SOFTWARE ENGINEERING & SYSTEM ARCHITECTURE | 91
Software can do more than process information and execute instructions.
When data, workflows and business logic are connected, systems can automate work, support decisions, anticipate what may happen next and respond to what is happening within the operation.
| 91 introduces intelligence into software systems using automation, decision logic, prediction and AI where each is appropriate.
We don’t start with AI. We start with the process, decision or problem that needs to be improved.
What is someone trying to accomplish? What information is available? Which decisions follow clear rules? What can happen automatically? What could be predicted from the available data? And where does human judgement remain important?
The technology follows from those questions.
Sometimes the answer is straightforward automation. Sometimes it is decision logic. Sometimes historical and current data can be used to predict what may happen next. And sometimes AI adds capabilities that conventional software cannot provide effectively.
Often, the right system combines them.
Repeatable processes can be executed by the system instead of depending on manual actions.
Business rules and operational data can determine what the system should do next.
We combine historical data with other relevant data points to identify patterns and estimate likely outcomes, helping systems anticipate what may happen next and act on that information.
Not every decision should be automated. Systems can organise information, identify what matters and give users better information on which to act.
AI can be introduced where it adds capabilities that rules, automation and conventional software cannot provide effectively.
Where appropriate, the outcome of an action can become input for the next decision, allowing the system to respond to what actually happened.
Intelligence becomes useful when it changes what happens next.
A system might execute a workflow automatically. It might evaluate data and determine the next action. It might predict an outcome and respond before something happens. Or it might present information to a person so they can make a better decision.
For that to work, intelligence cannot operate separately from the rest of the software.
It needs access to the right data and needs to connect with the workflows and systems where the resulting action takes place.
That is why we treat intelligence as part of the wider software system.
There is no reason to use AI for something that can be handled reliably with straightforward automation or decision logic.
Equally, conventional rules are not always enough when a system needs to work with large amounts of information, identify patterns, make predictions or interpret more complex inputs.
We choose the approach based on what the system and its users need to achieve.
The objective is not to add intelligence as a feature. It is to make the software more useful.
Different systems require different forms of intelligence.
Blackcurrant Finance
Blackcurrant Finance uses automation to connect financial processes that would otherwise require coordination between separate systems.
Banking, payments and accounting information can be brought together and used across payment, reconciliation and associated financial workflows, with processes automated where possible.
Here, intelligence is less about an algorithm making a decision and more about the system using connected data to determine and execute what needs to happen next.
Growdt
Growdt uses marketplace data, decision logic and automated execution to make pricing decisions on bol.com.
The platform continuously processes marketplace information and observes actual Buy Box outcomes. Its pricing engine uses those outcomes to determine what price movement to make next.
When Growdt wins the Buy Box, it can test whether a higher selling price is possible. If that change loses the Buy Box, the result becomes information for the next pricing decision.
The system observes, decides, acts and evaluates the outcome as part of the same operational process.
AI creates new possibilities for how software can work with information, language and complex inputs. But it is one option within a much wider system.
For AI to become operationally useful, it still needs the right data, integration with existing software, clearly defined responsibilities and a way for its output to influence what happens next.
Sometimes AI will be central to the solution. Sometimes a rule, calculation, prediction model or automated workflow will do the job better.
| 91 designs automation, decision logic, prediction and AI into software systems based on what the system and its users need to achieve.
We start with the desired outcome and determine what the software needs to do to get there.
