Could fintech innovation be the best thing that ever happened for financial consumers? Or will it make their lives worse? Or maybe… some of both?

These were the questions we dug into at CFA’s Financial Services Conference in Washington on November 30. I was delighted to moderate a panel with Lauren Saunders, Associate Director of the National Consumer Law Center; Elizabeth Kelly, SVP of Operations for United Income; and Greg Sager, Associate General Counsel of LendUp.

Our premise was simple. The financial sector is being transformed by the same huge currents of technology — from big data and artificial intelligence to smart voice assistants and cloud computing — that are reshaping every other aspect of our lives. People who care about financial consumers need to look closely at these innovations and consider whether some could help solve problems that have bedeviled consumers for decades, while also examining the risks they may carry.

Consumers struggle in their financial lives for many reasons beyond lack of money. Challenges arise from an intricate interplay of factors that include the inherently high cost of financial services; complexity and opacity of many services; product designs that can hide penalty terms; discrimination and abusive practices; lack of access to services like bank branches; lack of consumer financial literacy; difficulty qualifying for loans due to credit history; lack of documents to authenticate identity; the slowness and unpredictability of the payment system in clearing checks and debits; lack of access to investment advice; and the onerousness of budgeting and the difficulty of saving, to name a few.

Today, fintech innovators are attacking all of these sources of consumer distress. They are creating affordable and user-friendly mobile delivery channels, online lending options, efficient loan approvals, instant payments, affordable robo-investment advice, “insure-tech” products, blockchains that speed up payments and processes, apps that make it easy and fun to save and budget, apps that handle bill-paying and whittle down debt, easy ways to compare financial options, products that are simple, products that have no fees, credit cards that encourage borrowers to pay the full balance each month, ways to smooth out volatile income and expenses, ways to pay gig workers immediately after their gigs, and easier ways to manage benefits payments. Banks can open accounts for people using new, more inclusive kinds of identity data. Remittances are becoming more affordable due to nearly cost-free ways to transmit funds over the internet. Small businesses, meanwhile, can easily digitize their financial records through mobile phone-based payment readers, and use the digitized data to qualify for affordable loans online. Also, across the board, “regtech” is making compliance easier and cheaper, reducing barriers to serving consumers affordably.

Our session discussion began by looking at this good news. We had two fintech representatives on the panel, who described what their companies do and how they use new technology to make their products easy to use, keep costs low, and solve problems faced by traditional financial firms. LendUp offers online lending as an alternative to payday loans, while United Income has an automated system, supplemented by advice, for managing retirement finances.

One promising innovation we explored is new underwriting tools that have the potential to be more inclusive. The panel noted that risk assessment systems grounded in credit scores work great today — for consumers who have high scores. For others, they don’t, including for the millions of borrowers who have no score or “thin files.” Traditional systems also screen out many people who have low scores despite being creditworthy, due to situations that are complex and can’t be easily analyzed in an automated review. In today’s economy, lenders are able to look at alternative kinds of data and analyze it in new ways with tools like machine-learning that can conduct more granular screening. There is particular interest in “cash-flow underwriting,” in which consumers give lenders permission to do an automated review of their bank account data, to assess whether income sources are sufficient and expenses are under control. This method is being used widely today by fintech lenders who believe it provides a lens for looking more closely at applicants they otherwise would have to turn down. A new nonprofit, FinRegLab, is conducting a study to assess this potential.

The panel also talked about technology driving down the cost of producing and delivering financial services, which can enhance affordability. In particular, mobile phones are democratizing finance worldwide. The World Bank has set a goal of one hundred percent financial inclusion by 2020, because cell phones are nearly ubiquitous and can now put financial access into everyone’s hands. Boosting that potential is the fact that these currents of technology change are converging with equally powerful demographic shifts, as millennials become the largest generation in history and as rising numbers of consumers take a mobile-first stance in most things, including finance. The mobile phone can also make it easy and instant for people to get information about financial questions and options, including to compare product features, pricing and ratings.

So yes, maybe some consumer dreams will be fulfilled by tech. However, the panel also looked at a range of potential nightmares. Like all forms of innovation, fintech solves some problems, and creates others. Panelists urged evaluating fintech under the same criteria used with traditional providers, such as interest rates under 36%, transparent pricing, and disparate impacts.

One of the dangers we discussed was erosion of consumer privacy as more and more data about people is gathered, stored, shared and leveraged for a range of purposes. Related to that is a weakening of cybersecurity, including the harm done by data breaches. Some of the technologies that are causing these problems may also help solve them — the “cloud” is more secure than on-premise mainframes, better encryption is emerging to protect data, and we will gradually move beyond passwords, which are generally the weakest link in the security chain. Still, there is extensive concern about how sensitive data may be accessed and used or misused, legally and illegally, by companies, governments, hackers and thieves.

Another key issue is whether new data and analytics will be fair, including from a discrimination standpoint. Machine learning algorithms can create “black box” situations where systems may appear to be producing sound outcomes, but humans don’t understand how and so can’t fully trust the results. While AI-driven underwriting could be much more inclusive, policymakers will have to worry about how they were built. Is algorithmic pattern analysis based on a sufficiently large set of data? The right data? Was the algorithm “trained” on data that mimicked human decision-making that was already be biased? Should AI have to be “explainable,” even if less explainable versions demonstrably expand financial inclusion and diversity? Do we need a data code of ethics?

We also talked about the challenges arising from the mold-breaking nature of many fintech innovations, which may not fit neatly into traditional consumer protection frameworks. There is risk that nontraditional advisory services could fail to uphold fiduciary requirements, which obligate investment advisers to act in their client’s best interests.There is risk that instant payments, while potentially helping people who now rely on expensive cash services for bill-paying, could also leave consumers unprotected if they make a payment by mistake. There are questions on how to disclose the reasons for adverse actions on loan decisions, if the reasons don’t tick traditional boxes. There are doubts, too, about the efficacy of regulatory disclosure mandates in helping consumers navigate this fast-changing landscape.

The panel also discussed concerns about “regulatory sandboxes,” an innovation being introduced by regulators themselves to conduct hands-on testing of new products and services. These sandboxes can be very valuable in facilitating rapid learning by regulators and early feedback for innovators. On the other hand, they could do damage if they waive consumer protection rules and are not properly supervised.

With so much upside opportunity entangled with so much downside risk, our discussion highlighted the need for advocates, policymakers, innovators, and industry to elevate technology change as a central policy issue in consumer financial protection. We will need much more study and a great deal more dialogue about how to secure the good while avoiding the bad. In areas where good and bad may prove to be inseparable, we will have to find the best possible balance between them.

Jo Ann Barefoot is CEO of Barefoot Innovation Group and a Senior Advisor to the Omidyar Network. She previously was Deputy Comptroller of the Currency and served on the staff of the U.S. Senate Banking Committee. Jo Ann is a cofounder of Hummingbird Regtech and has been a Senior Fellow at the Harvard Kennedy School Center for Business and Government, working on a book and series of papers on innovation and regulation in consumer finance. She writes and speaks on these issues throughout the world.