Buy or Build an AIOps Solution.

Buy or build is the essential question most CIOs ponder when considering a software innovation. As enterprise IT investment policies realign in a post-COVID environment, these questions are far more relevant than before. Enterprises need fast, practical outcomes now - not aspirational goals that develop into value later. This is reflected in enterprise investment goals - Over 30% of organizations are increasing investments in cloud management platforms and AIOps tools(Gartner Research) in 2021. This is basic cause and effect. COVID-19 has forced greater adoption of cloud infrastructure and services, leading to greater need for these tools. This brings us back to the essential question. Buy or Build - how should enterprises approach an AIOps implementation?
To answer this, let's look at the core aspects of each approach keeping in mind the ground realities of AI development. The first aspect to consider is cost. Simply put, it's cheaper to Buy. The reasons are obvious - vendors recuperate their costs by targeting multiple customers and an enterprise is only targeting internal stakeholders. But that's true of any software development - and not a limiting factor for large enterprises. What makes it costlier with AI? Firstly, there's the high project failure rate. According to Project Management Institute, on average 14% of software projects fail. In the case of AI projects, the most conservative estimates for failure rate are as high as 30%(International Data Corporation), with most estimates closer to 70%(Capgemini) and 80%(Dimensional Research). The high costs of Build approach also include building infrastructure like data lakes, deployment and monitoring mechanisms as well as support activities like maintenance and upkeep.
The second important aspect is time to value. One of the biggest complaints of AIOps adopters is that time to value can be as high as 6 months or over a year, even when trying out vendor solutions. Most of this time is lost in deployment, configuration and training. Building a solution can expand this timeline to multiple years and leave you far behind the competition. Buying, on the other hand, has to be carefully researched as well. Decision makers must choose vendors that support integrations with common interfaces, automated discovery, and pre-designed workflows so that they can get you a running start with AIOps adoption.
The third aspect to consider is the availability of skilled resources. After cybersecurity, data science roles have the highest difficulty in finding recruits with the right skillsets. However, AIOps requires more than data science skills - It also needs years of experience in ITOperations to understand the technology and its challenges. AIOps solutions must be designed by people with a strong fundamental knowledge of data science partnered with a high amount of domain expertise in IT Operations. Finding this combination of skillsets is extremely challenging, costly, and time-consuming, making "Build" a much harder approach than "Buy".
Finally, let's explore the Transformation and Upkeep aspect. AIOps solutions started as diagnostics assistants and evolved into root cause detection, log analytics, and automated remediation providers. And AIOps has not matured yet. The technology involved is rapidly advancing and the use cases are broadening into new fields like development in addition to Operations Management. Keeping up with this technological transformation is in itself an expensive challenge and requires dedicated teams working on incorporating the latest technological developments into new solutions. The other half of upkeep is the longevity of AIOps solutions. ML models that form the core of these solutions do not remain accurate forever. They have to be constantly tested and vetted for performance and even replaced when necessary. Maintenance and upkeep of AIOps solutions is going to be the largest long-term challenge in AIOps adoption as AIOps becomes increasingly domain-agnostic.
Every point we've made above seems to favour the "Buy" approach. So when should you go for a "Build" approach instead? Building an AIOps solution works for large enterprises that already have the strong foundations to make and maintain AIOps solutions. That is, enterprises whose core business activity is centered around IT Operations and who have a large talent pool of data science professionals to draw from. It can also work for organizations focused on a small number of focused use-cases with limited data variety. These organizations can use a domain-centric approach to AIOps with limited application. It can also be important for organizations that want complete control over their solution. After all, no vendor can meet 100% of your requirements out of the box. Apart from these exceptions however, "Buy" approach is effective for most organizations whose core operations are not centered around generating revenue from AIOps activities. Organizations can also choose to "Partner" with vendors that provide Managed AIOps services if they have exacting requirements that are not met by out-of-the-box solutions.
To learn more about Algomox AIOps, please visit our AIOps Platform Page.