For business decision-makers · AI in practice
Where to start with AI in your business
Choose a specific situation, see what AI can prepare and decide who will judge whether the result is useful.
A customer request may pass through email, documents and another tool before someone can reply. To explore AI in that work, it helps to understand where information gets lost and which parts require a decision.
At Devmojos, that starting point connects business needs with software and AI. It may lead to a clearly scoped task in existing tools, a new system or a change in the development process. The choice depends on the work that needs to get done.
Figure 1 as text
The request and authorised notes are the inputs. AI organises the available information and prepares a draft. The responsible person checks the information, handles missing details and decides the permitted action. The authorised reply follows the agreed route and is logged. This is an illustrative workflow, not a client project.
Choose the starting point
You need software to deliver a service. A website, an application or a connection between systems requires users, functionality and acceptance criteria to be defined. The Build route lets you discuss that project, its architecture and delivery. A well-defined project can move on to scoping.
You already have tools, but repeat a task across them. It may make sense to explore an AI workflow that prepares content or organises information. The data it can access, the result and the permitted actions need to be agreed. A bounded experiment helps assess usefulness when it still needs to be demonstrated.
You already have a team developing software. The starting point may be the team’s process: context for agents, tests, review or handoffs. AI Engineering includes a process assessment and a scoped adoption experiment. The team decides which changes to keep.
Figure 2 as text
For a repeated task across tools, consider an AI workflow in the existing tools; AI Editorial Desk is running for live publications. For software to be built, consider a project with an agreed scope; Tribeca Festival Lisboa (SIC) is a launched client project. A development team starting AI adoption can consider AI Engineering: a process assessment and scoped experiment. The harness and Devbus are internal engineering practice. These examples have distinct statuses.
Follow a task through to the decision
Consider this illustrative example: a team receives requests by email and prepares a reply using approved information about its services. Use the scenario to identify the decisions involved in a first task.
The request and authorised notes are the inputs. AI can organise the available data, flag what is missing and prepare a draft. A person responsible for the reply confirms the information and decides whether to use it. Access to tools and any sending or system updates are defined within the scope of the work. After the decision, the authorised reply follows the agreed route and is logged.
If an essential detail is missing, the draft should make that gap visible for review. The experiment may allow for a question to the customer or return the request to the team. An apparently complete reply that invents information fails the objective.
To assess this task, compare draft quality, review effort, exceptions and cost with the current process. If the result needs too many corrections, it may be necessary to adjust the sources, reduce the scope or stop the experiment. If it is useful, the next step still needs a decision.
See what already exists
The AI Editorial Desk is a real application for analysing sources and preparing articles with source attribution and clear AI labels. It is running for publications, including Devmojos News & Stories, with controls configured for each publication. It shows a concrete workflow; it does not demonstrate the result another business would obtain.
The Brief to Shortlist addresses a different problem: turning a customer’s request into a property shortlist with information for a property agent to review. It is in use with test agencies; it is not a production service. The agent still decides.
Describe a first task
A useful description identifies what comes in, what needs to be prepared, who accepts the result and what to do when information is missing. Include the tools involved and the difficulty you want to assess. You do not need to begin by choosing a model or automating the whole process.