How Will AI Help Us With KPIs?
by Stacey Barr |There are lots of promises about how emerging AI will transform how we use data in business. How exactly will AI help us with KPIs?
Artificial Intelligence (AI) or machine learning are powerful technologies that are inevitable. But my concern is that we might put too much false hope in what they can do to help us with using KPIs to improve performance. I’m concerned that we’ll expect AI to do the critical thinking that only we can do.
But more importantly, I’m interested in your thoughts, ideas, hopes, dreams, expectations, concerns, and anything else you want to say about AI and KPIs.
Please share your comments on this Measure Up blog post. I’ll respond to each one, and use your brilliant ideas to help co-create a rewrite of this article, to thoroughly answer the question: “How will AI help us with KPIs?”
How exactly will AI help us with KPIs? Share your hopes and dreams and expectations in the comments.
[tweet this]
DISCUSSION:
Do you see any risks or bad assumptions about how AI will help us with KPIs?
Speak Your Mind
Connect with Stacey
Haven’t found what you’re looking for? Want more information? Fill out the form below and I’ll get in touch with you as soon as possible.

167 Eagle Street,
Brisbane Qld 4000,
Australia

ACN: 129953635
Director: Stacey Barr
Interesting question about A.I.’s usefulness in helping develop KPIs?
My experience with designing and implementing something as mundane as ERP and Accounting systems has shown me that the Bennett-Jeneret’s (from Dune) admonition that “fear is the mindkiller” applies even more when KPI’s are constructed and promulgated. People often feel powerless to change systems that they are presented with. Perhaps using AI to collect information about the implementation of the KPIs and in also in reporting how well the organization is achieving those goals would be helpful.
I advise not using AI during the creating of the KPI’s. .
I’m with you on not using AI to create KPIs. People need to buy in, and they won’t if they don’t have control or at least influence. I’ll spend more time learning whether/how AI could do data interpretation and draw conclusions from its interpretations. Thanks for sharing these thoughts, Free.
AI is going through the phases that the computer itself went:
1. There are those who understand what AI is and what it can help them achieve, and they know its limitations. But,
2. There are those who think AI will solve all our problems that require us to think and decide. Just like it was when computers were new.
3. It is a huge field work for consultants to lead organisations find their way in AI productive utilization.
4. There are going to be very expensive AI projects that will end up to be white elephants; just as some IT projects still are.
Wise observations, Leonard. And I agree. I’ve seen millions spent on KPI dashboards that people believed would end their manual reporting and accelerate their decision-making, only to ditch the dashboard within weeks because it didn’t tell them what they needed to know. And mostly because those people didn’t realise the thinking was up to them, as you point out! I guess that begs the question, how well can we realistically expect AI to ‘think’ and what kind of thinking?
On ‘thinking’, I can say it is about remembering what is right in a given situation for routine work. AI no doubt will excel in this. On the other hand, for complex tasks, thinking may include:
1. comparing options in mind to decide on suitable or preferable action when faced with unfamiliar situations.
2. coming up with new ideas on how to accomplish something, solve a certain problem or overcome new challenges.
As performance measurement is almost routine given a methodology, use of advanced tools in data collections and analytics can help in having clear visibility of actions’ impact on our processes. Therefore we humans must first have a good understanding of what we want to achieve, how we are going to achieve and how we are going to measure it, before automation. So, as Free put it, automation of KPIs comes later. By stretching this concept, even AI will come later to provide input into performance improvement programs.
I think of AI in two buckets Correlation Engines and Knowlege Trees. (Yes I know that is is harder than this but…) Correlation engines look for statistical significance in data patterns; knowledge trees assist by removing human error in decision making. I believe Correlation engines can help us answer questions like what are the limits of systems, so we can set good targets and develop good projects (experiments) to improve results. Knowledge engines can help with enabling people to do more complex tasks with a higher degree of efficacy or freeing people from mundane processing tasks or monitoring and correction where results are not being met. My largest concern is AI designed and used without good systems thinking. Like the Wizzard of Oz, people will walk up to it and ask for answers AI cannot provide unless the questions have been anticipated and trained in the system. My theory; expect the same rate of success (26%) in AI implementation seen in any major enterprise initiative (Six Sigma/Lean, Reengineering, Agile Development, etc.) Only by understanding what is necessary and sufficient for success and delivering it is a viable platform for implementation and growth established.
Tom, I like the correlation engines and knowledge tree concepts. Are these your own, or do you have references where I might learn some more? Also, thanks for bringing up the point about humans needing to first decide what they want before any AI application can work well. I reckon that deciding what we want is the type of thinking I wouldn’t personally want AI to do for me (that goes for Free’s and Leonard’s comments about not letting AI take over the decision about what to measure).
Hi Stacey,
For sure like any tool it has it’s pro’s and con’s:
Pro’s:
– It will help process in a smart way large amounts of data (through machine learning and algorithms)
– It will remove the emotional bias we tend to have a humans.
– It will help answer accurately the what is happening questions.
Con’s:
– Algorithms and machine learning is set based on data provided by humans if the sample chosen is already biased the AI system will also be.
– Relying only on technilogy and not using human intervention wisely.
– It wont be able to answer the question why it is happening and what to do about it accuratly (at least not yet)
At the end of the day AI is another technology that we are programming, we can use it to have an efficient and unbiased computation of data,
But it can never decide what the right measure should be (PUMP process methodology) and how to improve results.
Regards, Cynthia
Cynthia, thanks for your thoughts. The emotional bias problem you mention would be a great one to resolve with AI… people don’t often know how to detect or correct for their own emotional bias when interpreting data. And that goes for political worldview/values biases as well! In any case, I’m not sure how AI can become immediately better, overall, at using KPIs until we have enough humans knowing what it means, and what it’s like, to use KPIs well!
AI is like Google, it’s a great tool if you know what questions to ask. There’s is a trend called Augmented Intelligence in which human intelligence interacts with AI and ML to find insights more easily. For example, there are platforms that allow you to answer questions about your data without even needing to build a dashboard or a chart.
My impression is that AI and ML will help with the questions:
What happened?
Why it happened? (Less)
The question How to Respond will require a human unless it is a routine task that is possible to automate.
My 2 cents 🙂
Thanks Juan – I hadn’t heard about augmented intelligence but it seems more practical to me than artificial intelligence. I wonder if there are any augmented intelligence platforms that would respond to a question about whether performance has improved or not with an answer based on what an XmR chart would tell us. Can AI know about interpreting data with an understanding of statistical variation if we don’t explicitly tell it to?
Good one! AI and its machine learning capabilities are becoming increasingly important in today’s market. The KPIs that have worked before may be obsolete as new indicators are emerging through automation and AI. CSAT.AI is an AI tool that helps the KPI’s with measurements that proactively improve agents First Contact Resolve and dramatically impact the CX. MaestroQA, ScorebuddyQA, and Salesforce Einstein are some other AI tools that are also working in this area.