6 marketing automation use cases where AI can help with data quality

Will AI help address age-old marketing issues or just add more complexity?

For much of 2023, the AI hype has focused on generative AI content use cases (copy, image, video). Some still question generative AI’s ultimate impact, but the mainstream adoption indicates that much of the focus on content-focused capabilities is warranted.

And yet, there’s an even more profound movement afoot: The infusion of AI into every marketing technology application.

For martech leaders, infusing AI into core stack components like CRM and marketing automation platforms (MAPs) will increase accuracy and productivity. Within that scope, my focus has been prioritizing data management, which most marketing operations leaders also recognize as the bedrock of the foundation.

Data management: The first (semi) natural language process

Before the AI inflection point, data management was the earliest “natural language” change that fueled martech growth. How? Through the no-code transformation that empowered us to create new database fields, a privilege previously reserved for IT. The ability to create internal and customer-facing fields integrated into landing pages and websites transformed digital engagement.

Even with automation, we rely heavily on human interaction and system interfaces to drive much of the input. And despite easier-to-use tools, training was still an adoption barrier to (proper) data input. Early AI algorithms impacted various data cleaning processes after data was entered improperly or was incomplete. But, we all knew it was most efficient to prevent inaccurate data from entering the system, which would result in erroneous results downstream.

I’ll use a common framework — garbage in, garbage out (GIGO) — to illustrate.

‘Garbage in’

1. Entering data

Martech leaders cringe when users say entering the data is hard. Empathy is deserved, especially when there have been changes to the interface over time. (If you’re a Salesforce shop, and still switch to Classic vs. Lightning, that’s your empathy reminder!)

Many leading vendors, including Salesforce, have recently predicted that the generative AI “prompt” revolution will forever change the user interface. Every UI now needs to process natural language, reducing the friction (or excuse, if you’re cynical) for users to enter data.

For example, ChatSpot (HubSpot’s AI interface) leverages the GPT model in its user interface. (While I’m vendor-agnostic, I have been leveraging the tool and will excerpt examples because it’s available to test in their public alpha release.)

Let’s start with the basics — adding a new contact.

Users won’t have to remember where in HubSpot’s standard interface to click “Add Contact.” Instead, they’ll use a simple prompt like this…

n three months of alpha, HubSpot has also added prompt templates that trigger actions based on common to-do’s, so you can now choose from a favorites list like this.

2. Researching and adding data about people and companies

Many MAPs pulled in basic customer information from websites. AI is simplifying this task, and now a summary version of key profiles to augment contact personas or supplement company firmographic info is a prompt away. For example:

3. Infused in spreadsheets

Approximately 70% of marketers spend more than 10 hours a week working on spreadsheets, according to MarTech’s 2023 Salary and Career Survey. They are foundational in martech stacks.

I spoke about how these tools (and their formulas, VLOOKUP capabilities, etc.) are still our secret decoders for working across multiple data sources in my March 2023 MarTech conference presentation. For many larger teams, a full-time data analyst supports these efforts. Smaller teams often have a data-savvy marketer with Excel expertise.

However, programming VLOOKUP is too technical for many. Marketers are now using generative AI prompts to create formulas. Several AI plug-in utilities infuse AI-created prompts directly into spreadsheets.

These natural language “no-code” capabilities will be the most powerful and most-used additions. They will be embedded directly into foundational knowledge work tools (e.g., Google Workspace Labs and Microsoft Co-pilot). Users will ask an AI assistant to extract domains from email addresses, pull out first/last names, companies, etc., and effectively create structured data through natural language prompts.

‘Garbage out’

Let’s now flip to the other side of the spectrum: Use cases where AI will help with data output.

4. Natural language interfaces for analytics

We’ve all been there. Rather than access the platform, someone asks you to export a report in PowerPoint or Google Slides. Getting the report from the application through natural language prompts will be a game-changer.

“Can you give me a report based on <fill in the blank>” will be a prompt that lowers the barrier for more people to access analytics directly.

Over time, if users are more inclined to enter the data and see it properly reflected, they will be more likely to provide quality entries. Instead of fixing the chart, perhaps users will fix it at the source.

5. Infused visualization capabilities

Creating visualization will also be infused capabilities. We’ll be able to prompt the platforms for these visualizations through plug-ins/interfaces.

Like many, I eagerly await access to OpenAI’s code interpreter capabilities. In the meantime, I’ve been following others piloting it, including Ethan Mollick, who provided a sneak peek at the capabilities in his One Useful Thing newsletter — excerpted in his recent newsletter post.

6. Accessible big data

All of these data entry and output benefits will not just be limited to the specific data that is “source-of-truth” in CRM/MAP.

Because we’ve lowered the barrier to entry for more data sources, then the outputs of one analysis may be linked in ways to others that were not considered previously — as other data augmentation and supplemental attributes will be accessible — through AI-based prompts as well.

Governance and training still needed to avoid blind trust

Martech leaders need to be careful not to rely on AI alone for data management and quality. Additional governance should be applied given the immaturity of the generative AI tools and their potential to impact data quality if not supervised.

The challenge for data management has twice the impact. Prompts may not inherit your organization’s guidelines for associating contacts with accounts; more advanced prompts that follow those guidelines may need to be developed.

Today, anyone who imports data into a spreadsheet does a sanity check after applying formulas. Typos can generate issues across thousands of records. But faulty AI-introduced logic can corrupt thousands of records if the users didn’t create the appropriate prompt to begin with.

__________________________________

This article is written by Martech and originally published here.

Author