Artificial intelligence has significantly reduced the time it takes us to arrive at an initial response. Whether in development, QA, design or support, we can resolve issues or make progress on tasks in far less time than before. This changes the way we approach many tasks and makes evaluating, selecting and improving results increasingly important.
A simple example is code generation. An AI can propose a technically sound implementation in seconds, but to decide whether it fits with the existing architecture, how it will affect maintenance or what implications it may have for the rest of the product, one needs to understand the project. The same applies to a proposed interface, an automation process or a support task: the professional’s knowledge enables them to assess the result within its context and decide how to use it.
At WATA Factory, we work with AI in this way. We harness its ability to generate and process information and carry out tasks, and combine this with the technical and business expertise of our professionals. Experience remains essential for framing problems effectively and making sound decisions based on the results.
The value of AI starts with choosing the right problem
A company may have many processes where artificial intelligence can add value. The first step is to identify which ones have a clear need and where an improvement can make a significant impact.
Understanding how a process works allows us to identify where the workload is actually concentrated and to define a specific objective. The opportunity may lie in facilitating access to certain information, automating a repetitive task, streamlining part of the workflow, or helping staff work with a better understanding of the context. Defining this objective clearly makes it easier to decide what we need from the technology and how we can subsequently verify whether the solution is working.
Expert knowledge provides information that is not always documented
A significant part of a company’s knowledge is captured in specifications, procedures, knowledge bases, business rules or technical documentation. However, over time, professionals also accumulate knowledge that stems directly from their experience.
There are things a person knows because they are documented and others because they have encountered similar situations many times. They know which requirements tend to raise the most questions, which data warrants particular attention, or which decisions may have consequences for other parts of the product. This tacit knowledge is built up through work and helps to interpret situations that, on paper, may seem very similar.
This knowledge is useful for deciding what information the AI should use, which aspects require special attention, and how to organise the context within which it will operate. Documentation provides one part, and professional experience contributes another. Taking both into account allows us to build solutions that are much better suited to the way a company actually operates.
With AI, knowing how to evaluate is just as important as knowing how to create
When we can obtain an initial proposal quickly, we can also compare more alternatives before making a decision. This means that the ability to evaluate results carries significant weight.
The criteria vary depending on the discipline. In development, we can assess maintainability, performance, security or compatibility with the existing architecture. In UX/UI, we can analyse the clarity of the user journey, accessibility or how a solution responds to the task the user needs to perform. In QA, we can prioritise scenarios based on risk, impact or the likelihood of a failure occurring.
Each specialism contributes its own criteria for assessing the quality of a solution. AI can help us to quickly generate and compare alternatives, whilst professional expertise enables us to apply the appropriate criteria to the specific needs and constraints of the project.
How much autonomy to give AI: impact and reversibility
When artificial intelligence is directly involved in a business process, we need to define which actions it can carry out autonomously and at what points we wish to incorporate human validation.
Two variables are particularly useful in making this decision: impact and reversibility. If an action has a limited impact and is easy to correct, we can consider a higher level of autonomy. When the consequences are more significant or it is more difficult to reverse an action, we can introduce additional checks.
A system can, for example, locate documentation or classify information automatically. For other actions, we can incorporate a confirmation step before they are executed, or ensure that certain decisions are forwarded to the relevant professional along with the context that the AI has already gathered.
In this way, we can define precisely where a person intervenes within the process. The human-in-the-loop approach allows professional intervention to be reserved for those decisions where context, experience or responsibility require a specific assessment.
A solution must also allow for alternative paths
Real-world processes involve situations that differ from those initially anticipated. Information may be missing, a new case may arise, there may be discrepancies between two sources, or it may be necessary to consult additional information before proceeding.
These situations can be accounted for at the design stage. Depending on the case, the system may request additional data, consult another source, ask for confirmation or hand the task over to a person.
It is also important to define how this handover takes place. If a professional has to continue the work, it is useful for them to receive the data collected, the sources consulted and the actions carried out up to that point. In this way, they can pick up where the process left off and focus directly on the part requiring their attention.
These cases also provide insights for improving the solution. If a specific situation frequently requires intervention, we can analyse whether there is a rule that should be incorporated, whether any information sources are missing, or whether that type of case requires a specific workflow.
How to tell if an AI solution is adding value
Once a solution is in use, we can check whether it is improving the process for which it was designed. To do this, it is important to measure the outcome we wish to achieve, rather than simply the activity generated by the system.
The metrics will depend on the objective of each project. We can measure how much time or manual work is saved, whether the correct information is retrieved, how many cases successfully complete a process, how much a proposal needs to be adjusted before it is used, or in which situations professional intervention is required.
Usage data also allows us to identify patterns that can help us continue to refine the solution. In this way, the evaluation remains linked to the problem we set out to solve from the outset.
AI changes where an expert adds the most value
The introduction of these tools also influences how professionals organise their work.
A developer can use an initial implementation as a basis and devote more attention to the architecture and the decisions that will shape the product’s evolution. QA can draw on an initial set of scenarios to delve deeper into the risks that require more specific review. A designer can explore different directions more quickly and spend more time understanding which one best meets the user’s needs.
Part of the work can thus focus on architecture, review, resolving complex cases and making decisions that impact the product. These are tasks where a thorough understanding of the discipline and the project is particularly important, and where AI can provide support to ensure decisions are made with greater preparation.
AI + expertise: from a possibility to a real solution
Applying artificial intelligence to a product or process involves making a series of decisions: identifying a specific opportunity, incorporating the right information, establishing criteria for evaluating results, defining the level of autonomy, considering different scenarios and assessing whether the solution is achieving the intended objective.
At WATA Factory, we approach these decisions from a range of specialist perspectives. We have professionals specialising in software development, QA, UX/UI, support and process optimisation, which enables us to analyse each project from the perspectives of technology, the product, the user experience and business needs.
For our clients, this means being able to integrate artificial intelligence with a team that understands both the technology’s capabilities and the environment in which it will be used. Our job is to bridge the gap between these two aspects so that AI addresses a specific need and can be successfully integrated into existing systems and processes.
As tools enable us to generate and process information and carry out tasks more quickly, our professionals can devote more attention to the decisions where their expertise adds the most value. At WATA Factory, we aim to capitalise on precisely this combination: artificial intelligence that expands what we can do, and professionals with the necessary expertise to use it judiciously.
