Data Science, Security, IT Ops
Not quite the Bermuda Triangle yet not an easy constellation
Modern business operations can feel a bit like a complex dance. We’ve got various groups swirling around with different goals and responsibilities, all within the same orbit. Our data scientists are focused on wrangling information into beautiful, usable models. Security teams are ever-watchful, working to protect sensitive material. The operations folks work their magic to make it all happen smoothly. Sometimes, though, these orbits cross over, creating tricky interactions.
Think of it like a cosmic ballet where sometimes dancers bump into each other, maybe gracefully, maybe not. Let’s investigate those critical intersections, and figure out how to navigate the turbulence that arises from those crossings.
The Problem of Siloed Work
Data science, a fascinating field, revolves around gathering, analyzing, and translating data. They’re the builders of these incredibly smart AI tools, creating models that can find patterns hidden in the information. They tinker in this wonderful, slightly chaotic environment and get that data humming like a well-oiled machine. However, when it’s time to move these prototypes to the operational phase, data scientists can occasionally have a blindspot concerning compliance and security protocols. Some of that fancy, complicated machine-learning code doesn’t follow the well-worn routes of usual software, causing confusion and questions. The security experts have good reason to get involved since any un-checked tool could spell major problems.
The security people’s whole game is, naturally, all about protection. They’re the guardians of information, ever aware of potential vulnerabilities and constantly working to lock everything down tightly. A system only as strong as its weakest point. Sometimes this can appear as if security teams just throw up barricades. Their work requires rigid protocols and strict compliance procedures which, while super important, may sometimes slow down data scientists’ urge for faster experimentation, causing, at times, mild friction between those two groups. They’re not making anyone’s job easier. Yet, if the teams ignore those concerns, there may be huge headaches awaiting. No one want a public data incident now.
Operations professionals make sure the IT systems are not going to fail on you. They work like the silent guardians keeping servers humming and ensuring everyone has the resources to get on with the task. They are about getting stuff up and going and keeping things running smoothly — which is an often under appreciated skillset. For those folks uptime is always the most important metric and they’ll always favor systems that have seen it all. To them, all of this sounds familiar, especially when it is in production or close to that phase, but they also bring deep understanding about what sort of environment the team can deploy to. If we start a brand new deployment strategy we can face quite a bumpy road ahead! Sometimes these concerns appear as delays when people who favor a “go-go-go” approach want things delivered ASAP and don’t see why someone would question their progress. The Ops team members worry a lot! But, what are we to do without them.
Key Friction Points
Now imagine the overlapping sections where these groups’ Venn diagrams meet in the image shown in your code. Consider the places where problems could emerge — not because someone is at fault, but where team interaction itself leads to problems. First: When Data Science and Security teams must figure out what actually was done to build these new shiny AI products, we run into the first difficulty. Data teams don’t always record or report all those iterative steps they perform, since to some of them their processes just naturally fall into place, almost by magic. On the other side, the Security Team always asks what tools were in use. They also care very deeply about who actually has the privilege to access these important datasets and these high powered AI tools and also who exactly changed the data sets over the weeks and months. And the question “Who touched the model?” becomes quite loud and important if ever we need to roll anything back. Second, there is the interaction of Operations and the Data Scientist. As the Ops team starts to wonder about the performance metrics for all of these new AI powered tools they might discover some unpleasant surprises. All of a sudden, data models appear to need unusual processing power, which must also be backed up, properly secured, made reliable and what else. Also, deployment to production could sometimes be quite a pain for teams used to pushing conventional software. As these folks get ready to start the deployment they ask “is that model ready to roll?”. Finally there is an important intersection between Security and Operations, a tricky place indeed. Those guardians need to know everything that has happened and where things are to do their best. They often ask questions like: “Is the system that is now getting all the buzz also really secure? And is that server with that high computational power really ready?”. Often, the answer to that query is ‘it will all work out’, which really should never satisfy any professional of Security or Operations. When it does, it will only lead to problems later.
These moments of overlap often involve tricky queries which, if ignored, result in slower delivery of products, or even catastrophic security problems. Often, these moments result in wasted time. Even when the three functions get along in general, there is sometimes no true collaborative spirit. There is always risk that one team blames another in a crisis.
Storytelling Power — Why You Should Pay Attention
A common saying in business is that, ‘if it ain’t broke, don’t fix it’. Well, there may not be smoke, but you might be in for a surprise! Most enterprises tend to accept existing arrangements until a bad surprise emerges. There always is an unspoken contract — each group tends to respect others, until things start to fall apart and then… people are in for a world of trouble. This kind of arrangement, where each group stays mostly within their area is pretty standard and makes many team-leads and middle-managers happy enough, because everything has gone according to “plan”. These moments of overlap are seldom acknowledged as key risks and are usually the place where most organizational problems originate. But, with all that, it becomes harder and harder for teams to deliver a great products quickly and securely. Often, the key to these issues does not come from any team within the individual departments but instead in-between the teams.
A change here will require more than a few memos sent around! We need a bit of a story for people to believe that change is good. Here is our story: There were three very distinct teams working very hard. In that story we have our Data Scientists — the visionary magicians; the watchful and diligent security folks; and also the folks who ensure everything ticks and keeps humming; these guys are often known as the “IT Operations team”. For the most part the groups each kept doing their own work well within their area. Each of them felt comfortable, yet their working relationship also suffered from silent resentment of a different kind and approach to work. Data Scientist are all about speed. Security thinks everything is high-risk by nature and should proceed carefully; and IT wants everyone to use existing stuff to make sure nothing goes boom when we all go home on a Friday evening. They would interact in various meetings but their mutual collaboration left a lot to be desired. At least no major catastrophe has ever occurred under the status quo, at least so far.
However, one day, there came an exciting and innovative new project that suddenly started to cut across each of the individual teams areas. A new project was set in motion! This project involved, on one hand, massive new computational challenges and a completely new architecture, while requiring absolute security from the very get go! This forced our Data Scientists to get really chummy with the IT folks on questions of deploying, managing and scaling the brand new platform. And there were plenty of those challenges. At the same time security experts discovered plenty of loopholes that were not very common, forcing all the teams to sit together and come up with a mutually agreeable set of rules. Now that they were actually interacting more, all teams soon realized what strengths every party really has! The new system was actually delivered surprisingly smoothly, just like they wished for it earlier in the organization! After that, the new level of cooperation became the new status quo of this enterprise, transforming not just how they are working on this project, but rather creating new processes across the entire enterprise, resulting in improved delivery and more efficient operation. As our character realized in this simple, made up story, true collaboration will often turn out much better that one ever thought! All of that has occurred because someone finally asked the question!
The Path Forward
For organizations, this means setting up environments for proper collaboration and exchange of knowledge, encouraging dialogue between teams, and breaking down informational siloes, because each team does in fact have great expertise, all different, all of it very useful. The first task is making sure we actually share information freely between our departments! It could just be regular communication meetings, project handoff guides or just cross team training. A simple question may be the trigger that will ultimately improve our organization. Here, it really is important that this effort is not only a formality. Rather, each group needs to get a seat at the table to truly contribute what they can offer for the best results. For instance the data scientists need a path to express what they need without sounding like divas, while IT operations must have room to raise very important red-flags regarding stability, something you do need to pay attention to before anything goes boom! Also the Security Teams require respect for what they contribute. The goal here should be more effective knowledge sharing, but also a deeper respect and awareness between these critical teams.
We also need tools for automation that bring teams together! For instance all steps of AI model deployment are perfect candidate to introduce proper tool chains that allow you to understand exactly what data and processing are going to be used before you send anything to production. If you want to monitor production processes you’ll also need software with robust monitoring features! Also if things fail you might require automated rollbacks — and guess what — even those would need to be setup with proper input from multiple teams. When choosing your tools think of collaboration in mind. A good tool, or rather suite of tools, should be that the data-people, the IT Operations, and security teams alike can easily integrate into their work. By creating one continuous, shared process we can stop playing “he said — she said” and improve our ability to work with each other.
Organizations can even put an expert in charge that actually covers multiple areas of business and teams! These are often folks who have cross domain understanding who don’t belong directly into any of those departments. Someone who is experienced enough to think “outside the box” and understands business problems holistically, can lead these groups to mutual beneficial outcomes. This team also ensures that not one specific approach triumphs in any of these cross functional debates.
This approach really helps bridge some of the very different opinions and creates mutual awareness within our group. We don’t want conflict — rather, we need robust discussions to create the best possible products and the greatest organization.
Ultimately, what this situation illustrates is that it is time to challenge and to really think critically how teams operate with each other in order to enhance mutual output! This doesn’t require massive disruption or sweeping overhauls; but, instead it comes from implementing small changes for improving mutual knowledge and collaboration across team structures. When the whole is bigger than its individual components, we will have taken one step further as a business.
A Final Thought
As a matter of fact, some problems arise because of our own inherent thinking patterns. It appears we are all, as individuals and teams, deeply drawn to our ways, as if pulled down a deep spiral. So often we accept established thinking rather than seek out truly innovative strategies. This way, problems become our shackles; until we make the radical effort to break these mental chains and explore different paths toward growth, efficiency, and innovation. To escape this deep spiral, we need not just good strategy and careful planning. In truth, we also require an active awareness of our place within this ever evolving organization. Because true liberation is within the realm of knowing why something really should be in a certain way, we must choose the conscious effort to look behind all known. Then we shall know that all true wisdom comes from acknowledging the potential limitations of your viewpoint, and from asking questions, in truth, to actually create new levels of understanding.
More shiny stuff on platformeconomies.com!
