What would you call an average wage?

[Tuppy_Glossop]

Black Box systems can be trained to behave like individual practitioners given the right data set and system design. However, they tend to struggle when the current input data is beyond the training set ranges, and they effectively extrapolate to produce their outputs. There are also non-black box approaches, such as fuzzy-inference systems, which are much better at operating outside the training input data ranges. These can be made to work well (again it is contingent on good data and design) as decision makers or, better still, decision support systems. Which means you don't need such highly skilled staff in all cases. You can save them for the high risk/reward cases.

EDIT: You could even have the fuzzy system decide if each case should be sent to a series of black box models or should go directly to the expert human. If the different BB models don't agree you can then kick it up to the expert anyway.

 

[Murf]

Finance Operations has been outsourced for decades. AI is competing against an ROI of $40/seat/day.

The standard set of rules you mention are also interpreted rather than directly applied. That is why accountancy is a skilled role versus lesser skilled finance operation roles.

Posting a journal is certainly basic but that isn't a job for AI anyway, that is RPA. Calculating the value and where it should go? Yeah no way a black box AI is doing that.

I'm struggling to see the distinction you're making between accountancy roles and finance operations roles. It's all accounting. Differing levels of complexity, different parts of the process maybe but it's still accounting. RPA is useful if you don't have a practical way to eliminate the process or create an ETL that can tackle it. It's the last resort of automation in my view as you're often replicating how a human would carry out a process rather than reworking the process entirely so it can be done automatically via ETL.

The skilled role in the future is the individual that defines how the ETL processes the rules. Or the accountant that provides strategic business advice. Calculating journal amounts is not where the value will be in the future. If you've got an existing ETL stack and in house developers then $40 a day for a process that is slower and prone to human error is not appealing.

And over time, core software will get better at handling most of this anyway. Think Making Tax Digital or Payroll Real Time Information. HMRC submissions now calculated automatically and submitted through digital gateways. There's only one direction of travel, especially where the primary focus is the application of standard rules.

 

[Greebo]

I'm struggling to see the distinction you're making between accountancy roles and finance operations roles. It's all accounting. Differing levels of complexity, different parts of the process maybe but it's still accounting. RPA is useful if you don't have a practical way to eliminate the process or create an ETL that can tackle it. It's the last resort of automation in my view as you're often replicating how a human would carry out a process rather than reworking the process entirely so it can be done automatically via ETL.

The skilled role in the future is the individual that defines how the ETL processes the rules. Or the accountant that provides strategic business advice. Calculating journal amounts is not where the value will be in the future. If you've got an existing ETL stack and in house developers then $40 a day for a process that is slower and prone to human error is not appealing.

And over time, core software will get better at handling most of this anyway. Think Making Tax Digital or Payroll Real Time Information. HMRC submissions now calculated automatically and submitted through digital gateways. There's only one direction of travel, especially where the primary focus is the application of standard rules.

Indeed. Accountants wont disappear entirely but like many other job sectors, the total number needed will be great reduced and their job role will change. A bit like automated trucks. They can quite easily do the main haulage from warehouse A with robot fork trucks loading them to warehouse B where they are unloaded but multi drops and variable drops will be human drivers for time.

 

[dlockers]

I'm struggling to see the distinction you're making between accountancy roles and finance operations roles. It's all accounting. Differing levels of complexity, different parts of the process maybe but it's still accounting. RPA is useful if you don't have a practical way to eliminate the process or create an ETL that can tackle it. It's the last resort of automation in my view as you're often replicating how a human would carry out a process rather than reworking the process entirely so it can be done automatically via ETL.

The skilled role in the future is the individual that defines how the ETL processes the rules. Or the accountant that provides strategic business advice. Calculating journal amounts is not where the value will be in the future. If you've got an existing ETL stack and in house developers then $40 a day for a process that is slower and prone to human error is not appealing.

And over time, core software will get better at handling most of this anyway. Think Making Tax Digital or Payroll Real Time Information. HMRC submissions now calculated automatically and submitted through digital gateways. There's only one direction of travel, especially where the primary focus is the application of standard rules.

I think we're on two entirely different wave lengths my man! I've never heard of ETL in this context, ever - and if you are suggesting to put business logic in the middleware then I suggest you stay out of architecture

 

[Murf]

I think we're on two entirely different wave lengths my man! I've never heard of ETL in this context, ever - and if you are suggesting to put business logic in the middleware then I suggest you stay out of architecture

I'm suggesting using a whole host of technologies to eliminate humans from standard, repeatable processes. Business logic will more and more be catered for within core software moving forward, but in the interim there are plenty of tools that can apply business logic more accurately, and more efficiently than humans. Especially where they're just moving data around in spreadsheets. This isn't architecture, it's process.

I mean, at an absolutely basic level (I'm assuming you're an accountant here BTW), think how prepayment journals used to be posted manually, where now the majority of prepayment journals are automated as the right data is captured at source on the purchase order so that the core accounting system can generate and post the journals automatically. What was once a human using 'judgement' to apply rules on prepayments to invoices is now automated. Then just start applying the same type of logic to every other process. Some will be easier than others to solve systematically. And what you'll be left with is the accountants that adapted and pivoted to a greater focus on providing strategic advice, developing expertise in niche &, complex areas, developing their design thinking skills etc.

 

[platinum87]

Accounting is about decisions. AI can't take decisions.

You are confusing software which can be automated with AI.