I need to evaluate the monetary potential of AI at my company for the year 2025 to justify a long term investment in a dedicated AI center of competence. The departments in which AI would be used later on (Purchasing, R&D, Manufacturing, …) are generally very interested and some of them have already built some prototype use-cases. However, they are unwilling to predict and quantify any future potential. The investments in AI can only be justified if we can state the monetary benefit.

Are there any creative approaches out there to estimate the monetary potential of AI?

Any help is greatly appreciated!

  • 3
    $\begingroup$ If the departments are not even willing to estimate benefits of the use cases, anything regarding RoI is going to be a wild guess. I cannot answer this question, but I'd challenge the premise that you need to aim directly for a long-term investment. Start small, with a little blue sky R&D that starts to answer the questions about feasibility and get better estimates on RoI. Get one or two experts in for now, not a "center of competence". Re-evaluate after 6 months or a year of R&D. $\endgroup$ Jan 17 '19 at 9:45
  • $\begingroup$ curious, but what industry do you work in? $\endgroup$ Jan 21 '19 at 5:18

It would be irresponsible to try. Several in the past, dating back centuries led to various trends that some call winters. I don't agree they are winters. They are just people returning to reality. Creating intelligence in computers has been progressing forward steadily since Charles Babbage. It has accelerated due to advances in digital electronics, in discrete mathematics, and in brain imaging.

It is possible to predict, with some unavoidable uncertainty, the impact on the profitability of existing products or services five years into the future, provided the AI efforts in the corporation are planned at a high level first.

The main issue is that what we mean when we say the word intelligence is really a collection of abilities. Tesla died with no assets yet his designs are in our appliances and the grids that power them. A billionaire may not be able to draw a sketch of himself worth a penny, and Michelangelo may not have been able to design a way to transmit power to homes and businesses.

There is no one thing called intelligence. It is many things. That's why computers exceeded humans in arithmetic long ago, but there are no combinations of hardware and software that can even come close to writing a physics textbook or even passing an arbitrary high school physics final exam. There is much more not known about intelligence then there is known about it.

Monetary evaluation may be necessary as part of the solution, but the end goal is monetary projection. That means one must evaluate the state and trajectory of something AI experts don't yet understand well. We can guess accurately that, if the direction of an AI department or AI groups within multiple existing departments is well managed, gains can be made. We can evaluate the use of AI in the past within the same business type and use the trend we see to make projections along the same line based on how technology trends usually progress. We can also factor in the emergence of competitive businesses also progressing with their use of AI.

That purchasing, R&D, manufacturing, and other departments are interested in AI, have done some initial play with examples and maybe built out some proof of concept code but are unwilling to quantify future potential is not surprising. It is not any easy financial modelling task. To begin the task, some information must be collected and scrutinized to ensure it is an accurate starting assessment.

  • What are the kinds of products and services of the business?
  • What are the critical corporate functions that support those?
  • What are the key technological, market, and production characteristics of the business that places it in a group of like businesses?
  • What currently developed and known-to-be-effective AI technologies most likely produce profit enhancements for this business?
  • What past technology advancements are most like those current ones?
  • Quantitatively, what were the financial outcomes of the leveraging of those past technologies over time?
  • How is competitive leveraging of the same technologies by competitive businesses likely to draw business and divide profit margins?

Once this information is gathered and its accuracy is ensured, a financial model can be constructed and implemented in software and projections of various scenarios can be made. It is a development project in itself. Many businesses are bypassing this project and just assuming that the current AI craze must have a high ROI because large corporations have made the choice down that road. This question exhibits a more prudent course of action.

In addition to the departments listed, there are a few more to consider.

  • Marketing
  • Finance
  • Strategic planning (at the board and senior management level)
  • Accounts receivable

The financial evaluation of the distributions of costs and benefits of establishing an AI department or group to support other business functions is probably roughly additive. Consider approaching each of these separately and always compare the three most obvious options.

  • Fund the centralization of AI research to delegate and reuse expertise inter-departmentally, such as with IT, HR, and accounting.
  • Fund the decentralization of AI research, such that specific expressions of AI are running independently and in parallel
  • Don't fund AI research

Of course there are funding amounts spread over the five years to be determined for the first two and they can be mixed as government is in federalism or as a mid-level corporation that has an accountant in each department that reports up its director but is also responsible to an accounting department.

These are a few functions the planning or control can benefit from AI in the categories specified in the question.

  • Optimizing purchasing thrift in conjunction with quality control
  • Optimizing manufacturing thrift in conjunction with quality control
  • New product or services or enhancements with AI involved in their design
  • New product or services or enhancements that run with enclosed AI components

The substantial overlap is evident in the description of potential benefits.

The investments in AI can only be justified if we can state the monetary benefit.

The department heads won't volunteer any projections because they will then be held to them, and AI has about the same level of success in past projections as photovoltaics (the panels of which were supposed to mostly replace fuels a decade ago) and robots (which were going to be doing our dishes for us around that same time). People are skittish for good reason.

Are there any creative approaches out there to estimate the monetary potential of AI?

One creative approach is to estimate the loss to the company if no AI is funded, which may be so dismal that it consumes the entire five year AI budget.

  • Loss in personnel that don't want to work for a company that doesn't have AI in the job description
  • Loss in credit and confidence because a prospectus without AI is like raising a flag with the bright red letters, "We're not progressive."

That may not fly, but if it did, the analogy of the departmental accountant and the accounting department to quality control and educate the entire corporate financial mechanism is likely to also fly to the same board or senior management.

The less creative approach is to find the most similar past trends to each of the applications of AI technology to the four items above and use them to project benefit. Cost can be based on the typical infrastructure needs plus the extra fast workstations with hardware acceleration plus the cost of labor, which can be estimated now that the term AI engineer is starting to appear in salary surveys.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.