By Mrunal Gangrade, Vice President of Software Engineering,J.P. Morgan Chase & Company
Artificial intelligence is changing customer experience faster than most companies expected. Businesses are moving past simple chatbots and scripted automation. AI can now work out what a customer wants, look at their history, suggest what to do next and in some cases act for them. That is a huge opportunity. It also changes how we think about security.
For years, customer experience and cybersecurity were kept apart. The experience teams cared about speed and convenience. The security teams cared about protecting systems, identities and data. AI is pulling those two worlds together. The more an AI system can do, the more we need to know what it can reach, what it can decide and what it is allowed to do.
Customer Experience Is Becoming More Autonomous
Old style customer service waits for a problem. The customer runs into trouble, gets in touch and waits for help. AI makes something very different possible.
A bank might spot an odd activity and warn the customer before fraud spreads. An airline might see that a delay will break a connection and offer another flight. A phone company might notice weak service and fix it before anyone calls. These moments save the customer effort, and over time they build loyalty.
The next step goes further. AI assistants are starting to do tasks, not just answer questions. A customer might ask a banking assistant to check recent payments and find repeating subscriptions, then ask to cancel one or challenge a charge they do not recognize. Now the AI is not just giving information. It is taking action. From a security point of view, that changes everything.
Identity Becomes a Core Control
When someone logs into a company system, we want to know who they are and what they are allowed to do. AI adds one more layer. Now we also have to know which AI is acting, whose permission it is using, what it can see and how long that access should last.
An AI assistant should not get every permission the person using it has. If a customer asks it to look at their account, it may need short term permission to read their transactions. If that customer then asks it to move money or change personal details, the system should require a higher level of approval.
This is where (the concept of?) least privilege matters. Give the AI only the access it needs for one job and take that access away once the job is done. As companies add more AI agents and connect them to more systems, this becomes a must, not a nice to have.
Personalization Has a Security Cost
Companies want AI to know their customers well, because better context usually means a better experience. That requires data. An AI system might pull in past payments, earlier chats, preferences, browsing history and service records to make an interaction feel personal.
For the customer experience, that is useful. For security, it means a lot of sensitive information gathered in one place. So, the question is not only whether the data is there. It is whether the AI truly needs that information to do what the customer asked. More data does not always mean a better experience. Often the safer and more responsible choice is to use only what the task requires.
AI Creates New Attack Paths
Generative AI also brings risks that older software was never built for. Prompt injection is the clearest one. These systems read customer messages, documents, web pages and internal knowledge bases, and someone can hide harmful instructions inside that content to change what the AI does. The risk grows the moment the AI can use tools or take action.
Companies need a firm line between what the AI is allowed to read and what it is allowed to obey. And security cannot stop at the model itself. The Application Programming Interfaces (APIs) it calls, the systems it links to, the credentials it holds and the data it pulls in are all part of what can be attacked.
This is why AI driven customer experience cannot be one team’s job. Cybersecurity, identity, privacy, risk, data governance and customer experience all need to be involved from the start.
Governance Can Enable Innovation
Governance is often seen as the thing that slows new ideas down. In real life it does the opposite. When a company knows which AI took an action, which customer allowed it, what data it used and what controls were in place, it can grow those tools with far more confidence.
Being able to audit matters most here. For any important interaction, a company should be able to piece together what happened. Who started the request. Which AI handled it. What information it saw. What permission it was given. What it actually did. When a customer complains, “the AI decided” cannot be the answer. A person still has to be accountable.
Trust Will Be the Differentiator
AI will keep getting easier to use. Soon most large companies will offer smart assistants and AI powered automation, so simply having AI will not make anyone stand out. Trust will.
Customers may never see how an assistant is built, but they can feel whether the experience is safe, accurate and respectful. The companies that lead in AI driven customer experience will not be the ones that give AI the most freedom. They will be the ones that know how much freedom to give it, and can show that the right identity, security and governance controls are already in place.
AI can make customer experiences faster and smarter. Security and governance are what will allow a company to grow those experiences without losing the thing that matters most, which is customer trust.
Mrunal Gangrade is a Vice President of Engineering specializing in AI, explainable AI, and cybersecurity governance in regulated industries. An IEEE-published researcher and Gold Winner at the Women in Tech Global Awards 2025, she has spoken at events across multiple countries on building trustworthy, transparent AI systems that deliver measurable business value.
