How I would use AI RIGHT NOW if I was a CMO

Monday will be no-AI day!

I began this week by sharing a WSJ article on "how marketers are using AI," which left me unimpressed. So, I decided to end the week by delving into how I would utilize AI as a CMO with the existing tools on the market to enhance my team's effectiveness.

The landscape is changing rapidly! I outlined a plan on Wednesday, and before I finished writing this on Thursday morning, I already had a few new tools I wanted to add to the toolkit. I'm preparing a presentation on how to use AI for Warburg Pincus portfolio companies in May, and I expect this will only be a small subset of what I present.

Individual Contributor

Let's begin with the tools I'm currently using as an independent advisor. Some of these are quite new.

Descript

I use Descript for all my podcast editing. I upload my audio file, and Descript identifies the speakers and transcribes it. I can then edit the podcast audio and text simultaneously. If I edit the text, the audio is adjusted, and vice versa. I can even add text, and Descript will insert the words in the speaker's voice and intonation, matching the surrounding audio.

Rewind

Rewind records everything you do on your Mac: every website, every email you open, every call you are on (transcribing the audio). Everything is stored locally, and it has one search tool that allows you to look for anything. The basic product is free and includes 50 searches. For $30/month, you get unlimited searches and access to a ChatGPT that can look through your history, answer questions, and link to relevant content. It's a game-changer.

(Related: Yesterday I wrote, "Microsoft Copilot already summarizes Teams meetings, is Zoom far behind?". Later on Thursday morning, Zoom announced an integration of Fathom. Fathom is a FREE Zoom app that records and transcribes Zoom calls. It has AI integration to summarize meetings, provide action items, and search for relevant clips and text.)

ChatGPT

I have previously discussed how I use ChatGPT to edit my mini-essays. See Wednesday's post. I also use it for writing comedy. It's excellent for expanding on ideas or brainstorming, and GPT-4 is a significant improvement over GPT-3.5.

MidJourney

After experimenting with various AI image tools, I believe MidJourney currently creates the best images most consistently. The images I used this week for posts were all MidJourney created. Last week, they were mostly from Dall-e2. Most of the improvement was NOT due to my enhanced ability to write prompts (although that played a role, and I'll have a post on that eventually). If you have ANY need for images in your life, mastering text-to-image prompting is now a crucial skill.

How I would Direct my Team

The tools I mentioned above are ones I'm using right now. But the real power of AI comes from how it enables you to extract more from a team. These ideas are NOT tested (I don't have a team right now and may never have one again), but if I were stepping into a new CMO role, these are the steps I would take.

Content Creation

GPT has been good for content creation for the past two years. Good, but not great. With GPT-4, it's now great, and there's no excuse not to use it. I would start by assigning one content generator to be in charge of AI content generation. The job breaks into two parts: (1) Content ideas and titles, and (2) Content creation.

Content Ideas

1. Collaborate with my head of SEO to obtain a list of targeted keywords and questions people ask on Google related to those keywords. Then, use ChatGPT to generate thousands of additional, related questions that might arise in the future.

2. To stay relevant, gather a list of pop culture and news events. Feed them into ChatGPT and ask the AI to create headlines linking the news events to your industry and business. Update this list daily with new events and happenings.

3. Consult GPT for white paper ideas connected to your business and industry.

By following these four steps, you can generate thousands of headlines per hour, with each headline serving as a potential article, blog post, or white paper. Now, let's move on to content creation.

Content Creation

Blog Posts:

1. Develop and test prompts that encourage ChatGPT to produce high-quality blog posts based on the generated headlines. Begin with a prompt like: “Using this headline as a starting point, write a comprehensive and entertaining blog post in the style of [industry expert]. The post should be accurate, and [brand voice elements]. The post should be [length].” Iterate this process until the AI consistently generates exceptional content with minimal editing required. Experiment with various writing styles, and don't be afraid to venture outside your industry (e.g., writing styles of “Ernest Hemingway” or “Mark Twain”).

2. After creating the content, have a human editor review the post. You may also try having ChatGPT edit it first (“Edit the following post for grammar and readability. At the end, summarize all the before and after changes made.”).

3. Initially, aim for ten blog posts per workday, but anticipate an increase in velocity. Publish each blog post and nest it to support the SEO pages. Distribute a portion of the content via email.

White Papers:

White papers hold great value. They can be gated to collect email addresses, generate SEO links, be broken apart for content, or transformed into webinars. With ChatGPT, a single individual could potentially create an outstanding white paper every day.

Start by selecting a white paper idea generated by ChatGPT. Then, ask ChatGPT to create an outline for the white paper (table of contents). Provide the AI with the outline and ask it to write the content section by section, using the prompt-writing skills developed for blog posts.

White papers can be put behind reg walls to collect email addresses, but they can also be used to generate SEO links, broken apart for content, or turned into webinars. At GA we had a team of people working to generate a white paper every two months. WIth ChatGPT I expect one person could create a great white paper every day.

1. Start with the list of white paper ideas generated by ChatGPT. Choose one.

2. Ask ChatGP to generate an outline for a white paper (table of contents)

3. Now give ChatGPT the outline and ask it to write the content section by section. Use the prompt writing skills developed creating the blog posts

Tools/Analytics

It is early days, but I would want to be testing to see what new tools would be effective. I would begin by looking for a "co-pilot for marketing” product. We know the original co-pilot doubles or triples the productivity of developers. I expect there will be similar possibilities in all fields (I have seen the data for legal contracts. I have some friends building “copilot for product managers”. There will be more!)

I have found this one: Lighthouse by TripleWhale. It’ calls itself a co-pilot for marketers or ecommerce companies that use Shopify. Some features:

• Anomaly Detection: Alerts you when something is out of the ordinary. I remember when my CEO wanted me to build this back in 2014. It was not easy back then. Now it is a feature within a bigger third party app.

• Generative AI: It looks at your high performing ads and suggests new ones to test

• Data Stories: It will write up narratives in plane English to explain what is happening based on your data.

• Audiences: It will create new audience groupings for FAcebook targeting based on your best customers cut different ways

• Rules recommendations: It will generate and suggest new ways of spending money that it believes will increase marketing efficiency.

Will it work and be as good as it sounds? No idea. But the technology is there that this, or something like it, is either already available or available soon. For decades we have had the ability to collect data. But we still needed a strong analyst and/or data scientist (usually both) to get use out of the data. AI may not replace those employees, but it can definitely augment them.

Sales

Branching out a little from traditional marketing, sales teams should now be all over AI tools.

Mass Outreach

Many sales orgs are already buying emails and either mass mailing to them with boiler plate (or fill-in-the-blanks semi-boiler plate) or having human sales people craft personalized messages. The former is less effective. The latter is very time consuming (and frustrating as the response rate will still be very very low to unsolicited personalized messages). AI should at least allow for the best of both worlds: Fully automated mass mailings that seem fully personalized.

The way to do this is to match available email addresses to LinkedIn profiles (most B2B email purchase tools do this). Then use AI to read the profiles and create personalized messages. I expect those messages will be better crafted than what a junior sales person can manage. Here is a tool that does this (I expect the right answer is to pay more to have it all done automatically through APIs). It generated the following based on my profile:

Hey Edward,

I came across your experience in scaling companies and I'm intrigued to know more about your unique background. I am particularly interested in your experience as a Senior Advisor at Warburg Pincus. I would love to learn more about how you're helping companies scale across their portfolio. I recently read your book, Marketing BS, and it sounds like you have a unique perspective on the world of marketing.

It’s not perfect (I STILL haven’t made the book publicly available), but it’s… not bad? Now combine that with stuff about the individual “sending” the message to make a connection, along with an ask. Again, test different prompts, different asks, and different connection methods and measure response rate.

A similar thing can be done with all incoming messages. For incoming messages speed matters. Most of the time there is a trade off between speed and personalization. But now you can do both. You can instantly respond to incoming leads, make a personal connection, and try and set up a call (using a tool like Calendly). No need to wait until the sales person is ready. I expect you still want a sales person to respond to the highest potential value leads, and to second and third follow-up conversations, but I will bet conversion rate goes up with an instant initial AI response (maybe have it wait 5 or 6 minutes so it seems realistic… At GA we found CR peaked with a 7-minute delay in response).

You can also re-purpose drip campaigns so instead of low quality marketing follow-up or expensive BDR emails, AI can take the role of BDR and then kick it to sales if there is some sign of life.

Executive Co-pilot

I hinted at this earlier in the week. I would work at finding a tool that transcribes and summarizes all meetings in my organization and sends that information to a centralized repository. The AI should then do a few things:

1. Send a daily (or twice daily) message to executives on what has been accomplished, recommended next steps, and important things to know

2. Provide a ChatGPT-type tool that allows the executive to ask questions about what is happening and what people are working on

Rewind will do this for individuals. Maybe it is as simple as having everyone on the team use Rewind and then feed those datasets up to their manager. The goal is not surveillance (although clearly I see why that would be a concern for many people), but rather data collection to be able to keep track of what is happening with a large team.

I believe there are other changes coming very fast, and I am sure a day-long offsite with my team would generate more ideas. But I think the above list is a good place to start.

Keep it simple,

Edward