Lighthouse Chart — Jordan Reyes, Ops Lead

1. Opening / Bearings

You are spending real hours every week on work that follows the same shape every time: find the right words for a partner you have never met, pull scattered notes into a coherent status update, write down a process that lives in your head so someone else can follow it. The goal here is not to hand that work to a machine — it is to build a small, reliable set of AI-assisted workflows that cut the repetitive load so you can spend your judgment on the parts that actually need it.

Four weeks. Three sessions a week. By the end, you will have sent five AI-drafted partner outreach emails, a weekly summary process that takes under 20 minutes, and at least one SOP in active team use with an AI-drafting step baked in. Those are the three outcomes this plan is built around, in that order.

You already know what a prompt is, you have used ChatGPT occasionally, and you work in Notion and HubSpot daily. That is enough to start. The gap between where you are and where this plan ends is not technical — it is about building the habit of giving AI enough context to do useful work, and then storing that context so you do not start from scratch every session.

That is what the next four weeks are for.

2. Track Context

Your work sits squarely on the Ops and Partnerships track. The defining feature of this track is that the highest-value AI use is not creative generation — it is systematic, repeatable output from structured inputs. You are not trying to write something original every time; you are trying to produce something consistently good from a known set of inputs (a partner name, a mutual benefit angle, a week's worth of raw notes) without burning 45 minutes doing it.

The two failure modes on this track are worth naming early so you can avoid them.

The first is the blank-page prompt. You open Claude, type "write me an outreach email," get something generic, decide AI is not useful for this, and go back to writing it yourself. The fix is specificity: the more context you give the model about the partner, the angle, and the tone, the more the output looks like something you would actually send. The ready-to-paste prompts in this chart are built to give you that specificity from day one.

The second is the one-session wonder. You build a great prompt, use it once, and then cannot find it the next week. The fix is storage: every prompt you refine goes into a Notion page immediately, and by week two it also lives in OpenBrain so it travels with you across Claude sessions. That is the whole system — good prompt, stored prompt, reused prompt.

The upcoming partner onboarding project is a natural extension of this track. It is not in scope for the four-week plan (you listed building a full automation platform as a non-goal, and that is the right call), but the stage-mapping work in week three gives you a foundation to build on after this plan ends.

One more thing: you have HubSpot in your stack. Claude cannot write directly into HubSpot, but the workflow is fast — draft in Claude, paste into HubSpot, send. That two-step process is the right shape for now. It keeps a human in the loop on every send, which matters for partner relationships, and it does not require any IT approval beyond what you already have.

3. Goals and Success Metrics

Three outcomes, in priority order:

First: Automate the first-touch partner outreach sequence with AI-drafted emails. This is the highest-leverage outcome because it compounds. Every partner you reach out to is a potential future workflow, integration, or referral. The time you save on drafting is time you can spend on the follow-up conversations that actually move deals. The success bar is concrete: five outreach emails drafted and sent using AI templates. Not five perfect emails — five emails that went out, were good enough to send, and were produced faster than your current process.

Second: Cut the time to produce weekly ops status summaries by 60%. Your current process almost certainly involves hunting across Slack, Notion, and your own memory to reconstruct the week. The AI workflow here is not about writing — it is about synthesis. You paste in the raw material; Claude structures it. The success bar is under 20 minutes from raw notes to finished summary. That is a hard number you can measure from week one.

Third: Build a reusable SOP template that includes an AI-drafting step. This is the one that outlasts the four weeks. A good SOP template means the next person who joins your team, or the next process you need to document, does not start from zero. The AI-drafting step embedded in the SOP is the part that makes it a force multiplier — it is not just documentation, it is documentation that tells the reader exactly how to use AI at the right moment in the process. Success bar: one SOP in production use by end of week four.

What this plan is not: This plan does not build a full automation platform — no Zapier chains, no API integrations, no code. It also does not replace your judgment on partner deals. The AI drafts; you decide. That line stays clear throughout.

4. Baseline and Pacing Note

You write with AI at a solid 3 out of 5 — you have used it enough to know it can help, but you have not yet built the muscle of giving it the right inputs consistently. Research and building are at the same level. Automation is at 2, which is fine: nothing in this plan requires you to build automations. The highest-ceiling skill to develop over four weeks is prompt specificity — the ability to give Claude enough context that the first draft is 80% of the way there.

You know what a prompt is, what a template is, and what an API is. You do not need to understand APIs to do anything in this plan. The term that will matter most is "system prompt" — a set of standing instructions you give Claude before your actual request. You will build one for partner outreach by week two.

Pacing: three sessions a week at 45 minutes each is 135 minutes of focused work per week. That is enough to build and refine, not enough to overbuild. The plan is sequenced so that weeks one and two are about getting the outreach and summary workflows to a usable state, week three is about the SOP template and onboarding prep, and week four is about getting things into production and doing a retrospective. Each session has one deliverable. If a session runs long, stop at 45 minutes and pick up the deliverable next session — do not let one session bleed into the next.

5. Tasks and Projects

Partner outreach emails

This is where the plan starts, because it is your highest-priority outcome and because it gives you an immediate, tangible result you can measure.

The current state: you write outreach emails from scratch, probably starting with a blank document or a vague mental template. The result is inconsistent quality and inconsistent time investment. Some emails take 20 minutes; some take 5. The AI workflow makes the floor higher and the ceiling consistent.

The workflow you are building looks like this: you have a single prompt in Notion (and, by week two, in OpenBrain) that you paste into Claude.ai. You fill in three inputs — the partner's name, what they do, and the specific mutual benefit angle. Claude produces a draft under 150 words. You spend five minutes editing for tone and accuracy. You paste it into HubSpot and send.

That is it. The prompt does the structural thinking for you. You bring the context that only you have: why this partner, why now, what the angle is.

In session one, you build and test the prompt. In session two, you stress-test it across two different partner types. In session five, you send five emails. By the end of week two, the first success metric is complete.

The outreach prompt and the partner onboarding stage map you build in week three are connected: once you have a partner interested, the onboarding workflow is the next thing they experience. That context — what your outreach promises, what the onboarding delivers — is worth capturing in OpenBrain by week three so it travels with you when you start planning the onboarding project in earnest.

OpenBrain binding: Your first-touch partner outreach sequence — the refined prompt, the partner context, and the five sent emails — will live in OpenBrain by Week 2. This is not a file on your laptop; it is a queryable memory layer attached to your Claude sessions. In any future Claude conversation, you can ask "what is my current partner outreach prompt?" and get it back without opening Notion.

Weekly ops status summaries

The current state: you compile the weekly summary by pulling from Slack, Notion, and memory. The synthesis is the expensive part — not the writing, but the hunting.

The AI workflow here inverts the process. Instead of synthesizing first and then writing, you dump the raw material into Claude and let it synthesize. Your job becomes two things: (1) making sure the raw material is easy to grab, and (2) reviewing the output for accuracy.

Session three is your first run. You will collect this week's raw notes, paste them into the summary prompt, and time the whole process. That time becomes your baseline. Session six is your second run — by then you will have a sense of where the time goes, and you can start to tighten it. Session nine introduces a note-gathering method: a single Notion page or Slack channel where you drop notes in real time during the week, so that by summary time you are copying from one place instead of hunting across five.

The 20-minute target is achievable. The bottleneck is almost always note-gathering, not writing. Once you solve the gathering problem, the AI handles the writing in under two minutes.

Internal SOPs

The SOP template is the most durable thing you will build in four weeks. A template with an embedded AI-drafting step means that every SOP your team writes going forward has a built-in forcing function: at this step in the process, open Claude, use this prompt, review the output.

Session seven is the first draft of the template, using the partner outreach process as the subject. That process is the right choice because you already know it well and you have just spent two weeks refining the AI prompt for it — the AI-drafting step writes itself.

Session eight extends the template to a second process (weekly summary or onboarding intake) and stress-tests the format for consistency. By session eleven, the template is in production: marked Active in Notion, shared with a teammate, and incorporating one round of feedback.

The SOP template is also the foundation for the upcoming partner onboarding workflow project. When you start that project, you will not be starting from zero — you will have a tested template, two working examples, and the onboarding stage map from session ten already stored in OpenBrain.

Upcoming: partner onboarding workflow

This project is not in scope for the four-week plan, but session ten is specifically designed to give you a head start. You will use Claude as a thinking partner to map the onboarding stages, identify where AI can assist at each stage, and store that context in OpenBrain. When the project formally kicks off, you will be able to open a Claude session and ask "what did I capture about the partner onboarding workflow?" and get the stage map back immediately.

6. Setup and Constraints

Devices: You are on a MacBook Pro and a Windows desktop. Claude.ai Pro works in any browser on both. Claude Desktop (for OpenBrain) is available for both macOS and Windows — install it on whichever machine you do most of your focused work on. If that is the MacBook, start there.

Accounts: You have Claude.ai Pro, Notion, and HubSpot. That is everything this plan needs. No new tools are required in weeks one and two. The only addition is Claude Desktop for the OpenBrain setup in session four, and that is a free download under your existing Claude Pro account.

IT approval: Claude Desktop and the OpenBrain MCP server are the only new software installs in this plan. Both are open-source and run locally — no data leaves your machine through OpenBrain. That framing is usually sufficient for IT approval, but check with your team before session four. If approval takes time, sessions one through three work entirely in the browser-based Claude.ai and do not depend on Claude Desktop being set up.

Budget: Everything in this plan runs on your existing Claude.ai Pro subscription. No additional paid tools are introduced. The Notion pages and HubSpot records you create are within your existing accounts.

A note on proprietary content: Your partner outreach context and ops notes are business-sensitive. Claude.ai Pro does not use your conversations to train models by default — you can confirm this in your Claude.ai account settings under Privacy. If your IT policy requires stricter data handling, raise it before session three when you start pasting raw ops notes.

7. OpenBrain — Foundational Setup

OpenBrain is a personal knowledge layer that travels with you across Claude.ai, Claude Desktop, Claude mobile, and any other Claude surface you use. Think of it as a persistent memory attached to your Claude sessions: context you capture in one session is available in the next, without copy-pasting or re-explaining. It is designed to be set up in under 30 minutes by anyone who can install a desktop app and follow a README.

Setup steps for your devices:

On your MacBook Pro (recommended starting point):

1. Download and install Claude Desktop from claude.ai/download.
2. Open the OB1 README at https://github.com/NateBJones-Projects/OB1 and follow the setup instructions to add the OpenBrain MCP server to Claude Desktop. The README walks through each step — it does not require coding experience.
3. Once set up, open Claude Desktop and start a conversation. You will see OpenBrain available as a connected tool.

On your Windows desktop: the same Claude Desktop installer is available for Windows. If you primarily work from the Windows machine, run the same setup there. OpenBrain syncs across devices, so context captured on one machine is available on the other.

Track-specific usage patterns for Ops and Partnerships:

First pattern — prompt storage. Every time you refine a prompt (outreach, summary, SOP), paste it into a Claude Desktop conversation and ask Claude to store it. Label it clearly: "This is my current partner outreach prompt — please remember it." Future sessions can retrieve it with a simple question. This replaces the Notion prompt library as your primary retrieval point, though keeping both is fine.

Second pattern — project context. Before starting a new project (like the partner onboarding workflow), open a Claude Desktop session, describe the project scope, the constraints, and what you already know. Ask Claude to learn it. Every future session on that project starts with full context instead of a re-explanation.

Third pattern — decision log. When you make a meaningful ops decision — a process change, a partner prioritization call, a template revision — describe it briefly in a Claude session and ask Claude to learn it. Over time, OpenBrain becomes a searchable record of your reasoning, not just your outputs.

The "learn session" loop:

At the end of any working session where you produced something meaningful — a refined prompt, a completed SOP draft, a stage map — end the session by telling Claude: "Learn this session." Claude will summarize and store the key outputs and decisions. The next time you open Claude, that context is there. This single habit is what makes OpenBrain compound over time rather than staying flat.

8. Resources and Links

The resources below are grouped by how much time you have. Start with the 5-minute reads before session one. Return to the 30-minute resources when you hit a specific friction point. The deep dives are there when you want to go further after week four.

5 minutes — read before you start

Anthropic Prompt Engineering Overview https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview The single most useful reference for understanding why some prompts produce useful output and others do not. Read the sections on specificity and context before session one. You do not need to read it end to end — skim for the principles, come back to the examples when a prompt is not working.

OB1 README — OpenBrain setup https://github.com/NateBJones-Projects/OB1 Read this before session four. The README is the setup guide for OpenBrain. It is written for non-developers. Budget 20 minutes for the actual install.

30 minutes — when you hit a specific friction point

Claude.ai prompt library (Anthropic) https://www.anthropic.com/prompt-library A curated set of example prompts across use cases. Useful when your current prompt is producing weak output and you want to see what a well-structured prompt for a similar task looks like. Filter for business writing and summarization use cases — those map directly to your track.

HubSpot CRM documentation — email templates https://knowledge.hubspot.com/email/create-and-send-email-templates Relevant when you are ready to store your finalized outreach email as a HubSpot template so you do not need to paste from Claude every time. This is a natural week-three or week-four optimization, not a week-one task.

Deep dive — after week four

Notion AI documentation https://www.notion.com/help/notion-ai If you want to bring AI assistance directly into your Notion SOPs and summaries, Notion AI is the natural next step. It is a separate add-on from your current Notion account. Worth evaluating after you have the Claude-based workflows stable — do not introduce it before then.

Anthropic cookbook — business use cases https://github.com/anthropics/anthropic-cookbook A collection of practical patterns for using Claude in business workflows. Most relevant after week four when you want to go deeper on structured output, longer document processing, or more sophisticated prompt chaining. Not required for anything in the four-week plan.

9. Closing / Signals from the Keeper

Four weeks from now, you will have built three things that did not exist before: a partner outreach process that takes minutes instead of an hour, a weekly summary workflow that runs in under 20 minutes, and an SOP template that teaches the next person on your team how to use AI at the right moment in a process.

None of that requires becoming a technical expert. It requires giving AI enough context to do useful work, storing what works, and being willing to edit the output rather than accept it wholesale.

The signal that you are on track: after session two, you should be able to produce an outreach draft you would actually send in under 10 minutes. If you cannot, the prompt needs more context — not more time, more specificity. Come back to the prompt engineering overview and look at what inputs you are leaving out.

The signal that you are ahead of schedule: by week three, a teammate asks how you are producing summaries so fast. That is when you hand them the SOP.

Start with session one. The path is clear from here.