Like most industries, artificial intelligence (AI) is changing the way we work, live, and go about our day-to-day lives, and identity governance is no exception. By taking a once manual, error-prone process and adding the gloss of automation and accuracy, we can remedy age-old problems with managing access and entitlements. But we must remember what age-old wisdom tells us: if it sounds too good to be true, it probably is. And many AI-enabled products over promise and under deliver—at least for now.
While there’s no shortage of AI solutions on the market, many look at such a finite set of values that they’re simply not able to live up to their hype. Put simply, a lot more data needs to be available, and most one-off solutions don’t have access to this type of data. Additionally, the narrow view of identity as strictly a function of security has also limited what is possible with AI. Identity touches every part of a business, from IT to HR and everywhere in between.
If identity is working in a security and compliance silo, enterprises will not realize the true value streamlined, AI-driven identity governance can bring. And the outcome of any AI initiative should be a more agile enterprise. This includes activities like improved workflow, seamless user experience (UX), and improved operations. When done right, this is exactly what intelligent identity management can achieve.
But how do we get there? Let’s take a closer look at where we stand and how we can get our identity data to start working for us.
While popularity of AI continues to grow, practices and maturity have remained relatively stagnant. Industry research shows lack of skilled people and difficulty hiring topped the list of challenges in AI. Pair this with the 25% of companies that have seen half of their AI projects fail (IDC) and it’s no wonder why we haven’t yet seen truly successful AI applications around identity.
If we drill down to identity-related AI projects specifically, the data tells another grim story. A survey from Gradient Flow shows that two-thirds of respondents indicated that their company uses AI/machine learning (ML) to improve identity management. Yet less than a third of respondents indicated that AI/ML yielded moderate to high benefits for identity management. It’s likely that far less than two-thirds of respondents are using AI in a real production environment. Even still, it seems that there is more perceived than actual value.
The problems are clear: we don’t have enough data, we see identity on an island of its own, and we lack the skilled technical talent to make it sing. As such, here’s where business leaders should focus and start approaching identity backed by real deep learning technology.
- Volume of Data: AI/ML can find patterns and extract value in vast amounts of data with a sophistication few technologies can match. That said, both AI/ML algorithms need massive amounts of data to understand what is normal and what is anomalous behavior. Many data inputs are needed to train and test the algorithms, then, once validated and put into production, there needs to be continuous amounts of data feeding the algorithms to remain accurate.
- Specialization: It takes an expert to determine what data should be part of an AI/ML initiative. Bypassing this level of specialization can result in limited insights. A data scientist will be able to advise and customize the algorithms for your business’ specific use cases. While low- and no-code solutions are growing in popularity and sophistication, it’s still important to have AI talent close by to ensure accuracy, consistent training and tuning of models, and to avoid model degradation over time.
- Integrating Identity Organization-Wide: By leveraging a business platform, organizations gain access to an entire data warehouse with information about not just identity controls, but IT Service Management (ITSM), Security Operations (SecOps), HR, and more, along with all the related service requests and approvals. There’s no need to perform multiple bulk exports and imports from different products or systems across your enterprise—it all lives in the same place. The best news? This functionality already exists within your existing tech investments in platforms like ServiceNow, Salesforce, Azure, and more.
We still have a long way to go before AI and identity are working together seamlessly, but we’re on the right track. By taking stock of the data you have access to, organizational silos, and prioritizing AI talent—whether in-house or by way of the partners and products your business uses—these should be the priority areas for those seeking to maximize their AI efforts around identity.
A version of this article first appeared in insideBIGDATA.