Artificial Intelligence

How Credit Karma is blending traditional AI with embedded GenAI to create more impactful user experiences

  • With Intuit’s infrastructure behind it, Credit Karma is turning up the dial on GenAI.
  • In this exclusive, Maddie Daianu, Head of Data and AI at Credit Karma, shares what that looks like behind the scenes.
close

Email a Friend

How Credit Karma is blending traditional AI with embedded GenAI to create more impactful user experiences

Intuit has been quietly building a fintech empire in plain sight, mostly by buying its way in. When Intuit acquired Credit Karma for $7.1 billion in 2020, it didn’t just enter the consumer finance space — it tapped into a massive user network and secured ongoing user engagement through tools like credit monitoring, loan offers, and tax prep via TurboTax.

In this exclusive, I speak with Maddie Daianu, Head of Data and AI at Credit Karma, about how the company is leveraging Intuit’s infrastructure to explore generative AI (GenAI), how she’s steering its data and AI strategy, and where traditional AI and GenAI come together to elevate the user experience.

What it takes to build AI-backed user experiences

Maddie Daianu heads Data and AI at Intuit Credit Karma, where she steers her team toward a broader goal: creating a full-spectrum experience for members beyond just personalized touchpoints.

Maddie Daianu, Head of Data and AI at Credit Karma

“We’re building end-to-end data and AI infrastructure that scales meaningful AI-driven experiences to our 140+ million members, and transforms technology into tangible business results,” says Daianu.

She notes the firm has prioritized AI from the very start, investing heavily in the platform and sophisticated machine learning models that personalize recommendations for users across their app, emails, and push notifications.

“By automating the AI lifecycle, we deploy over 22,000 models monthly, generating 60 billion daily predictions. We’ve built a data-driven culture into Credit Karma’s DNA from day one, empowering teams to experiment, learn, and innovate through our integrated platform,” she notes.

Credit Karma’s Data and AI team structure: Credit Karma’s Data and AI team is made up of over 100 employees across Data Science, Machine Learning Engineering, Data Platform, Business Intelligence, and Experimentation Platform. 

Dismantling organizational silos has enabled the formation of an integrated framework, where Consumer teams, like Data Science, work closely with Provider teams such as ML and Data Engineering. “This unified setup brings both technology consumers and providers to the same table, fostering empathy and alignment from the outset, to tackle problems with clarity and purpose,” notes Daianu. 

She says the idea that ‘you ship your org’ has long guided her approach. Through the development of a cohesive Data & AI function, the team has established a strong operational backbone that supports quick innovation and collaboration.

“In the GenAI era, team structure must be resilient,” asserts Daianu. Given the pace of change in tech, Daianu and her leadership circle now emphasize flexibility in team structures, enabling faster decision-making and continuous adaptation instead of being stuck in predefined roles or rigid frameworks that might slow them down.

Rooted in honesty and accountability, Daianu’s cultural approach is encouraging a fail-fast, iterative mindset. She believes this approach empowers her team to openly assess what’s not working, let go of ineffective efforts, and pivot quickly, fueling ongoing improvements.

Building on Intuit’s GenOS: Daianu explains that in recent years, Credit Karma’s impact has grown faster by integrating with Intuit’s proprietary Generative AI Operating System (GenOS), which gives them the foundational infrastructure and tools for quick experimentation with GenAI.

“This has been essential for launching the first GenAI embedded product experiences at Credit Karma, helping power contextualized recommendations that provide clarifying explanations as to why we recommend a certain product or tool to a member, tailored to each member’s unique financial profile and needs,” she notes.

Why the embedded GenAI experience: Through extensive qualitative and quantitative research, Credit Karma found that because each person’s financial journey is distinct and personalized, one-on-one guidance is essential to be effective. Despite significant advancements in using data, machine learning, and AI to help members understand and manage their finances, many US consumers still lack the confidence to make sound financial decisions.

Recognizing this gap, Daianu says they’ve expanded GenAI capabilities on Credit Karma with the embedded approach. This approach involves integrating GenAI-generated insights directly into the app experience, so that as members scroll, they’re presented with personalized explanations and guidance. The goal is to boost financial awareness, confidence, and decision-making, whether members are exploring the right financial product for themselves or responding to changes in their financial situation.

An inside look at how Credit Karma is applying GenAI across use cases

One way Credit Karma is leveraging GenAI is through Intuit Assist. This GenAI-powered financial assistant provides financial guidance and capabilities across the company’s product ecosystem, including TurboTax, QuickBooks, and Mailchimp.

For Credit Karma users, any shift in their credit report sets off a GenAI-powered contextual summary that outlines what changed, how it affects their score, and what they can do about it. They also get a clear breakdown of the reasons behind the score movement.

“Since the unveiling of Intuit Assist in September 2023, we’ve learned so much about members’ biggest financial pain points and how we can iterate our GenAI offerings to better solve their problems,” says Daianu.

Another feature is ‘See Why,’ powered by Intuit Assist, which offers personalized explanations behind credit card recommendations. By tying these insights to a member’s financial data, it helps them better understand their options.

“These are just a few examples of how we’ve embedded the reasoning and explainability capabilities of GenAI, married with our rich financial data, to meet members where they are and equip them with personalized information,” she adds.

AI and generative AI work together to power Credit Karma’s Recommendations Engine: Credit Karma’s Recommendations Engine is built on its advanced machine learning infrastructure, which ranks content and offers based on each member’s individual financial profile. 

The idea is to surface the most relevant tools, insights, and products at the right time. For example, if a member is exploring loan options but has a high debt-to-income ratio, the system may highlight this as an area for improvement for them to take actionable steps toward securing better terms.

It is important to note that the Recommendations Engine itself doesn’t rely on generative AI. The Intuit Assist–powered ‘See Why’ feature backs the recommendation, demonstrating how the company combines traditional AI with GenAI. “While our recommendations engine intelligently serves members relevant content and offers, Intuit Assist helps contextualize to the member as to why we suggest a certain insight or offer,” says Daianu. “So, they work together to essentially equip our members with the information they need to make more informed decisions about their finances.”

But, “There’s no good AI without good data.” While GenAI has made it easier for Credit Karma to explain the ‘why’ behind its recommendation to members, as Daianu puts it, “there’s no good AI without good data.” High-quality data remains the foundation of everything.

“Data only becomes truly powerful when unlocked by AI, and conversely, AI reaches its full potential only when fueled by high-quality data,” she asserts. 

She further explains that Credit Karma’s progress in GenAI is built on a strong foundation of early investment in data infrastructure. The firm has built specialized infrastructure that links large language models with the right data sources, using de-identified member and product information to deliver personalized, relevant experiences.

“Our success doesn’t just lie in the quality of financial data we have on our platform, but also its historical aspect – the longitudinal, multi-year tracking of our members’ financial journeys within our app,” she notes. 

Daianu points to Credit Karma’s rich data history — spanning over 15 years — as the key differentiator in its AI strategy, forming the backbone of both its Recommendation Engine and generative AI innovations.

“Tax data is a huge competitive advantage for the Intuit ecosystem and will help us further advance our members’ financial goals in an impactful way,” she notes.

Testing, learning, and overcoming challenges of GenAI-powered solutions

Credit Karma continuously iterates on its data and AI systems — testing, learning, and evolving. 

Even so, a few challenges still need to be addressed. Unlike traditional software development, where updates are more clearly defined and straightforward to deploy, working with GenAI — especially in areas involving sensitive financial data — requires a more careful and layered approach.

Because of this complexity, “major upgrades” in GenAI aren’t always black and white. Instead, the company follows an ML-first lifecycle mindset, emphasizing continuous experimentation and rigorous validation to drive responsible innovation, shares Daianu.

“Much of the progress my team has made happens behind the scenes,” she says, “where the stakes are incredibly high given the sensitive nature of the financial information we work with.”

Adding to the broad spectrum of challenges is the unpredictable behavior of large language models (LLMs), which introduces new variables at every step.

To overcome these challenges, Credit Karma takes on strategic approaches:

  • Interoperability & ROI: Integrating GenAI into existing systems poses challenges, from computational resource requirements to stringent data and compliance standards. Despite industry-wide efforts, GenAI has yet to achieve widespread product-market fit, often resulting in lower-than-expected ROI. 

To overcome this, Credit Karma started small and empowered agile teams to creatively leverage GenAI, focusing on solving real user problems. “We found success by shifting from a chatbot to embedded GenAI experiences, like ‘See Why,’ that contextualized financial recommendations within our app, helping members navigate their financial journeys with more confidence,” says Daianu.

  • Safety & Governance: Fintech and highly regulated industries face challenges in leveraging GenAI, largely due to strict regulatory frameworks. To address these challenges, the Data and AI team focuses on building compliant infrastructure and deploying advanced techniques to ensure accurate experiences. 

“Given GenAI’s non-deterministic nature, which poses governance challenges like hallucinations and bias, we’ve made evaluation frameworks central to our GenAI stack, investing in scalable methodologies to detect and remove bias, and ensure high-quality outputs,” Daianu shares.

  • Expertise & Customization: Daianu points out that working in this space means operating in a state of constant evolution — it demands agility, deep expertise, and significant investment.

Successfully deploying GenAI internally isn’t as simple as plugging in a model; it requires a multidisciplinary team of specialists — from data scientists and engineers to ML and software developers — who can fine-tune the tech to specific products and member needs.

She emphasizes that building this kind of specialized GenAI capability doesn’t happen overnight – it takes time and significant resources. Without that foundation, companies risk misapplying GenAI to problems where traditional machine learning could be more effective. “To avoid unnecessary complexity, we’re investing in our talent at Credit Karma and developing hybrid approaches that balance deterministic and non-deterministic solutions,” says Daianu.

“While having a robust GenAI strategy is critical in today’s fast-paced landscape – especially if you want to maintain a competitive edge, for us, it’s about building dependable capabilities that balance innovation with quality, safety, and reliability,” adds Daianu. 

She explains that pursuing this path requires heavy engineering involvement and infrastructure overhauls, which can occasionally — and temporarily — slow the pace of innovation.


[Sidebar]: More GenAI features coming soon for Credit Karma users

Daianu explains that Credit Karma’s innovation efforts around generative AI have primarily taken place at the infrastructure and platform level to support building and scaling such experiences. When it comes to using GenAI to deliver more advanced experiences for members, she says the company has made significant progress — momentum that will enable it to roll out more member-facing, GenAI-driven features in the near future.

Daianu says: “Though I can’t share specifics just yet, we’re committed to pushing the boundaries of embedded GenAI experiences to transform member interactions into intuitive and fully personalized ‘done for you’ experiences.  We’re really excited about the future of GenAI and its immense potential to empower our members to tackle their most difficult financial challenges with less effort.

Resources like Intuit’s GenOS platform have enabled Intuit’s brands to do the unimaginable when it comes to scaling GenAI development. Everything from providing components that provide developers with dedicated environments to rapidly experiment and refine GenAI experiences, the ability to leverage and build capabilities that enable agentic workflows, moving beyond basic prompting to autonomous planning, reasoning, and execution to tackle complex business workflows, and a vast user experience (UX) framework that designers and front-end developers can leverage to build consistent and intuitive experiences. Building on this momentum, we’ll continue to invest in modernizing Intuit’s Consumer Ecosystem, driving AI system- and product-level innovation, while developing the next generation of data products that will unlock the full potential of this innovation.

I’m also immensely proud of my team’s creativity in unlocking breakthrough innovation while maintaining the utmost accuracy and safety in our GenAI products. Achieving this delicate balance is a remarkable feat, especially in a highly regulated industry where staying relevant and competitive demands constant innovation and resilience. 

More progress to come soon!”

0 comments on “How Credit Karma is blending traditional AI with embedded GenAI to create more impactful user experiences”

Artificial Intelligence

Why 70% of banks don’t report the outcomes of Gen AI deployments 

  • Major banks like Bank of America, Citizens, and JPMorgan are actively deploying Gen AI tools for employee productivity, but nearly 70% of AI use cases don't have any reported outcomes or measurable ROI.
  • The lack of results stems from difficulty separating Gen AI progress from overall growth, challenges in mapping cost impacts from internal efficiency gains, and an inability to translate the AI hype into consumer and societal impact.
Rabab Ahsan | August 26, 2025
Artificial Intelligence, SMB Finance

How Intuit is designing embedded AI agents in QuickBooks to serve SMBs

  • Traditional automation has helped with repetitive tasks, but often falls short when workflows get complex or unpredictable. In addition, key data remains siloed.
  • Intuit is tackling this issue by embedding AI agents directly into QuickBooks, supporting core functions like payments, accounting, finance, and customer support to better serve SMBs.
Sara Khairi | July 31, 2025
Artificial Intelligence, Banking, Member Exclusive

How Bank of America cracked the code on AI adoption by making Erica indispensable to both customers and employees

  • Bank of America transformed its customer chatbot Erica into an employee productivity powerhouse, achieving 50% IT service desk automation by strategically targeting common pain points and building adoption incrementally over five years.
  • Learn the adoption secrets behind getting 90% of employees to embrace AI tools, including how Bank of America overcame the adoption hump and integrated generative AI with 25 proof-of-concept projects now entering production.
Rabab Ahsan | July 08, 2025
Artificial Intelligence, Business of Fintech

Inside Shopify’s company-wide Gen AI push under Tobias Lütke

  • While traditional banks and credit unions have adopted a "slow and careful" approach to generative AI, limiting deployment to back-office operations, fintech companies like Shopify are taking radically different paths by integrating AI throughout their entire organizations and customer-facing products.
  • CEO Tobias Lütke's leaked manifesto reveals a mandatory AI-usage policy that affects everything from employee performance reviews to resource allocation, but this aggressive approach highlights unresolved challenges around error management, equitable implementation across organizational hierarchies, and the need for proper safeguards in this new paradigm of AI-augmented work.
Rabab Ahsan | July 01, 2025
Artificial Intelligence, Lending

Meet the AI lending officer: How algorithms are taking over credit decisions

  • Small business lending represents the last major frontier for AI transformation in financial services.
  • Visa's Jonathan Kolozsvary and Uplinq's Patrick Reily explore how AI lending officers are using transaction data to unlock capital for underserved businesses.
Shane Galvin | July 01, 2025
More Articles