Five AI skills that turn your CRM into a system that remembers everything. This guide walks you through connecting an AI assistant to your HubSpot portal, setting up a folder of files as long-term memory, and installing five ready-to-run skills with full copy-paste prompts. No code required.
Any AI assistant that can do three things:
Several desktop AI tools can do all three today. Check that the one you pick is approved by your company for customer data before you connect it.
Client context lives in your head, your meeting notes and scattered docs. Every AI chat starts from zero: explain the client again, paste the context again. Generic AI tools don't know your pipeline, your deals, or what you promised last Tuesday.
The problem isn't your CRM. It's that nothing connects your CRM, your files and your day.
The HubSpot connection. Live access to your contacts, companies, deals and activity. Real data, not guesses.
A folder of files. A folder on your computer that becomes the assistant's long-term memory. Notes persist between sessions, which fixes the "AI forgets everything" problem.
Skills. Saved instructions for repeatable work. Build the workflow once, then run it with one line.
Teach it how to behave before you teach it what to do. Paste this into your assistant's project or custom instructions. These six rules prevent most AI-in-your-CRM horror stories.
You are my AI assistant for [your company] and our HubSpot portal. Rules: 1. Read HubSpot data freely. Never create, update, delete, or mass-edit records without my explicit approval. 2. Never send emails or messages in my name. Draft only. 3. Before building anything substantial, write a short plan and get my sign-off. 4. Take notes aggressively. Save important context, decisions, and client details to my files so you remember them next session. 5. If my request is vague or risky, push back and ask questions. 6. If an action is hard to reverse, stop and confirm first.
Add the HubSpot connection in your assistant's settings and sign in to your portal. Access matches your HubSpot user, so the assistant sees what you can see and nothing more. Test it with one question: "How many open deals do I have, and what's the total pipeline value?"
Start read-heavy. Let it earn write access after two weeks of output you trust.
This is the step most people skip, and it's the one that changes everything. Create one folder and point your assistant at it. Every meeting note, decision and client detail gets saved as a file. Next session, it reads the files and picks up where you left off. Copy this starter structure and adjust it once you see what you actually use.
/hubspot-ai
/clients one folder per client or account
/acme-co
context.md who they are, goals, preferences, history
/calls notes and summaries from every call
/briefs daily pipeline briefs
/reports monthly report narratives
/audits data hygiene audit results
/sops your processes, in plain English
/templates email templates, proposal language
Rule of thumb: if you'd want to remember it next month, it goes in a file.
A skill is a saved, reusable prompt with the steps, rules and formatting built in. Instead of retyping a 200-word prompt, you type one line: "Run my daily brief." Save the prompts below in your assistant, or as files in your folder, and say "follow the daily-brief instructions."
One morning summary: pipeline movement, at-risk deals, today's meetings with context, and your top three priorities. It saves 30 to 45 minutes a day of clicking through HubSpot. If your assistant can run scheduled tasks, set it for 7am on weekdays.
Create my daily HubSpot brief. From HubSpot, pull: 1. Deals that changed stage in the last 24 hours 2. Open deals closing this month, sorted by amount 3. Open deals with no activity in 14+ days, flagged as stale 4. My tasks due today and anything overdue 5. Today's meetings with the associated contact and company Then check /clients and /briefs for notes on anyone I'm meeting today. Format as: Pipeline Moves / At Risk / Today's Meetings (with context) / Top 3 Priorities. Save to /briefs/[today's date].md.
A one-page prep sheet that combines the live HubSpot record with your own notes from every previous call. It's the one skill that uses all three building blocks at once, and it works for quarterly reviews too.
Prep me for my meeting with [name / company]. From HubSpot: the contact and company records, open deals (stage, amount, close date), the last 5 logged activities and emails, and any open tickets. From my files: everything in /clients/[company], including prior call notes, decisions, and open action items. Give me one page: 1. Relationship snapshot 2. Where the deal stands 3. Open commitments, mine and theirs 4. Suggested agenda 5. One risk to watch Save to /clients/[company]/calls/prep-[date].md.
Drop in a transcript or rough notes and get a structured summary, a draft follow-up email and a CRM-ready log. This skill feeds the memory, so every processed call makes the next meeting prep better. Paste the whole raw transcript, mess and all.
Here are my notes / transcript from today's call with [company]: [paste] 1. Summarize: decisions made, action items (mine vs. theirs), open questions, and risks 2. Draft a follow-up email in my voice. Draft only, do not send. 3. Write a short call summary I can log on the HubSpot record 4. Save the summary to /clients/[company]/calls/[date].md and update /clients/[company]/context.md with anything that changed Ask before changing anything in HubSpot.
A monthly report card on your data quality with a prioritized cleanup plan. It's read-only by design: it reports, you decide. Run it monthly, because the trend matters more than any single audit.
Run a data hygiene audit on my HubSpot portal. Report only. Do not change any records. Check and count: - Contacts missing an owner, lifecycle stage, or email - Companies with no associated contacts - Open deals missing an amount or close date - Open deals with a close date in the past - Likely duplicate contacts (same email or name) For each category: the count, the 10 worst examples, and the source if visible. End with a prioritized cleanup plan, quick wins first. Save to /audits/hygiene-[date].md so we can compare next month.
Turns your monthly HubSpot numbers into a plain-English story for leadership: what happened, why, and what to do next. After three months of saved reports, ask "what trends do you see across the quarter?"
Build my monthly HubSpot report for [month]. From HubSpot: new contacts created, deals created / won / lost with amounts, pipeline value now vs. start of month, and top lead sources. Compare against last month's report in /reports. Write a plain-English narrative for leadership: 1. What happened (with the numbers) 2. Why it happened (what changed) 3. Three specific things to do next month No jargon. One page. Save to /reports/[year-month].md.
These are the same rules we set on every client implementation.
| When | What to do |
|---|---|
| Day 1 | Set up your assistant, paste the operating instructions, connect HubSpot. |
| Day 2 | Create the folder structure. Add a context.md for your top five accounts. |
| Day 3 | Run the daily pipeline brief, then schedule it. |
| Day 4 | Use meeting prep before one real call, then process it with the post-call processor. |
| Day 5 | Run the data hygiene audit and pick three quick wins. |
| Month 1 | Run the report narrator with a full month of data behind it. |
We set this up for HubSpot customers: the HubSpot connection, the memory structure, custom skills for your team, and the guardrails that keep your data safe.
Book a free 30-minute review