AI is big news for all our futures, but difficult for Business Leaders to firstly understand with respect to their business, and secondly to action. The reality is that it builds on a very long history of improvements in software and hardware over time, and setting out on the AI road involves doing things that are less foreign than you may think.
What will AI do?
At it's heart, AI will do pretty much exactly what we do every day, and very much take for granted. That's slightly scary on one level (machines taking over), but at the level of understanding we can provide a very well understood paradigm to understand an emerging one.
- We start doing a few things from the moment (or even before) we are born.
- Process and store data inputs
- Match data inputs to processes and patterns we see around us, and which we create in our minds
- Modify either our understanding, our actions, or both in response to what we learn
Machines will do - are increasingly doing - exactly the same thing. If we park the complex software algorithms that drive pattern matching, what are we left with as the foundation? Data and processes. Ah - these now look very much more familiar, particularly if you're in or closely linked to IT projects.
Data and Process - The Stairway to AI
Software will continue to become smarter at correlating 'things in bits and bytes' with 'things in the real world', but we can lend a huge helping hand by providing software with good data. Yes, we can process and query billions of transactions, but if we call one thing two names, or three names, or four, we're making life much more difficult than we need to. We're also using software cleverness where it's not particularly needed. We're perfectly capable of assembling systems that provide one clear, consistent view of our operational data. We're also somewhat capable of doing the same with external data.
This thought process puts us on staircase to AI which we can picture in terms of starting with distinctly non-AI systems (good, old-fashioned reporting). These use basic data and process building blocks, and then present the output for a human to evaluate and act on.
I'll briefly walk through the 5 steps outlined above which will demonstrate you're 'AI-ready'.
Step 1 - Automate Standard Reports
Automate your existing reports. Why is this important, we're definitely not doing anything new here? This is important because we take humans out of the basic data collection, report production cycle. We have machines which are programmed to Extract, Transform and Load, we capture reporting definitions in code or in a database, and we schedule and automate the generation and distribution. And crucially - we trust the results.
Too often, organisations think their data is in better shape that it is, where we find duplication, inconsistency, gaps and ambiguity. These realities are often hidden by the hand of man / woman, where staff work during production to reconcile and correct the data. This is probably the hardest step for any organisation to take - in large part because there's limited awareness it even has to be done, as opposed to any inherent difficulty in the doing of it.
Step 2 - Construct Enterprise Dashboards
The next step is constructing a set of enterprise dashboards with KPI reporting and tracking. This has the effect of making the organisation think through what should be tracked in order to deliver on their vision, mission and goals. It also provides consistency in understanding the same across divisions and functions. We now have a 'consistency of purpose'.
Step 3 - Build An Ad Hoc Query Ability
Having machines which output standard reports, and dashboards which track our operational reds, ambers and greens, we now embed an Ad Hoc Query ability within the organisation. What does this do? Firstly, it brings the full width and breadth of input data into focus. We want to be able to query sales, procurement, inventory levels, service levels on any dimension that makes sense. It also begins to embed the human side of what will ultimately by augmented by Artificial Intelligence - an Analytics capability. We can now unleash trained human minds on our business problems and opportunities, and provide actionable recommendations.
Step 4 - Report and Action by Exception
By this stage we may be drowning in data. We can ask any number of questions, and we will supplied with answers in a consistent manner from a coherent set of governed data. We may have tens or hundreds of automated reports with a very low audience level, and where the audience can't keep up with the data. The solution is to start handing off the pattern-matching and decision-making that humans are doing to the machines. Define the patterns which elicit a response. Define the response.
Step 5 - Artificial Intelligence and Machine Learning
We're now in a great place from which to step into the world of AI. We have high quality data at our finger tips. We have already 'codified' much of the thinking which AI will need to build on. We can now take this data and process foundation and use the software of the (very near) future to start looking for new patterns. To evaluate huge pools of information vs. outcomes to predict future outcomes. To use those predictions to change the future for the better. We're flying!