lessons_learned_ocean
Chris Heineken

Chris Heineken

Atrium: Year 1 Lessons Learned

We just closed our first year at Atrium with a goal of placing our Company on the map where the frontiers of Machine Learning and business connect. It’s an incredible place to observe. The convergence zone is characterized by big expectations, diverse terminology, new technologies and labor pools that have never collaborated closely. Getting mathematicians, programmers and business leaders to work in alignment around real world use cases can be challenging Over the last twelve months we conducted over 150 briefings / discovery sessions across the Machine Learning (ML) ecosystem with companies that span all industry verticals. Across all these discussions there is a universal desire. Demystify the unknown and get started on the journey.

Three trends emerged throughout our discussions which framed some big takeaways:

  1. Mathematics at scale is a future pathway for IT / business leaders who want to expand their influence
  2. Machine Learning is a multi-year journey, not a destination.
  3. Success will be defined by companies that employ strong enterprise coordination skills between the technology and business domains


Get Comfortable with Mathematics at Scale

Much is written about two important questions related to Machine Learning and technology trends. How is Information Technology as a profession staying relevant as technology is commoditized? Is the promise of Machine Learning worth the investment? Both of these questions came in sharp focus for Atrium as we worked with companies to construct roadmaps for their journeys. Firms like McKinsey have evaluated the contrast between the projected winners and losers to emerge from this trend (see below).

Our learnings this year bring us to the same conclusion as McKinsey. There is significant money on the table for companies in almost any setting.

In retrospective, the briefings, discovery and strategy workshops we facilitated consistently generated simple investment opportunities in Machine Learning that represented high impact outcomes for our clients. These investment options centered around using math at scale to help enterprises become more productive across common challenges like lead scoring, win rate conversion, forecasting, customer attrition and cross selling.

From a career growth perspective, expanding individual skills around ‘Math at Scale’ as a discipline represents unique career opportunities for those looking to improve their internal influence and span of control. One of the more consistent trends we observed was the continued relegation of technology professionals to more commoditized activities inside the enterprise. However, Machine Learning represents a pathway out of technology commoditization and back into the innovation space. Those placing themselves on technology investments with Machine Learning more often than not, find themselves with a seat at the executive table. One of my favorite quotes from Atrium’s ’Year 1’ comes from a highly successful executive regarding the changing power dynamic “the business teams used to give me hell [about wasting their time with IT systems and adoption challenges], now I give them hell for wasting my time [after deploying ML algorithms that isolate productivity issues to specific business teams needing to improve performance].” This is one of many good examples where the value equation changed by placing technologists on par with business stakeholders, ultimately resulting in partnership The powerbase inside the enterprise is evolving and those who can use concepts like ‘Moneyball’ with Machine Learning will drive the innovation investment agendas.

Machine Learning is a Journey

The concept around Machine Learning as a journey vs. destination shares many common traits from past software trends. CRM, Supply Chain, Analytics investments have always been about journeys instead of destinations as the line of demarcation between success and failure often hinged on an enterprise’s ability to evolve software at the speed of business. As cloud technologies emerged, the definition of successful evolution speed moved from yearly software updates to quarterly releases. Regardless of the software release interval, the mindset (even in Agile shops) was predicated on stability and precision. ‘Measure twice, cut once’ is often the mantra associated with software development

Machine Learning challenges these notions of stability and precision and if anything, pushes enterprises to evolve faster. Rather than ‘measure twice, cut once’ it is all about ‘cut, cut, cut.’ Once you have deployed your first predictive analytics model two things will happen. First, you will be in constant search for new features that make your model more accurate and second, you will learn new insights about your business that will invariably push your teams to consider new predictive models. There is no end. Rather, it is a race to improve business productivity based on continuous Machine Learning principles. The implication for business leaders is that you need to be ready to play the long game as it relates to incorporating ML development techniques into your innovation investments It is not about limited ‘point in time’ connections between data science, business and IT teams, but rather a constant state of evolution with a starting point that assumes sufficient preconditions vs. perfect preconditions for success.

Business : Math : Programming Coordination Skills Win the Day

Companies that break through the communication silos between business, data science and IT stakeholders are the most successful in taking advantage of Machine Learning. Frequently, the challenge becomes a coordination puzzle balancing clarity of business intent, practicality in building predictive models and ability to adapt existing technology platforms. Not easy given all the elements to balance. Some industry estimates put the success rate on predictive models being deployed into a production environment in the range of 10% to 30%. Most of the best conceived concepts go to die in the laboratory and never see the light of day. Why is this? The root of the problem breaks down to three simple concepts in vetting use cases that frequently are not followed.

  1. Do we know our business well enough to pose questions that will provide high impact if answered?
  2. Is the problem we are solving actionable and associated with significant data volumes where speed of response matters?
  3. Can these insights from #1 and #2 be surfaced in an organizational workflow that can be distributed in a broad sense.

Many Companies can nail points #1 and #2 but fall short on the coordination skills required to adapt systems around scaling these powerful insights into an organizational workflow. Companies looking to dramatically increase their success rates will need to employ ‘enterprise translators’ that can help span the complicated topics across math, programming and business concepts to bridge the difficult final mile of these programs.

Atrium: Year 2

Atrium set course a year ago to be placed in a space of hyper-learning and we found it! As we set course for our second year we are thrilled to be in a position to help our customers take advantage of the above concepts and chart their own journeys. Equally, we are excited to create career defining opportunities for our team and our customers as we strap in for the rocket ship ride brought to us by innovations in Machine Learning and Artificial Intelligence.

Share this post

LinkedIn
Twitter
Facebook