Enterprises are in the midst of an application explosion and a transformation acceleration. This means their underpinning architectures – made up of interactions between people, processes and technology – are growing and changing more quickly than ever. These all need regular oversight or management.
What kind of organization and decision-making challenges are we looking at? Well, let’s consider just cloud applications. According to security provider NetSkope’s 2020 Cloud and Threat Report, the average enterprise uses 2,400 distinct cloud services and appsi. And cloud now represents around a half of all IT infrastructure; with cloud migration also continuing apace.
As well as the layers of infrastructure, servers, bespoke software and off-the-shelf solutions which keep a business humming, enterprise architects also manage how these interact with external systems, with processes, with people and the numerous documents and data sources that go with these.
Large amounts of technical and financial data are also continually generated. Big data keeps getting bigger. All this needs sensible monitoring as significant costs and risks are at stake: the job of architects.
Many companies are already comfortable relying on machine learning to boost business data analysis or sales workflows. For instance, as part of their CRM, or their cybersecurity defenses.
A 2021 study by Algorithmia found that 46% of enterprises are using artificial intelligence and machine learning to combat fraud. Other top use-cases include customer experience and process automationii.
We can also use machine learning to quickly generate insights from other enterprise datasets.
Enterprise architects who need to identify trends, bottlenecks and opportunities are turning to tools including algorithms and machine learning to get their arms around their technical datasets and business strategy.
Machine learning also allows architects to extract an additional layer of value from the very structure, connections and complexity of the architectures they’ve put time and effort into documenting.
They can use it to draw information from multiple layers of the enterprise architecture, look for patterns to ‘fill in the blanks’ and bring this all together intelligently into straightforward, everyday recommendations.
Experts are also expecting AI to benefit from graph database technology; as a recent article in Information Week explained “AI’s future focuses on graph modeling. Graphs encode intelligence in the form of models that describe the linked contexts within which intelligent decisions are executediii.”
Faster Digital Business Decisions
Machine learning is math not magic. It is highly data-driven and based on patterns and relationships in data.
Upload a spreadsheet of applications, lifecycles or financial data and link in other architectural content in ABACUS and the machine learning engine “Ask ABACUS” will use this dataset to provide a quantitative prediction of the values which belong in any empty cells. For instance, where an ‘empty cell’ is the TIME (Tolerate-Invest-Migrate-Eliminate) recommendation for an application, machine learning will propose a recommendation.
Valuable early use cases of machine learning in enterprise architecture include:
- More efficient data-entry, and better data quality: machine learning suggests values for empty cells
- Quickly assigning ownership: where a new application might logically belong, be managed or support a function or team
- Zeroing in on obsolete systems
- Identifying when two systems behave in a similar fashion; should they be combined?
- Indicating where an application or technology may support a key process
- Identifying which applications or processes should have high criticality or risk ratings
- Identifying opportunities to improve processes
“Ask ABACUS” Predictions & Recommendations
“Ask ABACUS” machine learning uses predictive analytics, natural language processing (NLP) and smart data discovery.
Machine learning was first introduced in ABACUS in 2019 and has been extended and enhanced with new releases.
It’s straightforward to enable and configure in–browser, and training intervals and other parameters can be also be adjusted easily. For example, users can set the minimum number of entities of a given type required before Ask ABACUS will include them in its training set.
Machine learning generated suggestions in ABACUS appear in cells with a red border. Machine learning provides a proposed value for that cell, and also the confidence interval for that recommendation.
Once enabled, “Ask ABACUS” suggested machine learning values will appear in cells with a red border, providing a proposed value and also the confidence interval for that recommendation. For instance, “Ask ABACUS” may suggest that the owner of a new application is “Jemima Black”, with a likelihood of 94.5%.
If you agree with that suggestion you can choose to accept it, and that ownership relationship is added to your repository.
This way, we can combine both machine intelligence and human intelligence. “Ask ABACUS” proposes values which can be reviewed by the architect, who might take into account politics, context and future plans. It’s important to note that Ask ABACUS provides just suggestions and architects still maintain final control over implementing ML recommendations.
Enterprise Specific Results
Every enterprise is different. Likewise, no two “machine learning brains” are the same.
The machine learning engine in ABACUS learns and adapts to the enterprise itself, because it references the tailored repository underpinned by a highly configurable graph database.
As the architecture team accepts or rejects recommendations, and adds data, “Ask ABACUS” also learns and self-tunes, improving its predictive power.
By maturing alongside the enterprise architecture repository, the machine learning engine will reference the context of business data and decisions, proactively making connections, filling out incomplete datasets, and recommending next actions.