The Autonomous Marketer: How to Create a Custom AI Assistant to Create On-Brand Social Content Ideas (Without Ever Leaving Slack)

Managing and producing content for multiple clients as an agency marketer can be a tall task. Marketing pros spend their days weaving and bobbing between different project management platforms, digital workspaces, and tools to complete their work creatively and efficiently.

But what if you could use AI to build a process that integrates directly with all your most-used tools and serves as a thought partner to help brainstorm content ideas?

That’s exactly what founder and creative director Laryssa Wirstiuk set out to build when she and her team of five at Joy Joya, an email + SMS marketing agency for women-focused ecommerce brands, were looking for a way to streamline their content strategy workflow and come up with ideas faster for their roster of clients.

The end result? Wirstiuk built an autonomous AI assistant that integrates directly with Slack and Notion through Make to fulfill the tedious task of brainstorming new social media content ideas, blending AI with human creativity. Oh, and as a massive bonus, they save between five and 10 hours a week.

With so many clients and multiple client briefs, strategy documents, brand guidelines, and editorial calendars, Wirstiuk wondered if AI might be able to assist her and the team streamline and improve their content process and pipeline.

Wirstiuk didn’t know exactly what this would look like in practice; her curiosity led her to reach out to an Upwork developer in the summer of 2024 to see what might be possible.

“Upwork is my go-to for those kinds of more exploratory projects, when I'm just trying to see, oh, what could we do?” she shares. “Maybe I'm not super clear yet on what I want it to be, but from there I can tap an expert and get their perspective.”

The developer had built a similar engine for another marketing agency with a different use case, but was familiar with the work and was game to help Wirstiuk get this agent up and running using Make, an integration and automation development platform.

“The system works like a hybrid AI assistant that sits in the middle of our content planning workflow,” Wirstiuk says.

The AI assistant:

1. Identifies the client
2. Pulls their brand information and content strategy from Notion
3. Sends everything to AI to generate customized ideas
4. Places those ideas into a Notion database where the team reviews, edits, and approves them.

When ideas are approved, a second automation automatically transfers them into the client’s content calendar.

“The developer designed the entire workflow and architected the logic end-to-end, including the integration points, the AI prompt structure, and how Notion data should be pulled and formatted,” she explains. “They also implemented the initial version of the Make scenarios, especially the more complex parsing and data-mapping components.”

Now, Wirstiuk collaborates with the developer to maintain the system — adjusting prompts, updating logic, troubleshooting edge cases, and refining the workflow as the Joy Joya client roster grows.

“From the team’s perspective, it feels like placing an order with an AI assistant and receiving a batch of tailored, on-brand ideas in Notion without ever leaving Slack,” Wirstiuk explains.

Wirstiuk says the output includes social media content ideas for Instagram, Facebook, Pinterest, and occasionally TikTok with:

• Storytelling ideas
• Educational posts
• Product highlights
• Behind-the-scenes concepts
• Content aligned with upcoming events or promotions
• Seasonal themes
• Customer-focused posts
• Brand-building content
• Email themes
• Blog prompts
• SMS concepts

Here’s how the sequence works in practice:

1. A team member posts a message in Slack inside our dedicated channel, following a simple pattern:

 “Generate ideas for [Client], [Monday date].”

2. Make listens for that message using a Slack module that monitors new messages in that specific channel.

3. AI parses the Slack message to extract the client name and the correct start-of-week date. That ensures the automation can’t run with missing or malformed input.

4. Make retrieves client context from Notion — things like brand voice, content pillars, product categories, current promotions, and any events happening that week. This makes the AI output specific and strategic, not generic.

5. The compiled context is sent to an AI model — OpenAI GPT-4o — in a well-structured prompt instructing it to generate social media ideas tailored to that client and week. “It guides AI into the role of the person who might be doing this kind of job,” Wirstiuk says. “Then it explains the task in great detail as well as gives it some boundaries in terms of date/time and what to reference.”

6. A second AI step cleans the output, ensuring second-person voice, platform-agnostic language, and consistent formatting.

7. Make writes each idea into Notion in a database called Content Idea Review, with status = “In Review.”

8. Her team reviews them in Notion every other week, edits as needed, and approves favorites.

9. A second Make scenario monitors for ideas marked “Approved.”

 When that happens, Make:

• Creates entries in the client’s content calendar
• Marks the idea as “Transferred”
• Notifies the team through a status change

That’s the full loop — Slack → Make → AI → Notion/Knowledge Base Review → Content Calendar.

“This setup helps us save hours of administrative time each week while ensuring that every idea aligns with the client’s brand identity and strategic goals,” Wirstiuk explains.

Plus, Wirstiuk says the tool allows the team to put more time and attention on strategic, higher-level initiatives for clients, so they receive even more value for their investment.

“We're able to accelerate results faster and also test more,” she says.

Working with a developer who understands the backend of this technology has been key, according to Wirstiuk.

“It’s a pretty sophisticated scenario with a lot of conditional flow built in,” she says. “Make is really acting as the backbone coordinating all these moving parts.”

• Routers in Make allow the workflow to handle different clients, specific content calendar structures, or unique brand rules.
• Filters ensure the Slack message contains the required information before the scenario continues.
• Conditional blocks determine how much client context is passed into the AI based on what’s relevant for that week.
• Iterators are used when AI returns multiple idea objects that need to be processed individually.
• Break modules are built in to prevent Notion API throttling and ensure each idea transfers cleanly.

AI isn’t perfect, and Wirstiuk and her team don’t always accept the ideas the tool suggests. For example, they often reject ideas when AI mentions a platform a client doesn’t use, recycles older topics at the wrong moment, are overly generic, and when they don’t align with the week’s strategy.

This workflow has been invaluable to Wirstiuk and the Joy Joya team.

The system forces all ideas to connect back to the client’s strategy because the AI prompt pulls directly from their brand documentation.

“That alone has increased clarity and cohesion in the weekly content,” Wirstiuk says. “Clients have actually commented that the ideas feel more consistent and more aligned with the themes we’re focusing on.”

With this tool…

⭐ Joy Joya can serve more clients without increasing team burnout — the team currently works with about 10 clients that benefit from this tool.

⭐ Team members spend their time refining and elevating ideas, not generating them from scratch, saving about five to 10 hours per week, on average.

⭐ Consistency and quality have improved — and as noted above, it’s definitely being noticed by clients!

⭐ Calendars are more strategic because humans now have more time for planning rather than admin work.

Wirstiuk says since the AI always gives them a large pool of ideas, the team is consistently able to generate 3–4 strong posts per week without dry spells.

Here’s something important to remember: this impressive workflow is super cool, but it requires the creativity and discernment of the human brain to work at its full capacity.

Remember that this entire process starts with deep research and a true understanding of each client’s background, goals, and voice. Without this documented in detailed brand profiles created by humans in Notion, the AI agent has nothing to work with.

At the end of the day, AI provides the Joy Joya team with 10-15 ideas and humans approve 3-5 that best match the brand strategy — about a 33% success rate. Ideas are strategically placed in the content calendar with storytelling flow, and final edits always happen on Joy Joya’s end before the client sees anything.

“We treat AI as a brainstorming partner, not a decision-maker,” Wirstiuk shares. “The AI creates volume and variety, but our team applies the strategy and taste.”