Machine learning features have the potential to transform your marketing and customer support, but they\u2019ve been out of reach for most companies. Now both Microsoft and Salesforce are building them into their cloud CRM tools, putting AI within the grasp of most enterprises.\n\nThe idea is to democratize the benefits of machine learning for businesses that don\u2019t have data science expertise. \u201cFor the vast majority of companies, it\u2019s too hard,\u201d Einstein general manager John Ball tells CIO.com.\n\n\u201cTo do data science, you have to have data, you have to collect and manage it and then you have to do some data wrangling. Then you have to hire a scarce data scientist, you have to build predictive models, you have to refresh them,\u201d Ball says. \u201cYou have to have infrastructure that\u2019s trusted and secure to run those models on and you have to maintain it. And then you have to actually take the recommendations and put those in the context of business users.\u201d\n\n[ Also on CIO.com: Essential CRM software features: A savvy buyer's guide ]\n\nPutting AI tools into the CRM makes them immediately useful. There are tools that can predict which leads will turn into opportunities and which opportunities will close and become customers. And there are tools that can offer suggestions to sales reps and marketers, from nudging them to follow up on opportunities they\u2019ve been neglecting, to detecting when a competitor is mentioned, to predicting when a customer will open an email or unsubscribe from a marketing campaign, to helping you craft an email that will appeal to the person you need to reach.\n\nSome of this will be fairly general. For example, when Salesforce does lead scoring, Ball explains, \u201cif an email is @yahoo.com then it\u2019s probably not a good lead.\u201d Some of is much more specific: The British brewery Marstons is using Customer Insights in Dynamics 365 to put customer details, from their favorite beers to whether they have unused vouchers, drawn from Facebook and TripAdvisor as well as its CRM on the handheld devices of pub staff. With that personalized information, the brewer is hoping to sell an extra meal and drink a day in each of its 550 venues.\n\nAI features compared\n\nAt first glance, the AI features in Salesforce and Dynamics are similar. Dynamics builds intelligence into the standard workflow, while Salesforce has specific tools in its different offerings.\n\nSalesforce\u2019s Sales Cloud has Predictive Lead Scoring, designed to suggest which leads are most likely to convert; you could use that to route better opportunities to more experienced sales staff. Opportunity Insights alert you when a deal is looking stronger or weaker, spot mentions of competitors in the conversation that might mean the deal is under threat, and even suggest if you\u2019re talking to someone who won\u2019t have the authority to close the deal.\n\nPredictive Scoring in Marketing Cloud tells you how likely it is that a customer will engage with the email you\u2019re sending them, coaches you on what language to use and even tries to send the message at the time recipients are most likely to look at it. Predictive Audiences will divide potential customers up into audience segments based on their predicted behavior, like disappointed customers you can win back, selective subscribers who look for specific products, loyalists you can rely on and window shoppers you need to appeal to directly.\n\n[ Also on CIO.com: Visualization analytics helps utility provider escape \u2018Excel hell\u2019 ]\n\nRelationship Insights in Dynamics 365 offers many of the same features, automatically capturing information from emails you send and receive in Office 365 to fill in opportunities, doing lead scoring and sentiment analysis and showing you the health of relationships and the strength of opportunities, as well as risks to deals like competitors showing up in the conversation, and suggesting the next best action. Bubble charts show the health of your relationships with customers as well as how large the customers are, to help you prioritize. You get alerts for customers who want to meet you as well as for opportunities that are going stale, among other smart notifications.\n\nAs you work in Dynamics, it fills in fields automatically with predicted information, like forecasting demand and inventory \u2014 and even suggesting when you should order more inventory and which other products you should restock at the same time \u2014 and suggesting products to cross sell and upsell\n\nSalesforce Commerce Cloud will create the kind of personalized product recommendations large ecommerce sites have; it can also personalize the order products are shown in and even what shows up in searches. Commerce Insights will suggest correlations to explain purchases, including the channel you reach people on, whether they buy after opening a specific email or even their email domain.\n\nDynamics 365 also uses machine learning to suggest products based on wish lists, click patterns and past orders. It can also personalize for each customer the information that's shown about a product.\n\nSalesforce Service Cloud Recommended Case Classification automatically fills in key fields so cases can be routed to the right staff. Recommended Responses highlight the information service staff should use when responding, and Predictive Close Times will predict how long it will take to resolve an issue. Similarly, Community Cloud will recommend specific experts, articles and topics, and if a post doesn\u2019t get a response, Automated Service Escalation automatically creates a support incident. That\u2019s similar to Dynamics\u2019 pre-emptive service feature, and there are scenarios in Microsoft\u2019s Azure ML service to built models that suggest relevant Knowledge Base articles and case studies.\n\n[ Also on CIO.com: 5 Dreamforce takeaways to guide your CRM strategy ]\n\nAnd the Einstein features in Analytics Cloud are broadly similar to the Customer Insights tool in Dynamics 365. Both let you pick measures and KPIs, like cost and profitability, and see correlations and insights. The Salesforce tools produce Word and PowerPoint documents of the customer journey; Customer Insights uses Power BI visualizations to show KPIs, insights and actions \u2014 and you can integrate them into custom apps like the one Marstons built for its staff.\n\nWhen can you get these features?\n\nSalesforce focused heavily on the Einstein AI features in the Customer Success Platform at its October Dreamforce conference, but it had actually launched them a month earlier. That doesn\u2019t mean you can actually use them right now; some are ready, some are still in development and they nearly all come out of Salesforces\u2019 recent acquisition spree in this area.\n\n\u201cSome Einstein solutions, like the recommendation engine based on technology from Demandware (acquired in June) are currently or will be soon available. So are machine learning-based data discovery and analysis services leveraging technologies from DemandCore (acquired in September),\u201d Charles King, principal analyst at Pund-IT, tells CIO.com. \u201cBut it is also clear that more advanced services and solutions won't be available until Salesforce more thoroughly integrates those acquired technologies, a process that could take several months.\u201d \n\nAutomated Community Case Escalation and Recommended Experts, Files and Groups are available now in Community Cloud Einstein, and they\u2019re included in the Community Cloud license. Product Recommendations in Commerce Cloud Einstein is also ready to use and is included in the Commerce Cloud licence. Smart Data Discovery (based on another acquisition, BeyondCore) in Analytics Cloud Einstein is also available now, but you pay extra (by the volume of data and number of users), and you need at least 10,000 rows of data to get reliable results so this is definitely intended for larger businesses.\n\nThe PredictionIO server is already available, and it\u2019s free, although you\u2019ll need far more machine learning expertise to build tools on that. Similarly, you can sign up for the Predictive Vision and Sentiments developer services, which are in beta (Salesforce acquired MetaMind in April), but while it\u2019s easy to drag and drop images into a web page to train your own deep learning model, selecting the right training set and evaluating how good the model is for the problems you want it to solve takes more machine learning experience.\n\n[ Also on CIO.com: CIO's move to chief customer officer role signals trend ]\n\nThe Einstein features in Sales Cloud, like Predictive Lead Scoring and Opportunity Insights, and Marketing Cloud\u2019s predictive email content (pricing based on the size of your mailing list) will be the next to show up. These are being built by Salesforce using technologies that came with its acquisitions of ExactTarget and Heroku. Commerce Cloud will get Commerce Insights soon, and Predictive Sort for products will come sometime in 2017. Also not yet available are Predictive Wave Apps in Analytics Cloud, the Service Cloud predictive tools, and IoT Cloud Predictive Device Scoring, Recommend Best Next Actions, and Automated IoT Rules Optimization.\n\nSalesforce is pushing Einstein as a platform service layer in its Customer Success Platform. So far, it\u2019s a discrete set of useful tools, only some of which are ready to use.\n\nIn contrast, most of Microsoft\u2019s intelligent features have either been available for some time or are ready for the November launch of Dynamics 365, King points out. \u201cMicrosoft's Dynamics 365 with AI is the result of the company unifying its ERP and CRM solutions, an effort announced in the summer of 2015. The Customer Insights cloud service is notable in that it allows customers to combine data from internal and external sources (including Microsoft partners Facebook and TripAdvisor). It can be used as a standalone service but also works with any external CRM tool with open APIs. Another solution \u2014 Relationship Insights \u2014 is built on the Cortana Intelligence Suite, which includes a recommendation engine, sentiment analysis and even facial recognition features.\u201d\n\nSmart notifications through the Cortana assistant are currently in preview, as long as Cortana is available in your country, initially in English (support for other languages is in development).\n\nIf you want to do predictive maintenance or remote monitoring, Microsoft has had preconfigured solutions for those in its Azure IoT Suite and Cortana Intelligence Suitefor over a year (the latter was initially called the Cortana Analytics Suite), and you can extend and customize those. Dynamics 365 includes IoT support for field service, and the predictive and prescriptive maintenance features will soon be incorporated as well.\n\nThe Cortana Intelligence Suite includes some of the 22 developer APIs already available in Microsoft Cognitive Services, like the voice recognition and natural language understanding APIs that fast food chain McDonald\u2019s is using to turn what customers say at the drive-through window into an order. There are already partner apps based on these services, like Veripark\u2019s tool for suggesting your next best action, available in Microsoft\u2019s App Source store.\n\nAnd if you want to build your own machine learning systems to extend what you do with Dynamics, the Azure Machine Learning Studio is a drag and drop environment with dozens of machine learning modules and solutions for everything from fraud detection to sensor data analytics, so you can start from scratch or adapt existing tools. It might seem a long way from CRM, but when you think of it as part of a digital business, it looks like more of a continuum. And that might give Microsoft the advantage in the long run.\n\nFor example, Microsoft is currently has in trial an intelligent customer support agent that uses machine learning. The support bot will suggest Knowledge Base articles that might fix the problem a customer describes in the chat. If that doesn\u2019t solve the problem, the customer can type in more details or ask to be handed over to one of the 2,000 human agents in the pilot. The human agent gets a summary of the session so the customer doesn\u2019t have to enter all the details again, and the support bot also learns from the solution the human agent suggests. Microsoft isn\u2019t making that system available as a product \u2014 yet \u2014 but it\u2019s something that would be a great tool for the Dynamics 365 service in the future.\n\n\u201cAt this point, Salesforce seems to be taking a broader and more granular approach to CRM processes and functionalities,\u201d King says. \u201cBut, while ambitious, its Einstein platform and services will take time to mature and achieve the full range of features the company envisions. Microsoft's Dynamics 365 with AI is the result of a multi-year process, and appears to be the more mature platform and solution set. More importantly, Dynamics 365 with AI is just one of what will eventually become a host of Microsoft AI-based solutions and services designed to enhance business processes. Salesforce's Einstein may eventually deliver a deep view and reach into CRM, but Microsoft is taking a far broader view of the implications and opportunities for using AI technologies in business.\u201d \n\nMicrosoft touts its cheaper pricing for Dynamics 365: The E1 plan costs $70 per month per user and includes customer and field service, project service automation, knowledge management and social engagement as well as sales. Salesforce puts those into different cloud services in Sales Cloud and Service Cloud and there are extras like Field Service, all of which cost $135 to $150 per user per month.\n\nGet smart\n\nAI and machine learning for CRM work on the same data you already have access to, but automating insights from your data can speed up responses and make sure you don\u2019t miss opportunities. A recent survey of financial institutions by LexisNexis found that about half of them turn away between six and 15 percent of potential customers (individuals and small businesses) because of the processes they use to manage credit risks and onboard new customers. Few businesses really have \u201ccustomer journeys\u201d in their sales and marketing processes today, so there\u2019s a lot of low hanging fruit that you can improve with smart tools.\n\nAnd there\u2019s a lot of potential for growth. When Accenture predicts that AI could double annual economic growth rates by 2035, it\u2019s talking as much about supporting existing employees and making them more efficient as it is about replacing them. \u201cAI can enable humans to focus on parts of their role that add the most value,\u201d Accenture said in a report on the growth of AI. \u201cAI augments labor by complementing human capabilities, offering employees new tools to enhance their natural intelligence.\u201d\n\nWhichever cloud CRM you use, the intelligent tools that Microsoft and Salesforce are adding aren\u2019t going to replace your customer support or marketing teams, but they might make them a lot more effective. And as AI is going to be a standard part of both services, you\u2019ll want to look at where it will be useful to you, because your competitors will be using it too.