Unique Genius / Strategic Partner Enablement
Teaching SPs to use an AI companion for real task completion, built from an analysis of what 156 Strategic Partners actually do every day.
Prepared by Paul Gregorio · July 28, 2026 · Internal review draft
The proposal in one slide
Train Strategic Partners to use Claude Cowork as a working AI companion, not as a chatbot.
Where this comes from
All 156 Strategic Partner intake responses parsed on July 28, 2026. Only 2 of 156 left the task field vague, so the data is unusually complete for a self-reported intake.
Finding 1 · What SPs actually do
Bar length is the count of SPs. The badge is how well an AI companion performs that task, which is a separate judgment from how common it is.
Finding 1
A roofing CSR, an executive assistant, and a bookkeeper are doing recognisably the same shapes of work on different data.
First plan
Four cohort courses, starting with Executive Assistants. Each one would have taught the same five routines with different examples.
Four times the build. Four times the maintenance.
What the data says to do
One universal course. Cohort routing becomes a single page that tells each SP which two or three routines to build.
One course to build, correct, and keep current.
Finding 2 · The most important slide in this deck
| Task category | SP count | Companion fit | Decision |
|---|---|---|---|
| Inbound and outbound calls | 55 (35%) | None | Excluded. Named explicitly in the limits unit. |
| Calendar and scheduling | 58 (37%) | Low | Excluded. Mostly real-time judgment. |
| CRM data upkeep | 26 (17%) | Low | Excluded. No integration exists. |
Finding 3
71 of 156 SPs volunteered to mentor, with named depth in the exact systems the work happens in.
| Expertise offered | Mentors |
|---|---|
| Marketing | 12 |
| Jobber | 10 |
| GoHighLevel | 10 |
| Estimates | 7 |
| Video editing | 7 |
| Lead generation | 5 |
Why mentor-led beats central delivery
The roster
Home Service is roughly 70% of the roster, so mentor recruitment and rollout sequencing should be weighted there.
| Cohort | SPs | Dominant roles | Main systems |
|---|---|---|---|
| Home Service trades Roofing, holiday and permanent lights, gutters, exterior cleaning, landscaping, HVAC, plumbing | ~110 | CSR, admin, office manager, estimator, supplementer, AR and billing, SDR | Jobber, GoHighLevel, HouseCall Pro, ResponsiBid, CompanyCam, ServiceTitan, AccuLynx, Xactimate |
| Executive Assistant | 18-22 | EA, EA to CEO, Executive Partner | Google Workspace, Slack, Canva, Trello, Asana, Dubsado |
| Accounting and Bookkeeping | ~18 | Bookkeeper, accountant, payroll, tax, process auditor | QuickBooks Online, Canopy, Gusto, ADP, UltraTax |
| SMM, Web, Creative | ~14 | SMM, designer, video editor, SEO, Meta ads, web dev | Canva, CapCut, Meta Business Suite, Shopify, Semrush |
The core idea
| A chatbot | An AI companion | |
|---|---|---|
| You give it | A question, or pasted text | A folder, and an outcome |
| You get back | An answer in a window | Finished files, plus a log of every step |
| Who does the work | Still you | It does the assembly. You do the judgment. |
| Your job after | Copy it out and use it | Verify before it reaches anyone |
The one rule that changes results
The biggest adjustment for anyone coming from ChatGPT. With a chatbot you learned to script every step, because it could only do one. A companion works better when you describe the finished result.
| Chatbot habit | Companion instruction |
|---|---|
| "Write a follow-up for an unpaid invoice. Now write one for 30 days late. Now 60." | "Read overdue.csv. Draft a follow-up for each customer, matching tone to how late they are. Put drafts in a folder and give me a summary table." |
| "How should I format an estimate?" | "Read the measurements and the price list. Build the packet using the template. Flag anything the measurements do not cover." |
The curriculum
Not on frequency alone. That distinction is what keeps the program credible.
| No. | Routine | Serves | What it produces |
|---|---|---|---|
| 1 | Recurring report pack | 38 SPs | Multiple exports to one client-ready report, arithmetic shown, data problems flagged |
| 2 | Collections and follow-up batch | 62 SPs | A full chase batch drafted, tone matched to how overdue each account is |
| 3 | Estimate or proposal packet | 44 SPs | Packet assembled from measurements and price list, with what is missing listed |
| 4 | Review and reply batch | 26 SPs | Drafted responses in brand voice, decisions escalated rather than guessed |
| 5 | SOP from a recording | 17 SPs | An SOP written from a recording, every unconfirmed gap marked |
The pattern that governs four of the five routines
Cowork does not log in to Jobber, HouseCall Pro, ServiceTitan, ResponsiBid, AccuLynx, Xactimate, JobNimbus, Aspire, Canopy, or QuickBooks.
What stays manual
Pulling the export. Two minutes, once per report.
What stops being manual
Combining, calculating, formatting, drafting, and commenting. That is where the hour goes.
Routine 2 · The largest task category on the roster
SPs do invoicing, AR, collections, commissions, or payroll
40% of the roster
An aging export becomes a full drafted chase batch, with tone escalating by how overdue each account is, a summary table, and an escalation list of accounts the SP should handle personally.
| Days overdue | Drafted approach |
|---|---|
| Under 15 | Friendly reminder, assume they simply missed it |
| 15 to 30 | Direct. Restate the amount and original due date, offer the payment link |
| 31 to 60 | Firmer. Reference previous contact, ask for a specific payment date |
| Over 60 | Formal. Never states a consequence not written in the SP's own context file |
Routine 3 · Closest to revenue
SPs build estimates, proposals, or bids
28% of the roster
Measurements and a price list become an assembled packet, with a customer-facing summary written in the client's voice.
The four rules that make it trustworthy
Routine 5 · The one Unique Genius should care about most
SPs currently do documentation work
Only 11% of the roster, which is exactly the point
The SP records themselves doing a task once while talking through it. That recording becomes an SOP written for someone on their first day.
Why nobody does it today
Writing an SOP from scratch is genuinely tedious, so it is always the task that gets postponed.
Why it matters
It protects a client when someone leaves, and it makes an SP visibly senior.
The course
Beginner throughout. No coding, no terminal. Written to paste straight into Talent LMS, one unit per LMS unit.
| Units | Content | Who |
|---|---|---|
| 1 to 3 | What a companion does, getting access, chat versus Cowork | Every cohort |
| 4 | The rules before the tools. Client approval, the five never-categories, approval-required mode, draft-until-read, honest EOD reporting | Every cohort, mandatory |
| 5 | Workspace and context file. The highest-value unit. Delivered live on a shared screen because this is where beginners silently get stuck. | Every cohort |
| 6 to 10 | The five routines | Routed by cohort |
| 11 | Where Cowork stops | Every cohort |
| 12 | Certification on real client work | Every cohort |
Unit 5 · The highest-value ten minutes in the course
Cowork has no memory between sessions. A context file describing the client, the voice rules, and the standing rules is what gives it continuity, and it stops SPs re-explaining their client in every task.
The five standing rules every SP writes into their own file
Expectation management
Stated up front, to leadership and to the cohort. The fastest way to lose a cohort is to oversell.
| Not promised | Reason |
|---|---|
| It will handle phone work | It does not make or take calls. 35% of the roster's day stays human. |
| It will run Jobber, HouseCall Pro, ServiceTitan, AccuLynx, Xactimate | No integration exists. SPs export and work on the export. |
| It will dispatch or schedule in real time | It prepares the information. The SP still makes the call. |
| It replaces review | Every output is a draft until checked. The course is built around this. |
| Unattended automation on client systems | Deliberately excluded on risk grounds. |
Risk
It is one SP sending unverified AI output to a client, and the resulting fire being attributed to Unique Genius training. Six controls address that directly.
| Control | How |
|---|---|
| Client approval first | A one-pager the SP sends their POC before pointing AI at any client material. Kept in writing. |
| Hard data rules | Credentials, full card or bank numbers, government and tax IDs, named medical records, and anything the client excluded. Never entered. |
| Approval-required mode | The tool pauses for approval before acting. Mandatory for two weeks, permanent for client-visible work. |
| Verification in certification | Cannot pass without a specific verification step and a real reported first-run error. |
| Honest EOD reporting | AI assistance is named, not hidden. |
| No browser automation | Excluded. It would need client production credentials, may breach platform terms, and could mis-invoice a customer or double-book a crew. |
Larger than the training itself
Push the routines centrally
On Team and Enterprise plans an administrator can make standard tool packages available by default or as required. That turns "156 people follow a setup guide correctly" into "the routines are already there when they sign in." It removes the largest failure point in a rollout of this kind.
The library becomes a Unique Genius asset
Context templates, the five routine prompts, and verified export steps for ten systems are reusable across every future placement and intake. Built once.
A positioning claim competitors cannot easily copy
"Our Strategic Partners come AI-enabled" is hard to match, not because the tool is exclusive but because the verified routines, guardrails, and certification are the difficult part.
Delivery
| Wave | Who | Prerequisite |
|---|---|---|
| 0 | Mentor session plus mentor deliverables | Starts as soon as seat access is confirmed |
| 1 | Pilot, capped at 8 to 12 volunteers from one cohort | Wave 0 complete |
| 2 | Home Service, about 110 SPs across 4 to 5 mentors in groups of about 20 | Pilot corrections applied |
| 3 | EA, Accounting, SMM in parallel | Mentors already trained |
What actually gates the rollout
The material is drafted. Two things gate delivery, and only one of them is a decision.
Seat access confirmed
Cowork requires a paid plan and cannot be opened on a free account. If SPs reach unit 2 and cannot proceed, the cohort's attention is lost and does not come back.
Export steps verified and tested by mentors
Ten systems. Each mentor exports from their own system, runs the matching routine against the real file, and confirms it works. Those steps are deliberately left blank in the course rather than written from memory.
Over to you
Decision 1 sets the start date. The rest can be resolved in parallel.
| No. | Decision | Why it matters | Owner |
|---|---|---|---|
| 1 | Seat access. UG seats, client-paid, or SP-provided? | The only true blocker. No paid plan means no Cowork at all. | Angela / Mitch |
| 2 | Company position on SPs using AI on client data | The course currently teaches "ask your POC". A company policy would supersede it. | Angela / Aljuri |
| 3 | Pilot capped at 8 to 12 from one cohort | Lets certifications be reviewed properly and the material corrected before 110 people see it. | Angela |
| 4 | Certification reviewer, and a live session slot | Unreviewed certifications mean the verification discipline is not enforced. | Mitch |
| 5 | Do we have Team or Enterprise, and can packages be pushed centrally? | Would remove the largest failure point in the rollout. | Angela / Rochelle |
Next step
Mentor recruitment against the cohort sizes can begin before that. Materials are drafted and ready for review: the course, the mentor track, the client approval one-pager, and the roster analysis.
Roster figures from all 156 intake responses, parsed July 28, 2026 · Tool capabilities verified against current vendor documentation the same day
Slide index