When I look at the problems in implementing AI in huge enterprise AI systems I am immediately struck by the relaization this is a software implementation problem that has not changed in 50 years.
Back in the 1980's you had the growth of MRP and Job Shop Scheduling programs that were supposed to be revolutionary very much in the way we talk about AI in ERP today. Productivity improvements, faster time to market all the buzzwords phrases.
Getting people to work with the software has always been the problem. Selling stuff without creating a sales order, removing inventory without reporting it, building weldments or machining stuff without a workorder. The list of things just go and on. All of it contributes to innacuracy
Whats the solution. The solution is where the problem is. You only have to take a look at Quantum Lean manfacturing and Shigeo Shingo's original Toyota quality system to see that any system needs to be built around humans.
Software be it AI or MRP or Excel spreadsheets have to support the workflow. In lean manufacturing you hang all the tools to do the job on a wooden job board with outlines drawin around each tool so you immediately know if something is missing.
You have already decided what tools go on the job board by looking at the flow through that workcell. You use the tools to complete one job and you dont do another job until the guy or gal doing the next process step has completed their work making space for you to drop your work with them.
Let's get out of the manufacturing shop floor analogy and take the process of creating an estimate. All the information about pricing and costing, available materials and processes has to be close at hand just like the tools hanging on the job board.
That information needs to be provided by the ERP system's estimating module, and it better be every bit as easy to use as grabbing a tool of the job board. If you need to figure how many panels you will need to use on a big architectural facade job the software should be able to tell you immediately how many standard sizes you need to bring in based on the facades height and width.
The amounts of money spent on modern day ERP implementations are staggering. We do have to realize that the prices reflect the digitization of every record keeping function in the company. Basically every department in a companywhere someone touches a computer has ERP behind.
The Clorox company for example spent $580 million USD on it's latest SAP ERP implementation. It's new system connects finance, sales, forecasting and planning to it's extensive supply chain for all it's products.
@leanmaniac Where the question is the answer should be#bradcairns #woodworking #quantumlean #continiousimprovment #culture #kaizen #manufacturing ♬ original sound - Brad Cairns
Advice to senior management so that ERP and AI works the way it should
@kylercheatham Send this to your #cio before they approve another stupid tech spend 🙃 or drop their 💌 and I will! #digitaltransformation #enterprisetech #executivedysfuntion #techculture #changemanagement #erpsoftware #aistrategy #aitools #automation #workflow #leadership #software ♬ original sound - kylercheatham
A huge gap is being created between building and deploying
Does AI actually make you money in the wild
Need hero engineers to come in and fix all you stupid broken AI, who is going to clean up all this shit.
Getting lot in how FD role demand have increased 1000% year on year
Palantir wildly recognized for seeing whats best for their clients namely billable hours, and pioneering the modern FDE role according to the Wall street journal they are the companies secret weapon.
AI Engineers typically make $288,000 or $300 an hour just to make AI work
Example of Palantir FTE's going to individual farmers and finding out what their weed control process actually looked like, then building an automated software process to support that.
OpenAI also use FTE's a lot Field Training Engineers, being sent out to a call centre and finding out how that business can be helped.
Healthcare framework most critical, Follow the experts around and build the software to support them
Dont need any more AI models they someone they can trust to properly implement them.
An MIT study discovered that 95% of generative AI projects produced no real value. CIO stated that over 45% of AI projects have been abandoned in 2024.
Exxon CFO talks about centralizing data and drastically removing the number of profit and cost centers so the ERP system becomes easier to manage. This seems to be completely the wrong path to take, something that could actually end hurting the business if it's one that cannot function on a centralized model.
Another huge problem with cloud storage is the US Cloud act of 2018 which effectively takes down barriers to government surveillance on any data stored in the cloud especially for cloud customers based outside the USA. European ERP cloud customers are particularly concerned about this.
@kylercheatham The most in demand tech jobs aren’t AI engineers they’re fixing the mess of so-called AI engineers... #frontendengineer #aiengineer #futureofwork #techjobs #devops ♬ original sound - kylercheatham
The City of Birmingham in the UK has spent 7 years and over 140 million pounds sterling on this conversion from SAP to Oracle Fusion ERP and the whole process is still a giant mess.
The city council could not even get their banking reconciliation module to work, leaving them in the dark as to how much money they had available in the ban
Canada a reference market aka guinea pig for Workday ERP based on its heavy regulartory compliance components
https://erp.today/workday-canada-investment/
ERP consultants create training hundreds of training videos for client staff and then cant find them because the system does not provide a searchable database.