Why I Joined SupportLogic: My Raison d être
We’re almost midway into 2021 and we have all seen several examples of data-driven decision making as a way to 1. Answer previously “unanswerable” questions, 2. Remove bias and be more objective in terms of answering questions, 3. Be more proactive instead of reactive (via forecasting, predictive analytics). Over time, these microscopic improvements in decision making are key to improving one’s “game” and moving further along the “excellence” vector.
Our digital lives (professional and personal) have increasingly led to:
- Large volumes of data being generated. (see Table-1)
- “Polyglot persistence” has become a reality over the past 2 decades.With growth of NoSQL, NewSQL, Hadoop, Data Lakes/Lambda/Kappa/CDW/Streaming/Batch architectures and a whole host of technology innovation, we can actually handle such volumes at acceptable cost curves (this was quite difficult, back when RDBMS was the only game in town).
- Cloud platforms have also realized the promise of utility computing, via IaaS, PaaS, SaaS so as to allow reasonable economics and improved velocity.
The other thing to note is the volume of unstructured data has greatly surpassed structured data (by orders of magnitude). Enter ML/NLP/AI innovation to process this (via supervised/unsupervised methods) and maximize Signal to Noise ratios. Since otherwise, this information overload cannot be dealt with by humans constrained by their existing cognitive capabilities.
I have spent much time geeking out on these cutting edge innovations via my various gigs in platform teams at high scale e-commerce companies (Walmart.com – with large number of concurrent users, transactions, data and Shutterfly – with large volumes of photos/image processing) and via my time at Terracotta (IMDG – in memory data grids and its usage) and at Slamdata (now Precog – 0-code transformations from raw data to analytics). However, increasingly, it felt like these were specific software-infrastructure plays that solved disjointed individual pieces of the overall problem.
What has increasingly fascinated me over the last decade is the problem of how to stitch all these infrastructure elements to really make data-driven decision making a reality for numerous personas within the enterprise.
How does one “main street”, “geek street” cutting-edge technologies? (Well, if finance people can have Wall Street – why not ;-)). As Andrew Ng has said: “AI is the new electricity” – and the question now is one of retrofitting every workplace, to now use the “new electricity”. Hence my time at Wiser Solutions (which allows e-commerce merchants to make decisions vis-à-vis commerce-catalog composition, pricing and promotion) and at Levadata (which helped bring cognitive dashboards, workflows and recommendations to the Procurement function in High Tech to manage supply chain risk and opportunities around cost savings). With SupportLogic I could see the same opportunity to bring such capability to the Customer Support and Success functions and usher in the CX revolution, on the back of our new fangled technological capabilities, alluded to earlier. As a matter of fact, this applies to other “bowling pins” in close proximity, to borrow Mckinsey/”Crossing the Chasm” phraseology, all of which speaks to large TAMs (Total Addressable markets).
While sales- and marketing-led growth were definitely the “stylish” vectors of growth and have seen significant VC investments, there is something refreshingly honest about good, old virtues of doing right by your customer – actually delivering real value when the rubber meets the road – thus running a great business with low to no churn (SaaS especially) and concomitant product-led growth (PLG).
So, throw out the old handbook, which thought of these as cost centers and embrace the new, where these functions really represent the voice of the customer (which often gets lost, given how siloed enterprise information systems are). This CX focus should, via product-led growth, bring down the cost of acquisition per dollar of ACV/revenue. Being a physics aficionado, I could not resist an analogy based off of the famous double-slit experiment 😉
Being a product guy, the product-based elements for success in such an endeavor seem similar at a high level (across verticals) as explained in this figure – and typically needs:
1. Very strong domain expertise:
- What questions need to be answered by which persona and what decisions are key?
- What’s the best way to consume these (i.e. what recommendations would be valuable) and what workflows would benefit and deliver business value ?
- What are the data sources and what transformations are needed to reduce raw data entropy and maximize SnR? How does one apply constraints of data perishability?
I glimpsed oodles of this expertise for the support/success function in Krishna’s head (Founder/CEO) and several of the staff here at SupportLogic.
2. Strong Data Pipeline & ML/NLP Augmentation:
- This is about a scalable, extensible pipeline that can support the classic 4Vs of data.
- Volume/VelocityVariety/Veracity: Large volumes of CRM and related data. While Batch ingestion could work, the goal at Support Logic is to be as quasi-real time as possible with intelligence embedded in all interactions. While structured feeds from CRMs are one source (most of it is unstructured text), semi structured and unstructured sources such as Slack messages, text transcripts of Zoom, Twitter messages and the plethora of channels via which modern-day customer interaction occurs all need to be analyzed. These extracted signals need cross-correlation and validation to improve accuracy.
- 11 patents speak to the strength of the team in terms of being able to handle unstructured text via NLP and ML models and these are just getting better with each day as the diversity and amount of training data is increasing via GTM growth.
The SupportLogic roadmap has already realized some of these and the Roadmap is visionary enough to differentiate “interactions” as the atomic unit (with CRM cases being just one modality of such as an interaction).
3. Embedding this intelligence via AI-Based BI & within workflows:
- Yes dashboards and Drill downs help, but increasingly “Dashboard fatigue” is setting in amongst decision makers. The experience end-users are hankering for are:
- Recommendations on what the ideal course of action is. Recommendations on what to focus on given the information overload (e.g. cases with a risk of escalation or which sorely need attention). And workflows with intelligence embedded. Along with a layer of explainability and drill down (for verification and trust).
- i.e. There is a strong need to automate the “cognition” and not leave it as an exercise to the end-user to understand several dashboards and connect the dots.
The architecture and design within SupportLogic products support this vision. A lot has been achieved and more needs to be done to deliver this promise of “post-CRM” functionality of turning around your efforts to be proactive instead of reactive (and being very metric driven about this) and of using the natural language and flow to maintain metadata (i.e. stop being a slave to demands of structured systems and instead free users to operate naturally). This has already led to exciting GTM traction with impressive customer logos (with real success stories of getting churn down and customer satisfaction boosts). The fact that not a single SupportLogic customer has churned thus far (knock on wood), speaks to that refreshingly honest virtue we earlier referred to (of providing value and deeply caring about your customers).
All of this meant for me that this is a great place to give that ambition of using AI to “electrify the workplace” further wings and hence, I am proud to be part of this effort. You can learn more here.