Common questions, with answers we can stand behind.
These are short answers. The source list at the bottom links to the fuller explanations, evidence, and current claim status.
Answers last verified 2026-08-19
Objection
"Why not just use Claude or the ChatGPT app?"
Those are engines and interfaces — we can run them. An agent alone is not a business system. It does not come with a dedicated always-on machine, a second agent that can check and repair the first, durable cross-agent memory, governed access to email and sheets and passwords, support when it breaks at 9pm, or backup. If the agent is the engine, ProductiveBot is the vehicle, the fuel, the mechanic, and the garage.
Objection
"What stops Anthropic or OpenAI from building this?"
Their business model depends on you renting intelligence from their cloud. They cannot become us without cannibalizing it — and they have no reason to build hardware supply, provisioning, SMB support, and a skills ecosystem for roofing companies. We are model-neutral: if the frontier labs win, we ship them; if open source wins, we ship that.
Objection
"Hermes works fine for me — why is this better?"
We love Hermes. We ship it. The question is what surrounds it: the workstation, the second agent, memory, security, connectors, support, backup, and the interfaces where real work lives. Our skills are built to work across engines rather than betting on one.
Objection
"You say I own it, but you still send updates."
You own the hardware, the data, the memory, and the workspace. Cancel a SaaS seat and your AI evaporates — unplug a ProductiveBot and you are still holding the entire system. Your iPhone gets updates and it is still yours. We are the update boundary that tests releases before they reach you.
Objection
"Are you HIPAA compliant?"
Separate architecture from certification: we are not HIPAA-certified and do not claim to be. Architecturally, data lives on hardware the customer owns, credentials stay on device, and sensitive actions require approval — a better starting point than sending PHI to a third-party cloud. Formal compliance is a roadmap item, not a re-architecture.
Never say yes. Do not state or imply HIPAA compliance in any surface.
Objection
"How technical do I need to be?"
This is the real #1 challenge and we should answer it honestly: the technology is capable, but a little coaching goes a long way. That is why the roadmap targets exactly this — Scout Guide, prebuilt workflows, training, Company Brain. Naming the challenge and showing the systems built against it lands better than claiming it does not exist.
05 · Proof points
What we can say in public.
Every claim below has a status. If it is not on this list, do not use it in public. Never invent numbers, testimonials, ratings, or customer names.
Statuses last verified 2026-08-19 · recheck when a reconciliation or permission closes
| Claim | Status | How to use it |
|---|---|---|
| Zero returns | Verified | Safe. Consistent across our record since March. |
| 60+ paying customers | Use the safe form | Use "60+" publicly. Reconciliation now closed: 65 paid orders, 64 reachable by email (one order completed without capturing an email); 61 was the earlier de-duplicated purchaser count. "60+" stays the safe public form. |
| Scout ~98% resolution | Define first | The 98% is measured over 185 classified outcomes, not the 1,259 total logged conversations — different denominators. Needs a stable public definition before marketing use; internally it is real operational data. |
| Pizza World — $90K in 3 months | Permission pending | Story is real, but still needs public-name / logo / quote / claim permission before external use. Do not put the $90K figure in front of a customer or partner until that package is signed. |
| Ships within 1–2 weeks | Interim — site conflict | Interim SLA agreed 08-18. The live site currently states two different things ("Ships within 1-2 weeks" and "Ships in 3-5 business days"). Use 1–2 weeks and get the site reconciled before promising either to a partner. |
| Jed — working prototype in ~45 min | Verified | Against $50K–$200K development quotes. |
| Apple distribution (Redstone, CDW) | Spelling check | Strong story: Apple initially shut us down; we now buy through certified Apple distribution. Confirm "Redstone" entity spelling before print. |
| "Best money I've ever spent" | Handle carefully | Real customer sentiment. Do not present as a quantified testimonial without permission and exact source. |
| Built with ProductiveBot | Verified | The website, support system, marketing, research, and presentations were made with it. Strongest available proof of the promise. |
| Flor static creatives — cost | Unverified | No invoice, rate, or contract exists in our record. Do not state a number. |
06 · Partnerships
What a partner is, what they earn, and what you may promise.
Everything in this section is live and documented. The commission tiers, cookie window and payout terms below are the actual program terms — not proposals. Anything outside them needs Alex before you say it.
Last verified 2026-08-19 · source: affiliate program terms + partner quick-start
Partner types we actually have
- Advocates / affiliates — individuals and creators who refer business. Tracked through GoAffPro, paid per unit.
- Agencies & resellers — serve SMB clients and place ProductiveBot into them. Same commission structure today.
- Skill Store developers — build skills for the ecosystem, keep 95% of skill revenue during the founding era.
Applicants are verified against existing-customer records and approved by Alex. There is no auto-approval, and the portal does not grant approvals on its own.
What a partner gets beyond money
- Founding-partner community access and a direct line to Alex and the product team.
- Association with a growing niche brand, and early access to what is shipping.
- A path to contribute skills to the Skill Store ecosystem.
In exchange we ask for authentic advocacy — real use, real recommendation. Commission is the reward for advocacy, not the reason for it.
Commission tiers — rolling 90-day
| Tier | Rolling 90-day sales | Commission per unit |
|---|---|---|
| Partner | 1 to 5 units | $100 per unit |
| Pro Partner | 6 to 15 units | $125 per unit |
| Elite Partner | 16 or more units | $150 per unit |
Flat dollar per qualifying sale — not a percentage of order value. Not paid on shipping, taxes, or refunded amounts. $150/unit is the ceiling on direct sales. Tiers reset on a rolling 90-day window; unlocked status does not reset.
Getting paid
Net-30 from date of sale, once the hold period closes. Minimum payout threshold is $100. Refunds and chargebacks inside the hold void the commission.
Attribution
60-day cookie. A sale inside that window is attributed to the partner. Coupon-code attribution also applies.
Required disclosure
Partners must follow FTC endorsement rules and disclose the relationship. This is not optional and applies to every social post.
What you may not promise
No custom rates, exclusive territories, guaranteed placement, co-marketing spend, or permanent tier status. The program terms explicitly reserve the right to change rates and tiers with 30 days' notice (90 days for Skill Store revenue share). Anything a partner asks for that is not on this page goes to Alex first.
Part two — how we are still deciding
In motionEverything below is a snapshot of live thinking, not settled policy. It changes. Do not quote it to a customer or partner, and do not treat a disagreement here as a decision. When you need one of these answered, ask Alex rather than assuming.
07 · Awareness levels
Start with what the buyer already understands.
Someone who feels buried in scattered busywork does not need to hear about an “owned AI agent workstation” first. Start with the problem they recognize, then show the relevant ProductiveBot help.
Paid-media context · 2026-06-29 brief — skip unless you write ads
Problem Aware
Knows they are overwhelmed. Not shopping for an agent.
Pain-driven hooksSolution Aware
Believes agents can help. Needs a credible solution.
Promise-driven hooksProduct Aware
Knows ProductiveBot, OpenClaw, or Hermes.
Proof-driven hooksMost Aware
Close to buying or retargeted.
Offer-driven hooksThe full hook library lives on the Value Map — every awareness level with example hooks, the recommended hook type, and the documented defect where we ran product-aware hooks at a solution-aware audience. Open the hook map →
08 · Competitive landscape
What buyers can get elsewhere, and where ProductiveBot still needs to be different.
This is a snapshot from 2026-08-11. Competitor capabilities change quickly, so check the current research before quoting anything here.
Snapshot 2026-08-11 · expires 2026-09-15 — recheck docs/competitive-landscape/ before quoting
A · Agent harnesses
ChatGPT desktop, Claude Code, OpenClaw, Hermes, and Buzz. These tools are changing fast and are easier to copy.
B · Cloud agent products
Viktor, Claude Cowork, and Perplexity Computer. They aim for a similar result in a system the vendor owns.
C · Agent hardware
Otto, Silvr Stone, and Silvr Mini. They show that people will consider an AI device they own.
D · Journey-adjacent
LM Studio, Buzz, model reviews, and paid AI groups. They are not direct competitors, but they shape what buyers expect.
What changed in August 2026
A well-known practitioner ranked the ChatGPT desktop app above OpenClaw, Hermes, and Claude Code. ChatGPT mobile can also control work from a phone. We cannot use either claim as our main difference.
What this takes off the table
“It controls your computer” and “message it from anywhere” are now common features in consumer products.
What ProductiveBot still needs to do well
It is vendor-hosted. Its demo has full access with no approvals. It does not solve setup for a non-technical owner, lasting memory, help when it breaks, or clear support. Those are the problems ProductiveBot must solve.
| Capability | ChatGPT desktop | Claude Code | OpenClaw / Hermes (raw) | ProductiveBot |
|---|---|---|---|---|
| Computer / browser use | Strong | Partial | Yes | Yes |
| Phone-driven remote control | Yes | No | Via channels | Yes, multi-channel |
| Model choice | OpenAI only | Anthropic only | Any | Any — chosen and maintained for you |
| Persistent business memory | Threads/projects | Project scoped | Files | TotalRecall, source-attributed |
| Permissions / injection defense | Vendor policy | Vendor policy | Config | PromptGuard · SkillGuard · SecuritySuite |
| Guided setup for a non-technical owner | No | No | No | Yes — this is the product |
| Diagnostics / repair / recovery | Vendor support | Vendor support | Community | Scout · Doctor · ProductiveBot+ |
| You own the machine and the data | No | No | Yes (BYO) | Yes, included |
Honest reading: the top rows are converging fast. The bottom rows are where we are genuinely differentiated — and they are exactly the rows a buyer cannot evaluate from a demo video.
Otto — the sharpest hardware comparison
Otto One: $599 one-time, $20 deposit, ships Nov 2026. This paragraph expires when Otto ships (Nov 2026) — recheck before using it after that. An always-on Linux mini preloaded with OpenClaw — the same engine we ship. Expect the price question. The answer is not a spec fight: it is a full Apple computer the business owns and can use beyond agents, plus guided setup, support, security, memory, and model flexibility.
Verified
Otto One pricing and ship date
Verified 2026-08-11 via competitor-watch
The unbuilt opportunity
Nobody productizes model routing. Practitioners hand-route across four vendors and five apps because no product does it for them. "We picked the right model for each kind of work so you don't have to" is a concrete value prop that turns engine-neutrality into a feature rather than a hedge.
09 · Open tensions
Decisions we have not made yet.
These questions are intentionally unresolved. If you need one answered, ask Alex instead of making an assumption in a customer-facing asset.
Reviewed 2026-08-19 · if an item here is unchanged in a month, nobody made the decision
Not resolved
Does "workstation" lead?
The documented line is "an AI agent workstation your business owns." That framing was directly challenged in the 08-10 strategy meeting — the argument being that buyers do not care about workstations and do not understand what "owning your AI" means, and that safety and security should lead instead ("sell medicine, not vitamins"). Alex received the challenge well.
Resolved 08-19
Pricing: resolved — $1,499 is live
The live price on productivebot.ai is $1,499, one price, no variants — matching docs/product-truth.md. A $900 one-time structure was discussed once in the 08-10 meeting and was never adopted.
Quote $1,499. Do not quote $900 — explored, not adopted. New pricing structures need Alex before use.
Live tension
Education vs. bottom of funnel
The accepted pivot is 80% organic / educational. But Alex's explicit warning stands: people in the Arch Grants room wanted to buy on the spot. Do not over-rotate to education and abandon the bottom of the funnel.
Appendix · internal only
Working notes, not guidance.
Kept for the people who were in the room. New readers can skip this entirely — nothing here is a decision or an instruction.
How the 08-10 strategy pivot was argued
The strategy pivot was argued from analogy and from the Arch Grants experience — not from ad performance data. No CTR, CPC, session, or view-through numbers were presented. 19 creatives launched into a shrinking exact-match budget with conversion tracking only just verified. That cannot produce a reliable read. "We have no signal yet" and "ads don't work for this product" are different problems and need different responses.
10 · Where it lives
Where to find the details behind this page.
This page is a summary. Use the documents below when you need the full reasoning, evidence, or exact approved wording.
These are internal repo paths, not links
They live in the ProductiveBot repository and are not reachable from a browser. If you do not have repo access, ask Alex for the specific document rather than trying to open the path.
| If you need… | Document |
|---|---|
| How the product actually works, day to day, plus six customer scenarios and creative prompts | docs/dora-questions-answered.md |
| Awareness levels, hook directions by theme, ad-creative matching, evaluation checklist | docs/rebeka-meta-positioning-brief-2026-06-29.md |
| Best phrases, avoid list, Scout/agent roles, first-run messaging | docs/productivebot-current-messaging-brief-2026-06-27.md |
| 22 hard questions with talk tracks, proof, and status flags | artifacts/arch-grants-2026/qa-index-answer-bank.md |
| Canonical product facts — what ships, what's included, channels, providers | docs/product-truth.md |
| Customer insight index — personas, onboarding reactions, support pain | docs/customer-feedback-and-insights.md |
| Brand thesis, say/avoid, visual system, channel guidance, QA gate | artifacts/pb-table-*.md (5 tables) |
| Current competitive snapshot + monitoring status | docs/competitive-landscape/ |
| Pitch-facing competition framing and market sizing | docs/arch-grants/ |
| Visual system, logo usage, product screens, content standards | The brand guide → |
Keeping this page honest
Competitive claims expire — recheck §5 against docs/competitive-landscape/ before quoting it after mid-September. Proof-point statuses change as reconciliations close. Open tensions should shrink over time; if one is still listed months from now, that is a decision nobody made, not a permanent state.