Five Minutes with Persado's Assaf Baciu

AI Street recently sat down with Assaf Baciu, co-founder and President of Persado, to discuss how generative AI is transforming marketing.

Persado, a New York-based company founded in 2012, has established itself as a leader in AI-driven marketing solutions. Working with eight of the 10 largest U.S. banks, Persado's AI technology harnesses vast amounts of customer behavior data to tailor marketing language down to the individual level. Clients include Chase, Ally, SoFi, and NatWest.

The company has generated more than $2.5 billion in additional revenue for its financial services clients over the past five years. Baciu shares his insights on implementing AI in a heavily regulated industry and broader trends in AI adoption among financial institutions.

This interview has been edited for clarity and length.

What does Persado do?

We are a Gen AI platform that has multiple capabilities. We have the ability to generate specific words, phrases, emotions and tones that will cater to a given audience, a segment, or even a person. 

How does Persado use Large Language Models?

We use LLMs as raw material, enhancing them to create domain-specific models for marketing in financial services. Our approach isn't just about using off-the-shelf LLMs, but about customizing and training these models on our knowledge base of customer behavior. This dataset comes from over a million content experiments conducted throughout a decade, enabling us to understand how specific language resonates with different audience segments. By combining open LLMs with our proprietary data and domain expertise, we create AI models specifically tuned for generating high-performing marketing content in the financial services sector—resulting in custom-built LLMs that reflect a brand’s unique voice.

We engage with about 64 million Americans every year and have analyzed their reactions to certain prompts. So whether you clicked on a campaign message or not, whether you purchased, started, or finished an application, all this data comes back to us anonymously, and we understand how to associate that action with the language that we showed. 

How is Persado used in financial services?

We're used across various aspects of marketing, from card acquisition and account upgrades to loyalty programs and early engagement. Our AI generates language for websites, emails, and other channels to drive customer engagement and conversions. Driving customer engagement can be non-transactional as well, such as going paperless, payments, or IVR.

Our generated content performs better than human-written, or other AI-generated, content 96% of the time, driving an average of 40% increase in conversions and we work with eight of the top 10 banks in the US by assets.

How does Persado tailor content for different audiences?

We developed our own knowledge graph that says certain words convey emotions in social psychology. About 10 years ago, we refined them to 15 that apply in the context of marketing. 

To see how this works, let's consider an example: If I'm selling a cruise, I may focus on wanderlust for one category of users, and convenience or health for another category of users. By tagging words and phrases with emotions relevant to marketing segments, we can predict how language will engage people across various platforms like web, email, SMS, or mobile.

What impact has Persado had on client revenues in the financial services sector?

Over the past five years, we've helped our financial services clients generate more than $2.5 billion in additional revenue. This is measured to the dollar through A/B testing [a method of comparing multiple versions of content to see which performs better] of our generated content against the client's original content. 

Who are some of your clients in the financial services sector?

We work with Chase and Ally Financial, as well as many other banks and card issuers. We also work with online lenders like Lending Club. Eight of the largest U.S. banks by assets, use our platform.

How does Persado navigate regulations?

Our content for financial services adheres to many regulations, like the Fair Lending Act. We can't directly associate protected classes [groups legally shielded from discrimination] with specific treatments. Instead, we create different narratives or emotional tones that are exposed to everyone, and then we observe which messages resonate with different segments.

How has the emergence of ChatGPT affected your business?

It's had two main impacts. First, we no longer need to explain that machines can write - everyone understands AI now. However, there's also a lot of confusion about what to do with AI and businesses  are still figuring out how to connect AI use cases with tangible value. ChatGPT has also exposed the need for differentiated enterprise content, governance, and models that don’t “hallucinate” or risk copyright infringement.

What AI adoption trends do you foresee in the next 12-18 months?

We expect companies to start scaling Gen AI in marketing and customer service. Many large companies will likely abandon in-house AI development in favor of specialized providers. It's hard for a single company to continuously invest in and evolve with the latest capabilities.

What has surprised you about recent AI developments?

Even for practitioners in the field, the capabilities of these models are still amazing. It seems like the bigger models behave in a way that certain areas of the brain get to be more specialized without any specific guidance to do that. So it's an awkward outcome, right? That a bigger model is better, but it's better because it actually develops specialization in certain areas of the layers. And this, even for a practitioner, it’s quite amazing.

Thanks for reading! Feedback? Reach out at [email protected].