Creators Column

The 9 AI Automations that Power Forward Future

Take a behind-the-scenes tour of how media business Forward Future uses AI workflows, including automations for every step of the content journey — from ideation to data analytics to repurposing.

Did you know HubSpot also has free marketing, sales, service, and CRM software?
Get Started Free

What does it take to run a successful media business and a YouTube channel with over half a million followers and a small team? A little gumption and a lot of trial and error. Oh, and AI. 

Here’s my behind-the-scenes tour of how Forward Future uses AI workflows. That includes automations for every step of the content journey, from ideation to improvement to repurposing. Follow along to see how they work, which tools we use, and how to build your own.

 

Screenshot 2025-12-03 at 10.58.27 AM

 

I’m Matthew Berman, the CEO and owner of Forward Future. I help businesses grow with AI by sharing the latest news, education, and practical tools. My goal is to make AI accessible, with practical advice. 

What is Forward Future?

The 9 AI Automations Forward Future Uses to Run Its Media Business

We’ve been working hard to scale Forward Future since its launch in 2023. Like all growing companies, the day-to-day admin began to detract from our core function: delivering high-quality, timely content. 

We were manually shuttling text between Google Docs and our CMS. We spent hours digging through recordings to find a single action item or memorable clip. Spreadsheets felt like they were tea leaves that we needed to decipher to find meaning. 

To solve this, we spent the past year building a comprehensive AI automation stack. But here's the thing, we didn't build it because we're automation zealots, or because we’re in the AI business. We built them to solve for our biggest pain point: limited time. 

The workflows we’re about to share are time savers. They’re strategic systems, designed to protect human capacity. They handle the repetitive drudgery of running a business and managing content so our homo sapiens brains stay fresh for work requiring judgment, intuition, and creativity. 

Let’s dive in.

One

The Research Assistant: Taming Creative Chaos

Commerce_Illustrations_Color_LightBG

Every writer knows that creative block has two sides to the coin. First, there's blank page paralysis, when you have just a seed of an idea but no entry point to start writing. 

Then, there's the manic draft, where inspiration strikes at 11 p.m., and you vomit 2,000 words onto the page. It's passionate, well-intentioned, and a structural disaster riddled with unchecked facts.

We built two Research Assistant workflows to solve both.

The Cold Start Generator is for when I need substance to get started. Here’s how it works.

  1. Trigger: I submit a topic through a simple Google Form, for example, "The future of VR in healthcare." 
  2. Research: Make.com connects to the Perplexity API, asking for a comprehensive research sweep. The output includes recent statistics, story angles, contrarian viewpoints, and key opinion leaders. 
  3. Filtering: A Python script filters out low-quality domains (those content farms that plague every Google search) from Perplexity and maps the survivors into a research packet.
  4. Generation: GPT-5 reads the packet and generates a structured brief. Here's where it gets clever. If the AI detects a gap, like if there’s no primary source found for a claim, it triggers a second, targeted search to fill that hole.
  5. Output: Make.com sends the standardized brief to Notion, pings me in Slack, and logs the metadata to our research journal for future reference.

Here, I have a full brief to review with the top perspectives and research summarized in one place. Having something to react to is much easier than starting with nothing, and I avoid research rabbit holes.

Tools used:

Google Forms, Make.com, Perplexity API, ChatGPT-5, Notion, Slack.

The Quick and Dirty Validator is for the manic drafts. I often write as fast as possible, focusing purely on voice and flow. As I’m writing, if I need a source or fact-check, I type [CHECK STAT] (or add a “Verify” comment in Google Sheets) and keep writing.

  1. Trigger: I add my raw, messy draft to a specific Google Drive folder, and Make.com feeds it to GPT-5 with browsing enabled. 
  2. Research: GPT-5 uses a prompt that instructs the AI to fact-check my content, but not to rewrite any of my prose. It isolates specific claims, names, and statistics, then queries Perplexity Pro to fact-check them against primary sources.
  3. Grading: Make.com compares the returned sources against my draft. For instance, if I claimed a market grew 200% when it actually grew 12%, it would flag it for me.
  4. Output: The automation appends a validation report to the draft document in Google Drive with a confidence score, direct links to sources, and suggested corrections for any inaccuracies. 

When I sit down to self-edit for version two, I can focus on structure. I can fix any problematic or unsupported claims flagged in the report.

Tools used:

Google Drive, Make.com, Perplexity Pro, ChatGPT-5. 

Two

The Media-to-Content Pipeline: Mining Silicon From Recordings

The written word is my backbone at Forward Future, but our best insights often surface in conversation. Spontaneous quotes arise from live shows with AI insiders, podcast interviews, partner calls, or even internal team calls. 

For months, those insights died the moment the Zoom call ended. Watching a sixty-minute recording gather digital dust felt criminal, but manual transcription took time I didn’t have.

The Media-to-Content Pipeline

I decided to treat every recording as a "campaign in a box" with potential themes, angles, and quotes. To make the most of these hidden insights, I built the Media-to-Content Pipeline automation to capture them. Here’s how it works:
  • Input: The workflow monitors specific folders in Google Drive (for Zoom recordings) and channels in YouTube or Riverside.
  • Trigger: When a new file is added, Make.com sends it to Whisper or AssemblyAI for transcription.
  • Transcription: Standard auto-captions often butcher technical jargon, so I set up dedicated APIs with a custom vocabulary glossary to minimize errors.
  • Analysis: GPT-5 receives the transcript with a strict context window prompt. It identifies the core theme, extracts the three most controversial or insightful quotes, and generates a summary.
  • Generation: Then, GPT-5 creates a suite of assets. That includes a 500-word blog post, a bulleted LinkedIn post, and SRT caption files for potentially viral clips — all in our voice and style.
  • Output: The workflow routes the assets to a Notion database called "Content Review," tagged by speaker and topic.

Every time I record, I have an accurate transcription, summary of key points, and fully-fledged drafts of content assets. I just need to review and distribute. 

Whisper or AssemblyAI for transcription, Google Drive, Notion, Make.com.

Alternatives: Otter.ai, Descript, or Deepgram for transcription, Castmagic or Oasis for all-in-one solutions.

Tools to Use

Three

The Content Upcycler: Using Every Part of the Digital Buffalo

I believe in using “every part of the buffalo.” Great ideas can be repurposed time and time again. So, I built a Content Upcycler to handle repurposing into different formats.


1. Trigger: When someone changes a piece's status to "Repurpose" in our Notion content calendar, the text passes to a Claude project. (We prefer Claude for rewriting, as it sounds less robotic than ChatGPT.) 

2. Generation: Claude writes drafts of social posts, newsletter articles, and any other format we request based on the original content. 

3. Formatting: My default Claude prompt contains specific character limits and formatting rules for each platform. For example, the prompt knows that LinkedIn posts need line breaks for readability. Meanwhile, newsletter summaries need narrative arcs.

4. Routing: Once ready, Make.com's router logic determines where to send the rewritten content. Newsletters go to my email drafts folder, X threads to the social scheduler, and quote graphics into the visual generator workflow (below).

5. Output: I receive one summary email with links to various drafts, ready for final human polish. 

Warning: Don’t miss this final step. Human intuition beats AI every day of the week. Even repurposed content needs your touch. Be sure to review and tweak all AI output.

Notion, Claude, Make.com.

Alternative tools: Repurpose.io or Munch for repurposing, Jasper or Copy.ai for writing.

Tools to Use

Four

The Visual Content Generator: Escaping Costly Design Cycles

Paying a designer for every single thumbnail hurts your bank account, and searching for stock photos can be soul-crushing. So, what should you do? Use AI, of course. 

We built two pipelines for images at Forward Future: a Fast Lane for speed, a High-Fidelity Lane for brand prestige.

  • Pipeline A: The Fast Lane (DALL-E 3)

    The Fast Lane approach works best for thumbnails, social crops, or internal decks.
    • Trigger: When we add a new post, Make.com passes the title to ChatGPT.
    • Prompting: ChatGPT writes a detailed image prompt describing a "minimalist, vector-style illustration.”
    • Generation: Make.com then sends the prompt to DALL-E 3 for image creation.
    • Output: The result auto-saves to our CMS and runs through a resizing module to generate 16:9 and 1:1 versions instantly.

  • Pipeline B: The High-Fidelity Lane (Leonardo.ai)

    Use the High-Fidelity Lane for feature headers, website assets, and guides.
    • Trigger: When we tag a post "High Fidelity," Make.com sends the concept to Leonardo.ai.
    • Style match: Leonardo.ai uses its style reference feature to match the image style to our brand’s color palette and reference hero images.
    • Generation: Leonardo.ai returns four options that look like they came from our in-house artist and don't scream "generic AI slop.”
    • Output: The result auto-saves to our CMS.

Make.com, ChatGPT, DALL-E 3 or Leonardo.ai for image generation, CMS.

Alternatives: Midjourney via Discord API, Adobe Firefly, Stability AI; Canva Bulk Create or Bannerbear for design automation. 

Tools to Use

Five

The Auto-Content QA: An Editorial Safety Net That Never Sleeps

Scaling production can mean slipping standards. Humans get tired. We’re a small team shipping a lot of content, and we go blind to our own typos. We sometimes use freelancers who don’t live and breathe our brand guidelines like we do. 

I wanted an editorial safety net that doesn't sleep, so I created the Auto-Content QA.

Service Agent_Illustrations_Color_LightBG
  1. Training: I trained a custom GPT on our 20-page brand style guide. It knows our voice (authoritative but casual), our formatting rules (Oxford commas only), and our forbidden words (never "synergy" or “revolutionary” or “game-changing” or “breakthrough”).
  2. Trigger: When I change a content’s status to "Ready for Review" in Notion, Make.com passes the draft to the custom GPT. 
  3. Synthesis: The model reviews the text not just for grammar, but for logic flow, tone, and consistency.
  4. Critique: Instead of silently fixing issues, it acts like a strict editor, inserting comments flagging weak headlines, passive voice, or fluff sentences that don't add value.
  5. Output: The workflow links the clean and dirty reports on the Asana project card.

Tools: Custom GPT, Notion, Asana.

Alternative tools: Grammarly API or ProWritingAid, Writer.com, or Acrolinx for enterprise.

Tools to Use

Six

The Performance Analyzer: Data That Tells Stories

I used to spend Monday mornings doing spreadsheet gymnastics, exporting CSVs from Google Analytics, YouTube Studio, and LinkedIn, then Pivot-tabling them into submission. It was torture. And worse, it was sterile data. Numbers don’t help if you can’t explain what matters and why it’s happening. Here’s how we do it now.
  • Trigger: Every Monday at 8:00 a.m., Make.com pulls the last seven days of data from all our connected APIs: Google Analytics, YouTube Studio, LinkedIn, and HubSpot (for lead generation metrics).
  • Cleaning data: Make.com standardizes the metrics in Airtable, converting views and impressions into a single reach metric for comparison.
  • Synthesis: This dataset feeds into Claude 3 Opus, which we chose for its superior reasoning on large contexts. The prompt asks specific strategic questions. Which topics outperformed the baseline? What's the week-over-week trend? Based on this report, what should we write or record next week?
  • Output: By 9:00 a.m., a color-coded HTML email arrives in my inbox with the report.

The weekly report reads like a memo from a strategist: "Engagement on labor market topics is down 10%; pivot to AI bubble content next week." 

Make.com, Claude 3 Opus, Airtable. 

Alternative tools: Looker Studio or Rows.com for data viz; Julius AI or ChatGPT Data Analyst.

Tools to Use

Seven

The Meeting and Task Syncer: Killing the Phantom Task

Strategy matters, but execution is where battles are won. Our biggest leak was the phantom task, when everyone would agree to something on a Zoom call, but no one would take ownership or write it down. Weeks later, it hadn’t happened. 

We built a bridge between our mouths and our project management tools, called the Meeting and Task Syncer.

1. Trigger: I record a client or strategy call using Fathom.

2. Analysis: GPT-5 parses the raw transcript, trained to spot key phrases like "I will," "Let's make sure to," or "Can you handle..." It extracts a high-level summary, key action items, and the responsible parties.

3. Assigning tasks: The workflow creates tasks in Asana for every stated action item with due dates, defaulting to 48 hours if undefined. The logic matches speakers to team members. If I say, "I'll handle the draft," it assigns the task to me.

4. Output: GPT-5 saves high-level notes in the Notion project page, and every phantom task has an assignment in Asana.

The result? Zero missed follow-ups. Every decision gets documented before we stand up from our desks.

Fathom, GPT-5, Notion, Asana.

Alternative tools: Fireflies, Otter.ai, Supernormal, or Grain for meeting AI; Zapier or Slack Workflow Builder for integration; ClickUp or Monday.com for project management; HubSpot's meeting scheduler and CRM integration to log action items and follow-ups.

Tools to Use

Growth_Illustrations_Color_LightBG-2

Eight

The Workflow Refiner: Building Self-Sharpening Systems

Outputs from AI workflows are only as good as their inputs and prompts. If something is broken, you need to know as quickly as possible. This workflow is the most critical piece of the entire stack. 

In the early days, we built automations that broke within a month because the prompts went stale, the models changed, or the output quality dropped. AI workflows must be able to adapt to feedback, so we fed our meetings and support tickets into an improvement loop.

  • Input: An AI agent scans transcripts from our Ops Review meetings and Zendesk tickets for negative sentiment regarding our internal content and tools. 

    Trigger: When an issue is identified (e.g., “the research briefs are missing citations” or “the generated images are too dark”).

    Generation: The AI agent drafts concrete fixes, rewrites the system prompts, and updates our Help Center checklists.

  • Approvals: Proposed changes land in an Improvements database. A human clicks "Approve” to greenlight them.

    Output: The automation updates itself by pushing the new prompt to the live workflow.

Zendesk and Fathom.

Alternative tools: Intercom, HubSpot, or Freshdesk for support; Notion or Guru for knowledge bases.

Tools to Use

Nine

The Lead Magnet Builder: Creating Bonus Content

Sometimes we realize we've written ten blog posts or recorded a short mini-series on a single topic. Essentially, we've written an ebook without realizing it. Rather than letting that content drift into the archives, we make the most of our intellectual property.

Input: The system scans our Airtable content library for tags. 

Trigger: When we reach five or more articles on a topic, GPT-5 compiles the articles into a longer lead magnet. 

Reformatting: According to the prompt, GPT-5 removes repetitive intros and restructures the flow to read like chapters.

Generation: GPT-5 generates a catchy title, table of contents, and an introduction and conclusion. It drafts landing page copy and email delivery sequences.

Output: A ready-to-design manuscript gets routed to our team, and we can launch a new lead magnet in 24 hours with almost zero writing effort.

ChatGPT and Gemini.

Alternative tools: Webflow or WordPress for CMS; Beehiiv, ConvertKit, or HubSpot for newsletters and lead nurturing.

Tools to Use

The Truth About Building Automations

Looking at this list, it's easy to see a well-oiled machine. But honestly? What you're actually seeing is a lot of late nights, broken processes, and AI trial-and-error that made absolutely no sense the first time.

This suite of automations didn't arrive in a neat package. We took the best use cases from different AI tools (Claude for writing, ChatGPT for analysis, and Perplexity for research) and found a way to plug them into our existing Notion and Asana workflows. It was built slowly, usually on Friday afternoons, born out of frustration with mundane tasks that were eating our creativity alive.

My top takeaway? Don’t try to automate everything at once. If you implement all eight workflows tomorrow, you won't have an automation strategy. Instead, you'll have a new job fixing errors. Make automation a habit, something you do or improve every week.

Start with the one thing in your week that makes you want to throw your laptop out the window. Maybe it's resizing images. Maybe it's summarizing meeting notes. Maybe it's staring at a blank Google Doc waiting for inspiration. Pick that one friction point, build a workflow to crush it, and reclaim those precious hours.

The goal isn't to let robots take over. The goal is to automate the robot out of yourself, so you can get back to doing work that only a human can do.

Customer Journey_Illustrations_Color_LightBG

The Real Value of Automating Your Workflows

AI automations aren’t about replacing people. They’re about protecting their time and preventing burnout. Well-designed workflows turn endless tasks into dependable systems, giving teams back the hours they need for strategy, creativity, and connection. 

Whether you’re a solopreneur, a small business like Forward Future, or simply a scrappy team within a large one, you can multiply your impact with smart automations.

_cta_contentblock_headshots_headshot_1

Turn These Creative Ideas Into Reality with Breeze

Start implementing these AI use cases today with Breeze — HubSpot's complete AI solution that understands your business context and automates workflows from marketing to sales to service, all in one integrated platform.