Amen. I have worked for a long time in ERP system implementation, and the biggest problem in every one I have dealt with is the business being unable to define and model their processes. The typical definition is a very high level, hand wavy power point deck gesturing towards what they do, when what is needed is a detailed step by step along with a data dictionary specifying "This field contains this data, controls this aspect, and this is what the data in it does by value." Then they are shocked when their system does work right.
It is probably true that most businesses don't need a high level of magic computer dust to work well. Hell, many make their business worse by trying to bend the operation of the business around the needs of the tool. But just as you say, many try to use the system to correct the disorders in the business, when the first step of setting up the system is a low level, nuts and bolts break down and organization of the business so you can model it. Whoops.
Another angle to this is that when a company grows via acquisitions, it tends to have disorganized and chaotic computer systems. I worked at Citigroup for a while, and Sandy Weill's legacy of building-via-acquisition was evident in the innumerable, often conflicting, and poorly designed, computer systems it had running the bank. You may recall a news story from a few years ago about an errant wire transfer that an Indian sub-contractor sent to (I think) Revlon, for something like $900 million. This does not give me much faith in the US banking system, irrespective of other issues with it, such as SVB-related bank runs, etc.
My friend owned a software company that made legal software for the court systems in many counties in California. He had happy customers who would get together and fund improvement and modification as needed. Then the state of California decided to make their own software for all the counties and gave a 5 million dollar contract to a consultant and prepared required specifications for all software functions. My friends software achieved all the specified requirements after a significant redo of the code. After many years of project cost escalation and repeated failure to meet their own standards for software performance and security and after spending a couple of hundred million being spent the estimate for completion went to 500 million dollars before the government bureaucrats pulled the plug. Meanwhile my friends company had the counties fully integrated, which allow citizens to pay tickets, fines, and file court documents, etc. on-line while access to all documents in the courts were instantly available, search capable, etc.
The irony is that he would have happily sold the company to the state for 10 million before they tried to put him out of business. He ultimately sold it to another legal software company for a lot more, after the state gave up trying to do their own system.
Electronic medical records have many issues, disparate systems is not really on the radar. One of the biggest, the underlying data is not uniform. Different instruments (particularly between manufacturers) give slightly different results. Any doctor confronted with a major decision will order their own test so they can correctly interpret the result. Reading another doctor’s diagnosis has the same problem. There are many other issues. People could carry the data themselves - most could fit their EMR on their phone. Every health system can read ICD10 codes and xl7. Apple and MS use xml data for their health data service. Everything is in place, probably the biggest issue is liability.
There is a more abstract version I have been thinking about for some time as regards the typical dysfunctional pathologies of the administrative bureaucracy and, perhaps, how they might be mitigated short of radical upheaval.
"Organizations get the computer systems they deserve."
I think I get what you are saying and I generally agree but I think "deserve" is a really bad word choice. Just because an action typically produces a particular result, it doesn't mean that result is deserved.
If someone makes a driving error, they don't deserve to crash. In fact, they usually don't.
If someone is poor and uneducated, they don't deserve a life of poor health.
There is very little in life we deserve. We often get what what we should expect from our actions but sometimes not even that.
Imposing order on an org that is chaotic, but well functioning, will certainly lead to some time of reduced functioning. Yet "well functioning" is usually a multi-variable function not fully described. Even "profit", or "revenue" or "cost" fails to describe short / medium / long term trade-offs, as well as relationships to other variables.
There was a great (Ben K?) comment on system optimization as compared to more limited sub-system optimization. Optimizing one sub-system often fails to optimize the system.
IBM spent more than $5 billion on trying to standardize on a semi-custom SAP system in a project called Blue Harmony. After disappointment in the first markets of China and Germany, that project was semi-cancelled, continued for new Africa markets and expecting slower migrations of established country systems, each previously in a not-consistent silo.
Their later system integration idea was to create a standardized "data lake" (not warehouse), from which each country's silo-kept data could be standardized and fed into. Then all the above-country reports would get consistent and compatible data from that data lake. My friends say this process is continuing and seems to be fine - but IBM "Big Iron" remains old, and getting older.
Then they bet BIG on ai - Watson (winning at Jeopardy!) But it is NOT an LLM, and while it was doing some stuff in medicine, reality was not as good as the PR. Watson medicine was sold off in 2022, "for parts".
" there is currently no way to connect that information, to link it to an individual across all the domains in which they get care, and then to develop a holistic picture of who they are, of what their diseases are, of what the best treatments are, and how to ensure that they get the best care at the lowest possible cost. There is no connectivity right now that can do that at scale. ...
the generation of technology that they had was nowhere near ready to accomplish the things that they set out to accomplish and promised that they could accomplish."
As tech changes, the software available changes and what the org gets, or deserves, change. Tweet seen: "AI eats software throughout the world"
There will be some far less creative AI systems that are highly, pedantically accurate - because there's a demand for truth; only the truth is a predictable answer about the past.
Even predictions about the future that come true are not exactly the same kind of "true" as the past which has already happened -- but what exactly did happen, truthfully, often remains in debate, thus somewhat unpredictable.
Amen. I have worked for a long time in ERP system implementation, and the biggest problem in every one I have dealt with is the business being unable to define and model their processes. The typical definition is a very high level, hand wavy power point deck gesturing towards what they do, when what is needed is a detailed step by step along with a data dictionary specifying "This field contains this data, controls this aspect, and this is what the data in it does by value." Then they are shocked when their system does work right.
It is probably true that most businesses don't need a high level of magic computer dust to work well. Hell, many make their business worse by trying to bend the operation of the business around the needs of the tool. But just as you say, many try to use the system to correct the disorders in the business, when the first step of setting up the system is a low level, nuts and bolts break down and organization of the business so you can model it. Whoops.
Another angle to this is that when a company grows via acquisitions, it tends to have disorganized and chaotic computer systems. I worked at Citigroup for a while, and Sandy Weill's legacy of building-via-acquisition was evident in the innumerable, often conflicting, and poorly designed, computer systems it had running the bank. You may recall a news story from a few years ago about an errant wire transfer that an Indian sub-contractor sent to (I think) Revlon, for something like $900 million. This does not give me much faith in the US banking system, irrespective of other issues with it, such as SVB-related bank runs, etc.
My friend owned a software company that made legal software for the court systems in many counties in California. He had happy customers who would get together and fund improvement and modification as needed. Then the state of California decided to make their own software for all the counties and gave a 5 million dollar contract to a consultant and prepared required specifications for all software functions. My friends software achieved all the specified requirements after a significant redo of the code. After many years of project cost escalation and repeated failure to meet their own standards for software performance and security and after spending a couple of hundred million being spent the estimate for completion went to 500 million dollars before the government bureaucrats pulled the plug. Meanwhile my friends company had the counties fully integrated, which allow citizens to pay tickets, fines, and file court documents, etc. on-line while access to all documents in the courts were instantly available, search capable, etc.
The irony is that he would have happily sold the company to the state for 10 million before they tried to put him out of business. He ultimately sold it to another legal software company for a lot more, after the state gave up trying to do their own system.
Electronic medical records have many issues, disparate systems is not really on the radar. One of the biggest, the underlying data is not uniform. Different instruments (particularly between manufacturers) give slightly different results. Any doctor confronted with a major decision will order their own test so they can correctly interpret the result. Reading another doctor’s diagnosis has the same problem. There are many other issues. People could carry the data themselves - most could fit their EMR on their phone. Every health system can read ICD10 codes and xl7. Apple and MS use xml data for their health data service. Everything is in place, probably the biggest issue is liability.
See: Conway's Law: "Organizations design systems that mirror their own communication structure."
Here's a great Casey Muratori video introduction to the idea.
youtu.be/5IUj1EZwpJY
There is a more abstract version I have been thinking about for some time as regards the typical dysfunctional pathologies of the administrative bureaucracy and, perhaps, how they might be mitigated short of radical upheaval.
"Organizations get the computer systems they deserve."
I think I get what you are saying and I generally agree but I think "deserve" is a really bad word choice. Just because an action typically produces a particular result, it doesn't mean that result is deserved.
If someone makes a driving error, they don't deserve to crash. In fact, they usually don't.
If someone is poor and uneducated, they don't deserve a life of poor health.
There is very little in life we deserve. We often get what what we should expect from our actions but sometimes not even that.
Imposing order on an org that is chaotic, but well functioning, will certainly lead to some time of reduced functioning. Yet "well functioning" is usually a multi-variable function not fully described. Even "profit", or "revenue" or "cost" fails to describe short / medium / long term trade-offs, as well as relationships to other variables.
There was a great (Ben K?) comment on system optimization as compared to more limited sub-system optimization. Optimizing one sub-system often fails to optimize the system.
IBM spent more than $5 billion on trying to standardize on a semi-custom SAP system in a project called Blue Harmony. After disappointment in the first markets of China and Germany, that project was semi-cancelled, continued for new Africa markets and expecting slower migrations of established country systems, each previously in a not-consistent silo.
Their later system integration idea was to create a standardized "data lake" (not warehouse), from which each country's silo-kept data could be standardized and fed into. Then all the above-country reports would get consistent and compatible data from that data lake. My friends say this process is continuing and seems to be fine - but IBM "Big Iron" remains old, and getting older.
Then they bet BIG on ai - Watson (winning at Jeopardy!) But it is NOT an LLM, and while it was doing some stuff in medicine, reality was not as good as the PR. Watson medicine was sold off in 2022, "for parts".
https://slate.com/technology/2022/01/ibm-watson-health-failure-artificial-intelligence.html
" there is currently no way to connect that information, to link it to an individual across all the domains in which they get care, and then to develop a holistic picture of who they are, of what their diseases are, of what the best treatments are, and how to ensure that they get the best care at the lowest possible cost. There is no connectivity right now that can do that at scale. ...
the generation of technology that they had was nowhere near ready to accomplish the things that they set out to accomplish and promised that they could accomplish."
As tech changes, the software available changes and what the org gets, or deserves, change. Tweet seen: "AI eats software throughout the world"
There will be some far less creative AI systems that are highly, pedantically accurate - because there's a demand for truth; only the truth is a predictable answer about the past.
Even predictions about the future that come true are not exactly the same kind of "true" as the past which has already happened -- but what exactly did happen, truthfully, often remains in debate, thus somewhat unpredictable.