The AI Dashboard Dilemma: Does More Information Improve Decisions?
Businesses have operated dashboards of all types.
Sales dashboards, operations dashboards, maintenance dashboards, inventory dashboards, and many more.
With AI, businesses can now take it a step further and include predictions, anomalies, recommendations, summaries, and more within their dashboards.
However, the real question that needs to be asked is “does this improve the operators’ ability to make decisions?”
An operations dashboard with multiple signals displayed can certainly show that something happened.
A rise in temperature, more vibrations in a piece of machinery, an increase in energy use, and various other events and alerts can fill up a dashboard of an operator or technician.
While they might have more information at their disposal, they might also start to ask more questions: “what is more important?”, “which of these changes deserve my attention first?”, etc.
This is where the application of AI can begin; when paired to the decision that needs to be made.
Rather than asking the question “what signals should the dashboard display?”, focus on “what decision does the operator need to make?”
For most applications, this will be something along the lines of:
should a technician perform an inspection?
or maybe
should an operations manager alter parameters due to changes in the equipment?
perhaps a purchasing manager needs to know if the inventory needs to be replenished.
With a clearer idea of what the operator needs to do, decisions can be made as to what information needs to be provided, when it should be displayed, or what analysis might be required.
A decision workflow could take shape as data → analysis → context → recommendation → human review → implementation.
While the AI might not have access to all information, it can work with what it does have to provide a recommendation that will save time and resources when paired with a human at the end.
Get Nayantara P S’s stories in your inbox
Join Medium for free to get updates from this writer.
Instead of displaying ten individual signals from different pieces of equipment, it could alert the technician that multiple changes occurred simultaneously, and provide a recommendation for them to review and confirm.
Context is key when it comes to recommendation engines.
A temperature change might make sense depending on the time of year or what production cycle the equipment is in; a change in vibration might be expected depending on a load change; inventory reductions might be anticipated depending on the season.
To make accurate recommendations or suggestions, AI will need access to supporting information; information that puts the data stream and previous results into context.
This is where the combination of AI and IoT can be seen as a strong one.
IoT serves to provide data about the physical world or process.
AI makes associations, connections, and recommendations based off of that data, and then can either alter a physical process directly (IoT) or make suggestions to humans about said changes.
In certain applications, the difference between the decisions that should be made automatically verses those that a human should make can impact how and when AI is implemented.
Low risk, easily reversible actions lend themselves to automated suggestions or implementation.
For higher risk or higher value items, having the proper recommendation engine in place without final implementation can go a long way.
This is less stressful for the operators and technicians (they aren’t expected to predict the future) and allows for human elements to be introduced if certain suggestions were made that rely on information only the operator at the machine would know.
Measuring change is often easier said than done.
A new dashboard could have a ton of charts and graphs, but it doesn’t change day-to-day operations if people aren’t engaging with it differently.
After implementing a recommendation engine (or whatever change is sought), metrics need to be tracked in order to ensure that change actually occurred.
Was time saved? Was cost reduced? Did maintenance crews become more efficient? Were operations run at a higher parameter after suggestions were implemented? Were decisions easier to make? Will inspections and evaluations be conducted earlier or less frequently due to the increased certainty?
These are just a few examples, and the best ones will always stem from an actual change that the business wants to see.
At Aperture Venture Studio, this is the type of business that we are looking to build.
Our portfolio companies focus on ventures that tie together the power of AI, and operational or operational issues that require a technological tinkerer to solve.
We are more concerned with providing information or analysis to operators at the point of need compared to simply giving them information that they might not need at a point that we feel is important.
AI-assisted disclosure: This article was created with AI assistance and reviewed for structure, clarity, and accuracy. The author remains responsible for the final content.
