Scaling technology and business processes are not equal, here is a framework for the business side of artificial intelligence.
Since the beginning of the enterprise technology scaling software has been a difficult task to get right inside large organisations. When it comes to Artificial Intelligence and Machine Learning, it becomes vastly more complicated. Data governance, business intelligence, algorithmic infrastructure, and even organisational charts has to change for this new approach to creating business value.
Matching (the right) data to (the right) problems.
Software is deterministically architected to create solutions to problems. Code comes first and the solution is born from running the software. Artificial intelligence is more about understanding the past to reproduce or predict the future. Data records of previous events represent the past, and more than likely some of those events represented value to the business. Aligning decision makers with data scientists on which data assets exist and how they correlate to a specific problem takes a deep organisational understanding and a fundamental understanding that being internally open with data is a requirement to create solutions.
Breaking down data silos
One of the major issues that small and large organisations operating with data and algorithms are dealing with, is definitely the transitioning from siloed work to interdisciplinary collaboration. AI and machine learning are the emblem of multidisciplinary activities at the intersection of statistics, computer science, economics and more specific concepts from the domain of interest.
The multidisciplinary nature of AI should be familiar not only to the direct builders of AI solutions – data scientists, computer programmers and project managers – but also to the higher levels of the organisation. In order to foster such an interdisciplinary approach, organisations that want to succeed with AI and machine learning must move from the “Data Cartel” mentality of hoarding data assets in favour of a more collaborative approach to data lakes. This approach of being more transparent with data assets will allow the curiosity of data scientists to cross pollinate data systems which often leads to impactful discoveries.
Investing the right amount of time and money for ROI
Another misconception of AI and machine learning in large organisations is about timing. It is a common belief that data driven solutions automating particular business tasks will deliver results right from the day after deployment.
When building software the requirements tend to be more finite. Code this thing so that when this other thing happens it responds with `X` or `Y`. When dealing with Artificial Intelligence the problem space is commonly distilled into probabilities.
For example, in finance, what is the probability that over the next three days I’ll spend my money on fuel for my automobile. These algorithms don’t immediately show results and often times they are made up of multiple algorithms working in tandem leading to a prediction based on a prediction.
These tactics take time to perfect and understanding that immediate value isn’t a normal case. Overtime, as the predictions receive more data and the data scientists have more time to manipulate the machine learning process, an increasing accuracy in the predictions will show.
It is common to hear large organisations working on an “agile” org chart and business process. The cross-functional aspect of AI and Machine Learning lends itself to a more agile organisations as it’s typical to find more knowledge sharing in a Pod like organisations chart as opposed to a traditional manager-managee environment. AI and machine learning are both agile by nature, as many problems are solved by trial and error. Applying more traditional and less agile processes to building data-driven solutions can lead to development cycles that can become just too long to sustain. This will, in turn, contribute to extending timelines and increasing disappointment at the end of each of such cycles.
It is essential to have leaders who are competent with AI – not necessarily at the same level as the builders of course – because sooner than later leaders’ decisions will play their role. This fundamental understanding of how data, algorithms, creation process, and iteration leads to a better understanding of how to create reproducible, production ready, solutions for the business. While leaders shouldn’t know the nitty-gritty of AI and machine learning techniques, they should be comfortable with the capabilities of AI and know of the existence of those algorithms that solve specific problems in order to identify the patterns of recurring challenges and their solutions.
Educating your executive team on Artificial Intelligence
Having leaders educated in AI and machine learning is leading successful data-driven companies towards establishing internal academies and providing training courses specifically designed for managers and decision makers. Great learning topics around AI are:
- Understanding the planning process of AI projects
- Interpreting business cases of AI and how they differ from typical software
- The difference between production software vs production AI
- High level techniques of Artificial Intelligence and how they differ from typical software applications.
Get started with Artificial Intelligence
Scaling, organisational readiness, data practices are all key to understanding how to be successful with the new tactics available with Artificial Intelligence. At Amethix, we deploy our approach called Crawl, Run, Iterate. Within this framework we ensure the base education of decision makers is there, we create our own business cases to evaluate the impact of our services, and then we start our process of Crawl: Data exploration & prototyping, Run: Deploy on production data sets, Iterate: Continuously iterate with new tactics and models to increase the business value.
With high levels of collaboration, the Amethix team works to create and maximise the value of your business and the people you serve.