How a fractional CMO uses AI

AI drafts. I decide.

I am Rolf Tjalsma, a fractional CMO for B2B SaaS companies. I use AI every day for market research, GEO copy, sales call analysis, meeting follow-up and recurring reporting. This page shows my rules, my tools and where a person always signs off.

Claude

ChatGPT

Mistral

Perplexity

Granola

Grok Bot

Claude

Drafting for Rolf

Draft

Summarize this week’s sales calls for Friday’s client review.

Draft summary

Three themes came up: pricing questions, a new competitor in two deals, and demand for a Dutch-language demo. I have drafted the summary for your review.

Principles and guardrails

Four rules for using AI in a B2B SaaS marketing team

01

Humans own the decision.

AI prepares options, drafts and analysis. Strategy, budget calls and anything with a client’s name on it stay with me.

02

Client data stays protected.

Client material only goes into tools that are fit for it, and sensitive details stay out of everything else.

03

Start with the workflow.

I map the job to be done first, then pick the tool. Most of the gains come from fixing the process.

04

Check before it ships.

Numbers, claims and sources get verified by a person before they reach a customer, a board or a website.

Tool stack

The AI tools I use as a fractional CMO, and when I pick each one

Core

Claude

My main workspace for strategy and writing, and the engine behind my recurring reporting. It is connected to the same tools as Grok Bot.

Second opinion

ChatGPT

A different angle on drafts, research and ideas, which makes it a useful check on my first answer.

Compare

Mistral

A European model I use and follow, so I can compare results across providers.

Research

Perplexity

Fast, sourced research on markets, competitors and buyers.

Meetings

Granola

Meeting transcripts, captured without a bot joining the call. Grok Bot picks up each meeting when it ends.

Execution

Notion, ClickUp and Obsidian

Where tasks, meeting notes and context live. Notion automations check each meeting for new and existing tasks, and Obsidian keeps local context so not everything sits in online storage.

Admin

Grok Bot

A team of bots for my admin: meeting details into Notion, a daily email check for tasks, calendar blockers, agendas and inbox. Nothing goes out without my approval.

In the work

How AI shows up in B2B SaaS client work

Research

Research and positioning

When a client needs a market view quickly, I run competitor and buyer research in Perplexity and Claude, then check it against what buyers say on sales calls.

GEO

GEO and website copy

GEO (Generative Engine Optimization) means writing pages that AI search tools can summarize and cite. Every page gets a nine-point GEO review before it goes live: category, audience, use cases, terms, answers, limits, comparisons, FAQ and structure.

Sales

Call and sales intelligence

Patterns across sales calls become competitor, objection and stack insights that feed messaging and pipeline decisions.

Reporting

Planning and reporting

Recurring status reports and task plans are drafted by Claude from the actual work across my tools, so my time goes to decisions.

From meeting to report

My Personal OS: Grok Bot, Notion and Claude turn meetings, email and calendars into tasks and reports

My Personal OS connects Grok Bot, Notion and Claude, with Obsidian for local context. Grok Bot and Notion automations handle capture and follow-up, and Claude runs the recurring reporting. This is how a client meeting flows through it.

Step 1

Meeting

Granola transcribes the meeting. When it ends, Grok Bot extracts the details and adds them to Notion.

Step 2

Tasks

A Notion automation checks the meeting for tasks and activities, and tells existing tasks apart from new ones.

Step 3

Inbox and calendar

Grok Bot checks my email daily for tasks and activities, and creates blockers or placeholders in my calendars where needed.

Step 4

Report

Claude runs my recurring reporting across the same tools. Obsidian keeps local context, so not everything sits in online storage.

My bot team

A team of bots in Grok Bot handles my admin, and none acts without my approval

In Grok Bot I run a team of bots across Dimastrata and my client work. Each one prepares, summarizes and flags. None of them sends, RSVPs, declines or publishes anything until I say so.

Meetings

Transcript Keeper

Copies each finished Granola meeting transcript into my Notion meeting database, so nothing lives only in a calendar invite or my memory.

How it runs

Checks about once an hour

Guardrail

It copies. It never sends or publishes anything.

Transcript Keeper

Extracts the details of each finished Granola meeting and adds them to my Notion meeting database, so nothing lives only in a calendar invite or my memory.

Calendar Sync

Looks across my Dimastrata and client calendars for double bookings, flags every clash and creates blockers or placeholders where needed.

Daily Agenda

A rundown of the day’s meetings plus any invites I have not answered yet, ready before the workday starts.

Inbox Digest

Checks my Dimastrata inbox daily for tasks and activities, sums up what needs attention first, and pulls the takeaways from the GTM newsletters I follow.

Chief of Staff

Keeps an eye on the other bots and tracks open decisions, then briefs me so nothing quietly stalls.

Client CoS

A chief of staff for Dimastrata and one for each main client, so every part of my work has someone watching the details.

FAQ

Questions about how I use AI

Which AI tools does a fractional CMO use?

My stack is Claude, ChatGPT, Mistral, Perplexity, Granola, Notion, ClickUp and Obsidian, plus Grok Bot for a small team of admin bots. Grok Bot captures meetings, email and calendars, and Claude runs the recurring reporting on the same tools.

Who approves what AI produces?

A person does. AI drafts and I decide. Nothing is sent, published, accepted or declined until I say so.

Is client data safe in AI tools?

Client data stays protected. I keep sensitive client data out of any tool that is not approved for it, and I check every output before it ships.

What is GEO and how is it different from SEO?

GEO (Generative Engine Optimization) means writing pages that AI search tools can summarize and cite. AEO (Answer Engine Optimization) means giving direct answers to the questions buyers ask. SEO aims at ranking in classic search results, while GEO and AEO aim at being quoted in AI answers.

What do you not use AI for?

I do not use AI for strategy, budget or pricing decisions, for sensitive client data in unapproved tools, or for anything sent or published in a client’s name without my review.

How does AI help with sales calls?

I use AI to analyze sales calls for buyer language, objections and competitor mentions. The findings feed research, positioning and GEO copy, and I check them against the calls themselves.

Who is this for?

B2B SaaS marketing teams and founders who want AI to save time while a person owns every decision.

Work with me

Want AI to pay back in your B2B SaaS marketing team?

I help B2B SaaS teams use AI where it earns its place, with clear guardrails and a human owning every decision.

Dimastrata

© Dimastrata 2026