Artificial Intelligence Will Never Transform Business Process Unless We First Fix CRM Data
This was originally published on LinkedIn.
At Dreamforce, Artificial Intelligence or “AI” was the buzzword of the year. We saw the launch of Salesforce Einstein and many other companies refining their messaging towards how AI will transform sales, marketing, and even customer success process. The vision behind this is bold and Einstein, if it delivers, will transform how people engage with customers. Yet I’m concerned about the viability of these products at scale: most CRM and marketing automation data isn’t high quality enough to deliver on the promises of AI. We’re all focused on how to make our applications smarter, but few, if any companies, have data governance policies with regards to how sales reps input data or how marketers capture inbound customer data. To make matters worse, it’s unlikely companies have a sense for how poor their prospect and customer data actually is because they’re comparing it to incumbent data providers like Dun & Bradstreet. These old-world vendors lack the quality and breadth to accurately determine where the gaps exist in customers’ CRM data.
Every sales operations team we encounter struggles with data quality and there are typically two current solutions: enterprises either (a) hire Acxiom, Accenture, Deloitte, or the like to remove duplicates, clean up data, refresh common elements or (b) ignore the problem and just buy more data. These two solutions end up being a disaster at scale — in fact I’ve seen enterprises with as many as five duplicates per Account stored in their Salesforce instance and sales activity that is duplicative across multiple opportunities.
If we don’t solve this fundamental data foundation problem, AI will fail to transform business process. It is impossible to predict outcomes or even make great decisions off predictions if the underlying data training the models is stale, inaccurate, or incomplete. Driving insight means we need to truly have a sense of our underlying data and tie those truths to our business decisions. The fundamentals of AI are predicated on learning and similar to humans, if we don’t have strong foundational curriculum in school, we end up making uninformed and unreasonable decisions when we enter the working world.
Many enterprise software companies entering the AI space seem to believe that hiring machine learning experts and more data scientists is the proper remedy for this problem and while that’s maybe half the battle, those experts will struggle with seeing performance from their work if the training and foundation data isn’t accurate. Microsoft is one of the few companies that understands — hence their $26B acquisition of LinkedIn.
The parallels here are strikingly similar to the creation of FICO scores back in the middle of the last century. Banks, financial institutions, and mortgage lenders needed accurate information to underwrite loans and as a result, the banks agreed to share credit history of potential customers through credit bureaus like FICO, Experian, Equifax, etc. These bureaus became the foundational truth and the intelligence the banks needed to determine whether a consumer qualified for a loan. In the case of driving intelligence for business process, we also need a standard or a “directory of truth” for basing predictions. This foundational data problem is however a significantly larger technical challenge as it requires billions of inputs across millions of accounts and contacts.
Our hope at Radius is that AI won’t die with the belief that algorithms and machine learning doesn’t work because of poor data quality. There are mass efficiencies and improvements AI can drive for companies at scale, but the sentiment will wane if we the underlying data in CRM isn’t improved quickly. For every customer we encounter at Radius, management teams believe they need AI today when realistically it’s a much more in-the-weeds data integrity solution they are in need of before predictive or AI can drive real value. In fact, the best algorithms alone won’t win, the best data plus the best algorithms will win.
How are we solving this foundational data problem? First, we focused on building a CRM data consortium for the benefit of everyone and leverage machine learning to make sense of the billions of customer-contributed inputs we have each day. At Dreamforce we announced that 99% of our customers contribute anonymized and aggregated data from their own CRMs to improve our core Radius Business Graph which benefits all of our connected customers. As we continue to get more contributions, all customers will benefit with exponential increases in data accuracy and AI effectiveness. It’s one of the reasons why we plan to expand our integrations to include Microsoft Dynamics CRM, Adobe Audience Manager, and many others that will allow customers to leverage the truth of our data to drive true intelligence and accurate predictions for their business.
The vision we all hold for how AI can transform how we work can only be achieved if the underlying data is accurate and fresh. Leveraging network effects, integrations into the platforms companies use, and building a true consortium data network for CRM will enable us to realize this vision.
A Product Executive’s Perspective On Unlocking B2B Data’s Latent Potential
Is Predictive Marketing Now Mainstream?
How To Elevate B2B Data Accuracy With Customer Network Effects