Potential Value from AI with TOC

(AI = Artificial Intelligence TOC = Theory of Constraints)

By Eli Schragenheim

AI is the term for tools, that learn from past data to take independent decisions, or to support them. It has become a buzzword for the future of technology to change our World.?

A lot of concerns about the abilities of AI exist. It may improve our lives but may also cause considerable damage.

I believe the Theory of Constraints contributes rationality to judgements. TOC seeks a system’s inherent simplicity to reveal the potential of apparently complex and uncertain situations.?Can the qualities of TOC significantly improve the potential value AI brings to the management of organizations?

TOC’s emphasis is to identify?the right focus?to achieve optimal performance. This also means to identify what not to focus on,?recognizing the limited capacity of our minds.

Can (will) AI significantly help us exploit our human capacity limitation better?

Humans should guide their minds to focus on what truly matters for their goal.?For organisations to achieve more of their GOAL (now and in the future) provides the objective to assess what to focus on now.??

No matter how cleverly we identify what matters?some important things may be missed.?An area neglected by TOC, (which no manager can afford to ignore), is to?identify, emerging threats early – not miss them.

Computers’ ability to process huge amounts of data is also limited – but the limitation is way above any human’s and the gap is continually widening.?We can hope that a human manager will continue to define the strategy. Clever use of software, particularly AI, should constantly check the validity of our focus and warn managers whenever a new critical issue emerges?

AI is widely used to replace human beings to do simple straightforward tasks, like the use of robots in large distribution centres.?Driverless cars are much a more ambitious target, and it is a task majority of humans do well (when not under influence).??Current management emphasis is to use AI to reduce costs by eliminating workers for simple repetitive jobs.?It would be amazing to show that AI could support the substantial Throughput[1]?growth of or even enhance strategic decision making.

The power of AI is its ability to learn from a huge amount of past data.?AI can also be trained to develop critical decision-support information. It observes correlations between variables, sees trends and sudden changes in market demand, suppliers, and flow.?Instead of performing relatively simple tasks and decisions, AI could be used to improve an organisation’s performance.?A good first target may be to improve the forecasting algorithms to also highlight the possible and reasonable spread of results. AI’s ability to identify correlations could show dependency between different SKUs, to significantly improve forecasts.??

More challenging tasks will be to analyse data to determine the potential impact of price changes, and other critical characteristics offered to the market.?Another worthy challenge is to find and highlight irregularities that require immediate management attention.?From a TOC expert’s perspective, it would be valuable to evaluate buffers’ effectiveness better than it can be done today.?Applying AI could be used to improve the intuition and the thinking of open mind managers!?A human manager may be able to use AI to validate or invalidate, assumptions and hypotheses. When this becomes possible it will considerably improve the evaluation of the ramifications of changes. The quality of management will improve.

An important downside of AI, especially from a TOC expert’s perspective, is AI does not use cause-and-effect logic.?The ability to check cause-and-effect hypotheses is a key to successful management.??

A second downside is AI’s dependency on training data, which can lead to erroneous results or decisions (driverless cars continue to make driving mistakes).?A key challenge will be to find ways to reduce dramatically reduce the probability of a significant mistake. The ability to spot such mistakes through cause-and-effect analysis may become vital.

The process to use AI effectively must start with the GOAL. The key elements that impact the goal must be identified to develop the objectives that will deliver valuable results. This list of valuable objectives, to enhance the organization’s performance, should be analysed to discover whether AI, possibly combined with other software, can overcome the obstacles that currently prevent achieving these objectives.

A key idea is to recognize and use AI’s potential. It could provide us with vital information, even new insights,?to become an integral part of our decision-making process.?

TOC wisdom can be useful to set worthy objectives and guide AI. Eli Goldratt’s title for his book “The Haystack Syndrome – Sifting Information out of the Data Ocean” has even greater meaning as data oceans and AI capability grow.

Computers and particularly AI are superior when handling complexity when many variables interact.?The tougher challenge to deal with is uncertainty.??‘Noise’, the common and expected variations, and the rarer risks can cause much damage.

This is the opportunity to use the emerging power of AI combined with TOC wisdom, to support the better planning and assessment of future moves.?

To guide AI to monitor predicted market trends and the impact of external forces, like changes in the economy should be valuable to decision-makers. Predicting the effect of increasing or reducing prices, could yield even more value.?Since many relevant data for such missions lies in external databases it will be likely that services to obtain the data will be required.?Cooperation between competitors that allows AI analysis of their combined data to be carried out by a neutral third party would be beneficial for all.?Such cooperation must ensure no internal data ever leaks between companies.?The analysis that highlights issues like price sensitivity, the impact of inflation, changes in government regulations, etc., could yield currently hidden knowledge so that future decisions will no longer be based solely on intuition.?Human intuition’s disadvantage is the long time it takes to adapt to changes.?AI a huge number of similar past changes should be better in predicting outcomes. AI should be much better if enough of the relevant data has not been made irrelevant by change.

There are two categories to focus on to?use AI effectively when making management decisions. These include the critical question of “what to focus on?”?These two categories have the potential to bring huge value to the management of any organization:

1. Sensing the market demand.?This includes forecasting current trends and?the prediction of the potential outcomes of proposed moves and changes.?AI would give a good idea of the impact of price changes, the economy, and the variety of choices a customer gets or could get.

Warning:??Don’t accept one-number forecasts!?An AI output must translate the result of the statistical analysis into a reasonable range, or a confidence interval, as defined by the statistical module. See my article?https://elischragenheim.com/2021/04/08/forecasts-the-need-the-great-damage-and-using-it-right/

1. Pointing to emerging threats.?TOC wisdom can yield a list of potential threats management should be aware of as early as possible.?There is a need to identify signals from the recent past, that testify a threat is developing.?Giving enough examples to an AI could trigger a deeper search for enough corroborating evidence.

For instance, if an important supplier begins to behave erratically, it may point to management problems, even the possibility of bankruptcy, or that our supplier or client has of us, is fading.?A change in the quantities, and/or frequency a big client buys, can signal a change in the client’s purchasing policies. AI can monitor for changes and cause management to act to corroborate or discount the evidence.

A problem with complex environments is the accuracy of their data.?If your AI module intentionally looks for outcomes that do not fit the data, then notifying the user to check specific data items can be meaningful.

A current existing example is monitoring the need for machine maintenance.?This is an Industry 4.0 feature that identifies when the current pace and quality of a machine deviates from the norm. AI sees the problem before it becomes critical, leaving more or enough time to plan the necessary maintenance.

Critical questions to continue the discussion:

• Are there other topics that AI, guided by TOC, can assist management?
• Is there a generic set of insights into how TOC can have a positive impact on the objectives, training, and actual use of AI?

AI may be guided to thoroughly check the quality of the constraint capacity consumption data including the capacity consumption of the few other critical resources. Comparing this to the capacity requirements of the past and incoming demand can help determine?whether our protective capacity is adequate.

• How can we make the above happen?
• And what training should people using an AI module get?

Eli Schragenheim wants to start discussing this issue of TOC and AI in greater depth. He sees a lot of potentials. Eli is seeking discussion partners with expertise - particularly AI but also other TOC expects. If you have thoughts and insights to add to the discussion please contact Eli.

[1]?Throughput = the rate at which a system makes money; for most of us it is the same as gross margin. A commonly used definition of Throughput is sales less totally variable cost (usually mostly materials).