How I Built My Marketing Stack with AI as a Team of One
I’m the only non-engineering hire at a seven-person seed-stage company. The others are six (incredible!) engineers.
And as wonderful as they are, they’ve got more important things to do than build me pipelines or make changes to marketing language on our website and documentation.
More than that, our small scale means that I’m working with a limited budget for marketing point tools (and no budget for additional hires), meaning I can’t simply hire out to products or people to fill in the gaps that I can’t reasonably do as one person.
Or at least, I couldn’t until now. I’m lucky enough to be working on the exact product that makes this feasible for me as a team of one.
Product marketing as a whole has transformed significantly in just a few short years with the advent of AI. Here’s just a few ways I’ve used it to form my own marketing stack and 100x the work I do.
1. Turning Every GitHub Release Into a Coordinated Marketing Event
Our team is constantly shipping (almost every other day). Some of what’s released are small and don’t mean much to our audiences. Some of them mean a lot. The problem is that without a system, every release gets the same level of attention: either we do something with it or it disappears into the GitHub changelog.
So I built a release triage workflow. Whenever a new GitHub release goes out, Friday reads the release notes, categorizes it, and decides what to do. This is pulled automatically from GitHub whenever a release happens (not manually triggered by me), and it breaks down against four tiers:
Tier 1 (major product or feature update): update the website and documentation, generate a draft blog post and associated social assets, and create a Notion doc with messaging, positioning angle, and asset list for launch
Tier 2 (meaningful improvement, visible to users): update the documentation, flag the change in a changelog post, draft short posts across all of our social channels
Tier 3 (bug fixes, performance): update the docs and changelog if relevant to end users but no additional marketing action
Tier 4 (internal, infrastructure): log it, do nothing
Friday doesn’t publish anything without me reviewing it. But it handles the triage, writes the first drafts, and queues everything in Notion so I can process a week of releases in 30 minutes instead of figuring out each one from scratch. And anything Friday writes for me, including socials, blogs, and website, all run through a series of skills to make sure its aligned with our content strategy: Friday voice and tone, SEO optimization, ICP and positioning.
It’s not quite there yet, but the next iteration as we scale up the company includes work to scan existing sales materials and flag whether a release warrants an update to battle cards, one-pagers, and team training.
2. Campaign Tracking Without a Dedicated Ops Person
Campaign tracking is the kind of work that either has a dedicated person or falls apart. By trade, I’d say I’m not the strongest GTM engineer, so I’ve relied heavily on Friday to support me in this process.
The workflow I built mirrors a stripped-down version of how a real GTM team tracks programs. I maintain a campaign brief in Notion. When a campaign launches, Friday reads the brief, creates the relevant tracking entries, and monitors progress across our channels: UTM performance from our analytics, social engagement, any inbound form fills or signups that match the campaign window.
At the end of each week, Friday pulls the numbers and writes a one-page summary: what the goal was, what the numbers say, what looks like it’s working versus what isn’t, and a few recommendations. The summary lands in my email inbox. I read it, add context, and decide whether to adjust.
Obviously, this isn’t sophisticated attribution modeling, but for a pre-PMF company with very limited traffic, it doesn’t need to be. I just need to know whether the thing I shipped this week got any traction and where the signal is strongest, so I can adjust the things that don’t work and repeat the things that do. And, I need that answer in under 10 minutes.
3. Competitive Intelligence That Runs Without Me
I come from a background where competitive intelligence meant maintaining a spreadsheet that was always two months out of date.
Now Friday monitors a list of competitors on a daily schedule: their websites for pricing or positioning changes, G2 and Capterra for new reviews, job postings for signals about where they’re investing, LinkedIn for announcements, and social posts from the community. When something changes, it updates our competitive tracking spreadsheet in Google Sheets and writes a short summary of what changed and why it might matter.
But working in AI, the competitive landscape is constantly changing. The hot topic 1 week ago is old news the next. So more than just keeping an eye on existing competitors, what we find most value in is finding new entrants. Not just random entrants, but the ones the community actually care about.
So we also have it scan across AI communities and popular thought leaders on Reddit, X, GitHub, and Hackernews to scan for new entrants. The agentic workflow space is moving fast, and the definition of who counts as a competitor is expanding. Friday searches across product directories, funding announcements, and relevant communities, then scores new entrants against a simple rubric: how much they overlap with our positioning, how funded they are, and how much traction they appear to have. Anything that scores above a threshold gets flagged with a recommendation: monitor, investigate, or respond.
4. Positioning Research From the Conversations Already Happening
Positioning is the part of my job I find hardest to do well in a pre-PMF stage. You have hypotheses about who your best customer is and what they care about, but not enough conversations, data, or time to validate them the traditional way. The workflow I built treats positioning development as a loop rather than a one-time exercise.
Step 1: Consolidate market signal Friday scans Reddit and X for conversations about AI agents, agentic workflows, and workflow automation. It organizes findings by ICP profile: what does this problem look like for the developer, the AI tinkerer, and the non-technical AI enthusiast? The output is a synthesis brief I can read in 10 minutes. Reading those briefs regularly builds intuition about which audience segments have the sharpest pain and which framings keep resonating.
Step 2: Draft positioning candidates I take that intuition and work with Friday to draft specific positioning statements that help judge who the product is for, what problem it solves, and why it’s different. It’ll build one set for each candidate.
Step 3: Score each candidate For each candidate, Friday scores it on three dimensions:
Specific: vague positioning fails on contact
Pain/JTBD fit: does it address something people are actually trying to do
Differentiated: is anyone else already saying this
And each score comes with a plain-English note on what’s pulling it down.
Step 4: Iterate Depending on how it goes, I’ll define a candidate, then score it again. Try a different angle for a different ICP, compare scores. The loop runs as many times as I want without scheduling a user interview or waiting for enough traffic to run a test.
By the time I do talk to users, I’m not starting from scratch. I have candidates with known scores and clear weaknesses, and I know which ones to pressure-test first. The conversations become more useful because I’m walking in with a hypothesis rather than an open question.
5. A User Research Knowledge Base That Compounds Over Time
The problem with user interviews isn’t doing them. It’s that the insight from each one mostly lives in a transcript that nobody reads again.
I have a folder on my computer where interview transcripts and call recordings land. Friday watches that folder and when a new transcript appears, it reads the full conversation and extracts: the pain points the person described, the exact words and phrases they used, what they’re currently doing instead, what would need to be true for them to switch, and any signal about their context (team size, role, how technical they are). Everything goes into a persistent knowledge base via memory in Friday, tagged by theme and audience type.
That alone would be useful. The part that makes it compound: every week, Friday runs a synthesis pass across all entries and surfaces patterns. Which pain points come up across three or more conversations? Which phrases keep appearing verbatim that we’re not using in our own copy? Which audience types are showing up most, and what do they care about that the others don’t?
It then compares those patterns against our current positioning and flags gaps: pain points we haven’t addressed, language we’re not using, audience segments generating signal that we haven’t built a narrative for yet. The output is a short brief, not a data dump. Something I can read in five minutes and act on.
What’s most powerful about this is that the knowledge base also becomes queryable. I can ask “what have mid-market users said about their best use cases in Friday” or “how do individual prosumers discover our product” and get an answer grounded in actual conversations. Every new interview makes the next positioning decision better informed.
A Note on Building This Without an Engineering Background
I’m not an engineer. My background is product marketing and product management. I learned to use Friday Studio the same way I learned every other tool: by describing what I wanted and iterating when it wasn’t right.
None of the workflows above required me to write code or configure a server. Each one started as a description of the problem and a rough idea of what the output should look like. Friday asked clarifying questions, connected to the tools I was already using (GitHub, Notion, Slack, our analytics), and built the automation. The first version of each workflow took less than 30 minutes to set up. The harder part was figuring out what output I actually wanted, which is a thinking problem, not a technical one.
And the best part of this is that none of these workflows are set in stone. Being this early in the building process means that as soon as I set a process for myself, its out of date. Friday lets me iterate against every automation so I can refine and rebuild each flow without having to build from scratch, via conversation. It’ll take my recommendations, update the configuration, and be just as reliable as it was the first time I tried it.
That’s the part I think gets undersold. The constraint for solo marketers at early-stage companies isn’t access to tools. It’s time to think clearly about what you want to automate and what the right output looks like. Friday handles the wiring and you supply the judgment.
Download Friday Studio free at hellofriday.ai or browse the source on GitHub.