At the end of July, I attended the Marketing AI Conference 2023 (MAICON) in Cleveland.
I expected to learn more about marketing AI tools, streamlined marketing workflows and how agencies are using AI tools.
The conference delivered on all of my expectations.
Here are my notes from the event…
MAICON was attended by 700 marketers representing brands and agencies of all sizes. Most of the attendees I met were just starting to experiment with AI (translation: they are playing with chatGPT) and wondering how to make the most of the AI tech that is coming.
Most people at the conference were just starting to use AI tools (mostly chatGPT). The attendees who were more experienced were using AI for prediction, mass personalization and automating marketing operations.
My main takeaway was: humans will do the thinking and machines will do the doing.
A few MACON speakers presented their frameworks for identifying tasks that should be done by AI to free up human marketers to do more creative work. To spend more time on the creative tasks, we must be able to separate them from routine marketing tasks. That’s not always easy.
For example, I recently spent many hours writing a 45-page ebook on the topic of SEO for Revenue. I spent my time revising the text, building a landing page and coordinating with my designer for the visuals. The deep work that required my human insight was just the time I spent fleshing out the topic in conversations with my team and refining the SEO concepts in the book.
The rest of my time was spent on routine tasks that could be automated by an AI driven workflow, including formatting the text, researching examples, and writing my ideas in coherent sentences. At Fire&Spark, we’re building streamlined marketing workflows in Make.com, Zapier and AirOps.
The AI advantage: If you are already using AI tools, press your advantage! The marketing world is just beginning to understand AI’s capabilities. You have a huge advantage right now if you adopt AI enabled workflows.
At Fire&Spark, we’re experimenting with AI for mass personalization, automated cold outreach, content generation, automated email follow-up, metadata prediction for SEO, video editing and automating marketing ops.
High throughput marketing: Marketing teams will soon be organized around the concept of high throughput marketing–which allows marketers to test more ideas, iterate faster, and amplify their best ideas.
More than a decade ago, I started using AdWords. For the first time, I had access to a platform that allowed me to efficiently test messaging (ad headlines) on a targeted audience. I was doing marketing for a software start-up at the time. AdWords dramatically reduced the time it took to test campaign messaging and value propositions.
Today, AI increases the rate at which marketers can build campaigns of all types. Testing ideas for content marketing, digital PR, cold outreach, and even organizing in-person events will become significantly easier. Marketers will build more stuff and prove out more great ideas.
Will we need less marketers? No. High throughput marketing will require human marketers to come up with more ideas worth testing to fuel AI-powered workflows doing the implementation.
AI-powered marketing teams use loosely connected AI-enabled marketing services orchestrated by a no-code automation tool like Zapier, Make.com, AirOps or Google Sheets. Every organization will have different marketing workflows for content, social, ad ops, etc.
Inexpensive prediction: I’ve thought of AI as inexpensive prediction since reading the book Prediction Machines. Marketers will use AI to predict which subject line will get the highest open rate, which topics will generate the most engagement, or who on your email list will purchase an offer based on intent signals. Current AI tech enables inexpensive prediction that can be trained on relatively small datasets. I got a demo of one AI tool that makes building prediction models easy.
The Akkio platform includes a no-code tool that uses autoML to spin up prediction models, from spreadsheets. This is huge for marketers who lack a coding or machine learning background. Any marketing problem that requires prediction can be addressed with an AI prediction model if you frame the problem properly and supply appropriate data. For example, predicting the open rate on email subject lines or predicting the best SEO keyword to target or predicting the click-through on ad creative.
Akkio allow marketers to easily build prediction models from spreadsheets. You give it a spreadsheet and it spins up a predictive model with autoML that you access via an API. At Fire&Spark, we are personalizing emails with GPT4 but we aren’t doing anything with the performance data. Soon, we’ll build a prediction model using open rate and click-through data to predict the best subject line and copy for each individual recipient. Then, we’ll try building a model to predict which keywords to target with SEO content based on out client’s ability to rank for each keyword.
For Agencies: Prediction models could be part of an agency’s secret sauce. If you build a prediction model around proprietary data, your model could be hard for your competitors to replicate. Agencies specializing in a specific industry (e.g. healthcare or education) have the opportunity to create prediction models tailored to their industry. This can be based on the data they gather while working with their clients.
AI Automation Specialist role: Someone at your organization needs to lead AI adoption. This person would research AI capabilities, educate the larger team and identify areas for process improvement. At MAICON, I met one agency who recently hired someone with “AI Automation” in their job title. I expect many more to follow. At Fire&Spark, we are experimenting with a dedicated AI automation specialist role within the agency. So far having someone with a dedicated focus on AI has allowed us to experiment at a much faster rate. Expect to receive LinkedIn messages from fractional CAIO services soon.
Revisit old marketing initiatives: A few years ago, my team at Fire&Spark produced a video interview series with marketers from purpose-driven brands. The campaign was successful at getting the word out about our Purpose-driven SEO methodology, but it took too much time to organize, shoot and edit. Today using marketing AI tools, we could cut the time required to shoot an interview in half. AI would dramatically reduce the time it takes to research guests, organize the live shoot and edit the final video. We will consider restarting our interview campaign now that the economics have changed.
Replace segmentation with personalization: At MAICON, I saw evidence that mass personalization can dramatically improve conversion rates. I wondered if Fire&Spark has enough first party data to personalize our email marketing. We periodically send an email newsletter, event announcements and thought leadership article emails. I ran an experiment using pref.ai at the conference. Pref.ai is a “preferences database” that you can query for “observations” about an email recipient that an LLM could use to write a personalized email. It worked well when I trained it on emails I had received form each recipient.
Adapting to AI: From my experience teaching marketing AI workshops… When marketers first start exploring AI tools, they inevitably begin to see their marketing workflows in a new light. They begin to distinguish between the creative tasks that require human inspiration and the repetitive tasks that AI can automate.
When they experience this lightbulb moment, they realize they can spend more time of their workday on creative work.
Spending time on many tasks starts to feel wasteful:
- Watching a webinar recording to find the most insightful shorts. Waste of time!
- Reading 3 months of email correspondence to craft the perfect follow-up email. I’ve got better things to do!
- Transforming your conference notes into a blog article. AI does a better job!
What will marketers do with the time saved?
Get better at coming up with creative ideas to fuel their AI workflows. Iterate faster. Test more messaging. Personalize content. Train prediction models. Help your team and clients understand the capabilities of marketing AI tools.
I spend my time generating leads and selling client projects for our agency. My best guess is that I spend 50% of my time researching decisions, making decisions, brainstorming for content creation, communicating with my team, nurturing partner relationships, etc. These are tasks that require deep human work. The other half of my time is spent coordinating with my team, formatting documents, drafting emails, attending meetings, reviewing documents, and more.
Those are my impressions from MAICON. At Fire&Spark we’re hoping to free up our team to do more creative work. AI tools should give them more time to think deeply, generate ideas and try more things.
Some of those things will be home runs. I’m really looking forward to unearthing those home run ideas.
