Everything You Need To Know About Building A Successful Marketing AI Team

I want to be upfront.

I am exactly zero percent objective about Artificial Intelligence.

I believe it is completely disrupting the entire world, especially business as we know it today. 

I am not one of those former Big Tech Presidents that hop on The {Name Redacted to spare us another Twitter war} Podcast to tell you that we’ll never reach Singularity and everything will be ok. While, at the same time, he makes all sorts of huge investments in the background, proving that he thinks ASI is barreling toward us faster than another superhero movie.

There’s no doubt in my mind that the earth will be scorched, and only the most prepared businesses will survive. (Notice I didn’t say biggest or best. I said, prepared. Half of this battle is about survival. The other part will be about adapting and disrupting.)

However, even if I’m 10% wrong, or half wrong, or all wrong, you already know that AI is changing our personal and professional lives and that, if nothing else, it’s prudent for you to have a solid foundational knowledge of what AI is, what it does and how it can benefit you.

Even if you never do anything AI-powered, AI-assisted, AI-enabled, or AI-related in your company, you know for sure that the search engines, social networks, biggest retailers, marketplaces, financial institutions, and the most used appliances and vehicles are — and will continue — to enhance and evolve with things like computer vision, natural language processing, and machine learning.

One way or another, you will need to accommodate it. Otherwise, you will risk being left behind. (Or, you know, incinerated.)

So, as a marketer, how do you get started?

START SMALL. BLOOM BIG.

The most successful companies using AI start with a tiny seed and grow it into a beautiful flower. (Or, you know, a rapidly-spreading Kudzu plant…) 

Sometimes the first AI project starts in Marketing. Other times, it begins in IT or Customer Service. And then it spreads from there.

At some point, you’ll decide to commit to “this whole AI thing,” much like we devoted ourselves to “the crazy WORLD WIDE WEB” schtick over two decades ago and you’ll want/need a team solely dedicated to the advancement of AI in your company. For now, however? Let’s just prepare…

What does an AI team look like?

Artificial Intelligence teams look different in every business. As in the early days of the internet, some companies have folks working on AI projects in their IT departments. Others have marketers working with outside vendors whose software/services are powered by AI. Many companies have a hybrid version – some inside stuff and some outside stuff. There’s no one right way to do this. You need to decide what’s best for your company based on your size, budget, talent, expertise in the subject matter, projects, bandwidth, and so on.

 

What does an AI team do?

One of the best ways to figure out what you want your AI team to look like is to determine what you want your AI team to do. For many companies, a solid AI team:

Develops and maintains the overall vision for AI. This includes knowing what can and can’t be done with the current technology and, more importantly, within the scope of your organization, its practices, and policies.

Solicits and screens ideas. Builds and prioritizes business cases. Allocates resources.

Develops, manages, and implements AI projects.

Ensures the data is right for the job.

Communicates the process and concepts to other divisions/teams so that everyone has a similar understanding of how AI is being used within the organization.

Works with Legal and other departments to develop standards for AI and its use. This includes things like privacy and bias policies and data and security standards.

Evangelizes AI within the organization. This includes working with other business leaders and stakeholders to identify new and improve existing projects and sharing the successes (and failures) of the company’s AI projects.

 

Should I outsource or build an in-house Marketing AI team?

There’s no one-size-fits-all answer to this question either.

Some companies do everything in-house for security and legal reasons. Or because that’s just what they prefer to do. Other businesses outsource everything because they want to be in pools of successful companies who are working on similar projects but in industries that aren’t competitive. Many organizations build blended/hybrid teams because they have challenges hiring enough people or paying market rates for salaries. Lots of folks augment their in-house teams with outside people because they want specific and/or overseas talent and experience.

 

What kinds of people are on an AI team?

AI is typically a team sport. It takes a lot of different types of people to build, deploy and perfect AI projects. The most successful AI teams collaborate with all the relevant stakeholders and divisions. (Please note that many AI teams start with one person and then grow from there. Please don’t let the list below scare you. You certainly don’t need all these people, especially if you use outside vendors/services.)

On Artificial Intelligence teams, you often see the following titles:

Data Scientists – identify use cases, determine appropriate datasets and algorithms and build AI models.

Data Engineers – Collect, manage and convert your raw data into usable information.

AI Architects, Product Owners, and Managers – typically the glue that holds everyone together; they oversee/lead the projects.

Machine Learning Engineers – create, design, deploy and optimize AI models/algorithms.

Solutions Architects – determine which technology to use and how all the technologies work with each other.

QA Specialists and Testers – make sure everything is tested and approved.

AI Translators (aka the Go-Betweens or the Middlemen) – connect the AI team to the organization’s other business units. Ensure that everyone understands the projects, how you’re determining success, and so on.

AI Analysts – analyze the results and then communicate them to the rest of the organization. Usually in charge of testing and control protocols.

Other titles you might come across: CAIO (Chief AI Officer), Data Modeler, Research Scientist, Deep Learning Specialist, Machine Learning Researcher, Domain Expert, Solutions Architect, AI Ethicist, AI Sociologist, Applied Scientist, DevOps Engineer, AI Mathematician, Change Management Researcher, Conversational UX Designer/Researcher, AIHR Manager, and AI Lawyer.

It’s important to note that many organizations have AI team members embedded throughout, not just in one department.

 

Do you need all those people when you’re first starting?

Hard no. Even the very biggest companies go lean-and-mean at the beginning.

The more projects you take on, the bigger your team will be, but most AI teams these days are still only a handful of people who are empowered to act quickly and fail often. I know the latter is cliché, but alas… With AI, if you’re not failing often, you’re not taking enough big swings.

The most successful companies using AI – both big and small – have well-rounded teams that include people with different talents and expertise. There’s usually someone who understands the business (domain expert); someone who understands the data (data scientist); someone who knows how to measure things (analyst); and someone who holds everything together (architect or project manager.) As you grow, an AI translator can really help. Translators are often the most effective at getting widespread adoption of AI within an organization.

Word of advice? Don’t build a big team before you know how AI will work in your organization. I’ve seen many marketers build AI teams only to realize that their data needs lots of work before they begin. (Work = data massaging, hygiene, compiling in a centralized location, and so on.)

Many experts say you need at least one person with deep modeling knowledge. In my experience, you need at least one (hopefully many) person who really cares about the data, its quality, how you’re using it, how you’re not using it, and so on. Someone who knows the data like the back of their hand can be way more helpful than a modeling maestro with no idea if/where your data sucks. AI is a prediction machine, and the best way to arm it is to give it solid historical and current data. (You can read more about data here. It’s critical.)

 

What’s the most challenging thing about building an AI team?

In the South, there’s this saying, “fixin’ to get ready,” which basically means getting ready to get ready?!? Sadly, that’s why a lot of AI projects never even get started. Folks spend oodles of time overthinking their initial projects. They develop eleventy bazillion plans about how things will go, but since they don’t have any experience with AI projects, it’s an exercise in futility.  In the beginning, this fixin’ time is just a waste. I’m not recommending you dive head-first into the pool without knowing how deep the water is, but I don’t recommend doing a 365-day water sampling of a chlorinated pool at The Equinox before you begin. When you’re starting out, you need people willing to jump in and start making some splashes.

As your team gets bigger, finding the right level of people becomes an issue. Marketers and IT folks tend to squawk rather incessantly about how hard it is to find good AI people as if AI folks are some sort of 1 in a 1,000,000,000 unicorn. They’re not. The most well-known people are hard to get right now, but it’s not that they’re not available; it’s that they’re uber-expensive (comparatively), and most companies don’t want to pay for them.

Instead of hyperventilating over the unicorns or lack thereof, find people who care about data and/or have strong math skills; are perseverant; embrace change and/or are adaptable; like to learn/grow, and are obsessed with success (but certainly not afraid of failure.) Then, get them resources and training.

The good news is that a lot of AI projects are cut-and-dry. You don’t need a Michelin-starred chef in the world’s most perfect kitchen to make good food. You need a solid cook who can put yummy meals on the table.

 

What’s one of the biggest mistakes that Marketing AI teams make?

Many companies treat AI as an IT project. Or a math/technology project.

It’s not.

AI is a company-wide project; even if everyone isn’t involved initially, folks must know what’s happening. I can hear the questions now… “If I own a hotel chain, you’re saying the housekeepers need to know that we’re using AI?” Not necessarily. However, AI is not just an IT thing or a Marketing thing; your senior people must know about AI because it will likely impact more than just a few areas of the business. You’ll know when it’s best to loop in your Operations people, your Financial people, and so on. Just remember, as it gets more successful in your business, it has the potential to change EVERYTHING.

But is that really the biggest or most common mistake that I see?

No. AI Marketers quickly learn that data is gold and communication and collaboration are key. They also tend to understand that time whizzes by in AI-land and that they need to move fast too. What teams often miss is that there are places it’s ok to cut corners and places that aren’t. For example, model training. 9.9 times of 10 training is NOT the place to cut any corners.  However, even if you have a team of 1000, it’s easy to talk almost anyone into “just shaving a few days off the training schedule” because few (read: nobody) appreciate(s) things like model training or data hygiene like they should. (Granted, both are zzzzzzzz—snoring.)

Even if you’re a team of two people, you need someone to play the role of AI Cop. This person (or people if you have a big team) will police the situation to ensure you’ve given the project its very best shot. If a project has been completed and everyone is second-guessing the situation/results, your AI Cop didn’t do a great job. If you feel you gave it your very best, then the AI Cop likely succeeded. Giving something your very best shot doesn’t mean it’s perfect or that you can’t list any possible improvements; it means that you left it all on the field and you’re confident that you’ll be able to review and act on the results.

 

What are some best practices for developing a successful Marketing AI team?

Ensure you have at least one domain expert per project. In this case, domain means subject matter and company. Domain experts are usually NOT AI/tech experts; they are people who know the company and the product(s) inside and out, top to bottom. Incidentally, many times you need different domain experts for different things, depending on what problem you are trying to solve. Remember, a deep understanding of your customers and their needs goes a long way when building AI projects. (At its core, AI is just a smart machine.)

Give the team its own separate budget even if other areas (IT, Customer Service, for example) have AI as part of their budgets.

Develop a structure for communication within the team and your stakeholders/shareholders, too. Don’t just assign it to one person. Design an automated process for sending out updates and/or results in summaries (successes and failures.) Encourage other team members to share how one AI project directly impacted their department.

Develop your own Questions List/Checklist. Get your team into the habit of asking yourselves specific questions before the project starts to well after the post-mortem. The questions will help you stay on track, maximize efficiency and improve your current/future processes. They’ll also be a great reference when reviewing/disrupting after project completion. Examples of questions: how will this change how our users/visitors view our brand? As the learning grows and becomes deeper, what things will tell us that we need to instigate disruption? What other kinds of data would improve this project?

Solicit feedback on an ongoing basis. (Not just when you’re launching.)

Recognize that there’s often a difference between formally trained data scientists and folks (marketers, managers, etc.) who are great with analytics and that there’s a place for both kinds of people on your team. Data will be one of the biggest keys to your success in any AI project, and it’s essential to know how to read it and use it to drive revenues, increase performance, maximize savings, reduce employee downtime, and so on.

Implement a post-mortem process for every project, whether it was a failure or a success. Recap how you did and how you could do it better next time. Make sure to ask yourself whether you took full advantage of your technology and data. Again, AI is new, and your first projects may be rocky, but they should get better in time. One of the ways you’ll keep improving your process is to acknowledge where you went right and what you could do better next time.

Some of the above are specifically for Marketing AI (other divisions, like IT, often need different things.)

 

What’s the #1 characteristic of the most successful AI teams?

Excellent communication.

A lot of this stuff is new, and people are learning as they grow. Throughout the journey, there are — and will continue to be – many successes and failures. The companies that have the most rock-solid AI teams all share one thing… They communicate efficiently and effectively with each other and the rest of the company/stakeholders. Prioritizing communication isn’t always easy, especially when a project fails or you need to relay bad news to your other team members, but it’s essential. 

Got the communication thing down pat? Work on mentorship, shadowing programs, and building stronger team connections so that everyone always has someone else they can go to for advice and support. This sounds very rainbows and unicorns rah-rah, but AI teams have high position turnover, and it’s best to be prepared if someone moves up or out of your organization. 

Are you building an AI team, or do you already have one? Have any tips you’d like to share? Questions you’d like to ask? Tweet @amyafrica or write info@eightbyeight.com

.