Salesforce and the Machine (Learning Revolution)
Machine learning is here to stay and Salesforce is poised to ride the nascent revolution of analytics into the future and beyond…if they can adapt their successes to other parts of their growth strategy.
The megatrend of machine learning is critical to Salesforce and its ability to compete in the sales and customer service space due to a variety of reasons that ultimately link back to the growth strategy of the company. The bulk of Salesforce’s growth strategy is focused on extending, expanding, and improving on its current positioning in various aspects of the business. Cross selling and upselling, extending existing service offerings, reducing attrition, strengthening their service offerings, and improving go to market capabilities are all integral to the future growth of the company and Salesforce has correctly identified an effective way to execute that vision via the widespread adoption of machine learning processes across their array of products [1]. The ability to improve and personalize their products and services, develop proprietary AI processes and metrics, enhance their customer service capabilities, and utilize this type of analysis is a key skill moving forward in the ecommerce and ecommerce support space where a culture of self-improvement is paramount to staying relevant.
In the short term, Salesforce has already released Einstein, an integrated set of AI technologies, and worked diligently to establish this flagship program as central to the Customer Relationship Management software it sells. It works to analyze and adapt to client behavior over time to provide a better sales platform experience. Additionally, Salesforce has also released many early betas of different extensions of the Einstein platform like Einstein Sentiment, which works to classify the “tone” of text communications like posts or comments to better understand feedback, Einstein Intent, which classifies inbound customer support queries to assist with customer service experiences, and Einstein Object Detection, which helps create applications capable of recognizing images and objects which could aid in the restocking of store shelves [2]. In the medium term, Salesforce is working to extend their vision by continuously improving their ecommerce cloud via machine learning to unify cross channel operations. Salesforce believes that 20% of future jobs will be within the Salesforce economy and platform and in order to establish that dominance, they are figuring out betters ways to embed machine learning algorithms in their processes to personalize the shopping experience for their client’s customers [3]. Additionally, they are capitalizing on their current momentum by extending long term relationships with key opinion leaders like Dell to overhaul their customer interactions and make them more predictive. This also has the added benefit of creating multi-channel experiences for Dell and emphasizes the inherent strengths of Salesforce’s architecture [4].
To further improve upon this success, I recommend Salesforce constantly reassess the ability to extend their newfound machine learning strengths into the other elements of their growth strategy. The solutions pursued currently all make good business sense, but in a constantly changing competitive environment new opportunities can arise quickly and Salesforce needs to be ready to execute immediately. Targeting specific vertical industries is a natural way of pre-planning specific machine learning strategies. A solution for financial services or healthcare on the shelf, but “ready-to-go“, when the chance appears could be the difference between a secured or missed contract. These processes can also be extended to aiding not just Salesforce’s customers, but to their own business tactics as well. Reducing customer attrition is a key goal of Salesforce’s growth strategy and the proprietary machine learning processes can be applied to their own client base to increase that retention in a meaningful and organic way. Finally, the risk factor of AI solutions from startup and established companies is a point well taken by Salesforce [1]. They understand the danger that more effective algorithms pose to their platform and should adjust their M&A strategy accordingly. Without a consistent effort to find the best budding CRM technologies, acquire them, and integrate the best practices into the Salesforce environment, the company runs a massive risk of losing their standing in the long term. This also reinforces their strengths as they implement their growth strategy in the coming years.
In the context of Salesforce there are an array of further questions that merit consideration when discussing the future and relevance of the company. Most are very tactical about the specific methodology the organization should pursue while building these algorithmic processes, but the biggest two are broader in scope. First, what should Salesforce establish as the goal of their research departments moving forward-incremental improvements or great leaps forward? Second, can a bet on developing innovative predictive solutions be consistently successful for year after year?
(736 words)
[1] Salesforce FY18 Annual Report, https://s1.q4cdn.com/454432842/files/doc_financials/2018/Salesforce-FY18-Annual-Report.pdf
[2] Salesforce Extends Einstein Machine Learning Features for Developers, By: Needle, David, eWeek, 15306283, 6/28/2017, http://web.b.ebscohost.com.ezp-prod1.hul.harvard.edu/ehost/detail/detail?vid=7&sid=7e0c174a-9d6b-4f2f-88e9-b3a887b51ab4%40pdc-v-sessmgr01&bdata=JnNpdGU9ZWhvc3QtbGl2ZSZzY29wZT1zaXRl#AN=123921761&db=bth
[3] Salesforce Extends Commerce Cloud Einstein, By: Ghosh, Sudipto, MarTechSeries, 5/17/2017, https://martechseries.com/sales-marketing/marketing-clouds/salesforce-extends-commerce-cloud-einstein-delivering-personalized-ai-powered-shopping-experiences/
[4] Salesforce’s Einstein AI to Power Dell’s Customer Interactions, By: Shaikh, Nahida, MarTechSeries, 5/25/2017, https://martechseries.com/predictive-ai/ai-platforms-machine-learning/salesforces-einstein-ai-power-dells-customer-interactions/
This was a really interesting, especially considering how difficult it can be for Salesforce to use feedback in a meaningful way. It seems that the best way forward may be incremental – it is about customers being comfortable with the way they interact with the company, as well as how their data is being used. I would hypothesize that companies are more sensitive to the way their data is used than individuals tend to be. A really insightful article, thank you.
Thanks a lot for sharing this article – Salesforce is in many ways on the forefront of innovation in the CRM space and the way we interact with customers and reading about their efforts to incorporate ML technologies is an exciting one. Coming to your questions:
First, what should Salesforce establish as the goal of their research departments moving forward-incremental improvements or great leaps forward?
I believe these goals are not necessarily mutually exclusive – The goal-setting should be aspirational i.e. focus on significant revenue opportunities or product improvements, however, certain findings from a ML perspective might be incremental, others may change the product works in the backend entirely. I feel it is critical to give developers enough freedom to think out of the box, so encouraging to look for bigger solutions makes sense.
Second, can a bet on developing innovative predictive solutions be consistently successful for year after year?
No, I don’t believe that the bet will reward them every year – there will be years in which no significant improvements will be made and there will be years in which multiple significant improvements will be made. As long as the core business and the core operations of Salesforce provides sufficient cash to fund these experiments, there shouldn’t be an issue however!
Thanks again for sharing!
Very interested topic. Machine learning relies on data. Salesforce is in a good position to gather business data since their products are embedded in many businesses across different industries. However, the concern is whether Salesforce’s customers are open to share their data which most likely will contain sensitive information. I guess one way to circumvent the problem is to gather data by industry verticals (but not individual firms) and share those data as industry market intelligence with their customers. This initiative will also help Salesforce’s attract more customers since they will be interested in knowing the real time trend in their industry too.
To take a shot at your first question on whether Salesforce should focus on taking incremental steps or great leaps forward, I suggest that they focus primarily on incremental steps while also establishing a small team in their R&D division that is entirely dedicated to pursuing great leaps forward. While Salesforce is certainly in a dynamic space and faces increasing competition, they are still an enormous player with a well-established global infrastructure and proven results. Any innovation in AI that Salesforce makes, even if it is small by Salesforce’s standards, will have an enormous impact on the market due to Salesforce’s sheer size. To me, the added benefit of focusing entirely on great leaps forward is not worth the risk of failure, especially when Salesforce already has such a hold on the market. However, I do think that adding a small team focusing on large ideas will allow Salesforce to identify opportunities for great leaps without risking the incremental and steady progress of it’s AI advancements.
A well written and well researched essay. I worked in B2B sales before and have firsthand experience with Einstein. While the product holds great promises, there are two important challenges to consider: (1) prediction effectiveness depends on input data, and currently Einstein is built as a Salesforce extension and does not integrate rich reservoirs of customer data sitting outside of Salesforce. Because of this, Einstein’s prediction accuracy is somewhat limited. (2) As you rightly pointed out, many startups are working to tackle sales AI. Important to note is that many large tech companies
(Microsoft, Google..etc) already have developed internal sales intelligence & churn mitigation tools. Google, for example, used its internal product usage and engagement data to flag key at-risk customers. Einstein, as a general-purpose sales intelligence tool, cannot compete effective against internal tools built using internal/proprietary data specifically for parent companies. I think adding to your list of near-term and long-term strategy is the need to expand their Business Development effort — working with client companies to build a large set of APIs to effectively integrate non-Salesforce customer data (for example, product engagement data) into Einstein to enhance prediction.
Taking “leaps forward” will have to be the approach that Salesforce takes with its adoption and implementation of AI-driven solutions. Salesforce has been a market leader in the CRM world and this move will allow it to stay ahead of the curve, especially at a time when countless open-source CRM platforms are entering the market. CRM platforms still require heavy input from a data/tech team at the client-end to design dashboards and make sense of heaps of data; providing access to “smarter” analytics would become it’s key differentiation.